idnits 2.17.1 draft-ietf-rift-rift-06.txt: Checking boilerplate required by RFC 5378 and the IETF Trust (see https://trustee.ietf.org/license-info): ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/1id-guidelines.txt: ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/checklist : ---------------------------------------------------------------------------- ** There are 134 instances of too long lines in the document, the longest one being 116 characters in excess of 72. == There are 1 instance of lines with multicast IPv4 addresses in the document. If these are generic example addresses, they should be changed to use the 233.252.0.x range defined in RFC 5771 Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year == Line 4598 has weird spacing: '...berType undef...' == Line 4602 has weird spacing: '...velType top_...' == Line 4604 has weird spacing: '...itsType defau...' == Line 4606 has weird spacing: '...velType leaf...' == Line 4607 has weird spacing: '...velType defa...' == (27 more instances...) -- The document date (June 23, 2019) is 1768 days in the past. Is this intentional? -- Found something which looks like a code comment -- if you have code sections in the document, please surround them with '' and '' lines. Checking references for intended status: Proposed Standard ---------------------------------------------------------------------------- (See RFCs 3967 and 4897 for information about using normative references to lower-maturity documents in RFCs) == Missing Reference: 'A' is mentioned on line 268, but not defined == Missing Reference: 'B' is mentioned on line 268, but not defined == Missing Reference: 'C' is mentioned on line 278, but not defined == Missing Reference: 'D' is mentioned on line 278, but not defined == Missing Reference: 'E' is mentioned on line 271, but not defined == Missing Reference: 'F' is mentioned on line 271, but not defined == Missing Reference: 'NH' is mentioned on line 2220, but not defined == Missing Reference: 'P' is mentioned on line 2415, but not defined == Missing Reference: 'RFC5880' is mentioned on line 4562, but not defined -- Possible downref: Non-RFC (?) normative reference: ref. 'ISO10589' ** Obsolete normative reference: RFC 7752 (Obsoleted by RFC 9552) Summary: 2 errors (**), 0 flaws (~~), 17 warnings (==), 3 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 RIFT Working Group The RIFT Authors 3 Internet-Draft June 23, 2019 4 Intended status: Standards Track 5 Expires: December 25, 2019 7 RIFT: Routing in Fat Trees 8 draft-ietf-rift-rift-06 10 Abstract 12 This document outlines a specialized, dynamic routing protocol for 13 Clos and fat-tree network topologies. The protocol (1) deals with 14 fully automated construction of fat-tree topologies based on 15 detection of links, (2) minimizes the amount of routing state held at 16 each level, (3) automatically prunes and load balances topology 17 flooding exchanges over a sufficient subset of links, (4) supports 18 automatic disaggregation of prefixes on link and node failures to 19 prevent black-holing and suboptimal routing, (5) allows traffic 20 steering and re-routing policies, (6) allows loop-free non-ECMP 21 forwarding, (7) automatically re-balances traffic towards the spines 22 based on bandwidth available and finally (8) provides mechanisms to 23 synchronize a limited key-value data-store that can be used after 24 protocol convergence to e.g. bootstrap higher levels of 25 functionality on nodes. 27 Status of This Memo 29 This Internet-Draft is submitted in full conformance with the 30 provisions of BCP 78 and BCP 79. 32 Internet-Drafts are working documents of the Internet Engineering 33 Task Force (IETF). Note that other groups may also distribute 34 working documents as Internet-Drafts. The list of current Internet- 35 Drafts is at https://datatracker.ietf.org/drafts/current/. 37 Internet-Drafts are draft documents valid for a maximum of six months 38 and may be updated, replaced, or obsoleted by other documents at any 39 time. It is inappropriate to use Internet-Drafts as reference 40 material or to cite them other than as "work in progress." 42 This Internet-Draft will expire on December 25, 2019. 44 Copyright Notice 46 Copyright (c) 2019 IETF Trust and the persons identified as the 47 document authors. All rights reserved. 49 This document is subject to BCP 78 and the IETF Trust's Legal 50 Provisions Relating to IETF Documents 51 (https://trustee.ietf.org/license-info) in effect on the date of 52 publication of this document. Please review these documents 53 carefully, as they describe your rights and restrictions with respect 54 to this document. Code Components extracted from this document must 55 include Simplified BSD License text as described in Section 4.e of 56 the Trust Legal Provisions and are provided without warranty as 57 described in the Simplified BSD License. 59 Table of Contents 61 1. Authors . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 62 2. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 5 63 2.1. Requirements Language . . . . . . . . . . . . . . . . . . 7 64 3. Reference Frame . . . . . . . . . . . . . . . . . . . . . . . 7 65 3.1. Terminology . . . . . . . . . . . . . . . . . . . . . . . 7 66 3.2. Topology . . . . . . . . . . . . . . . . . . . . . . . . 10 67 4. Requirement Considerations . . . . . . . . . . . . . . . . . 12 68 5. RIFT: Routing in Fat Trees . . . . . . . . . . . . . . . . . 15 69 5.1. Overview . . . . . . . . . . . . . . . . . . . . . . . . 16 70 5.1.1. Properties . . . . . . . . . . . . . . . . . . . . . 16 71 5.1.2. Generalized Topology View . . . . . . . . . . . . . . 16 72 5.1.3. Fallen Leaf Problem . . . . . . . . . . . . . . . . . 26 73 5.1.4. Discovering Fallen Leaves . . . . . . . . . . . . . . 28 74 5.1.5. Addressing the Fallen Leaves Problem . . . . . . . . 29 75 5.2. Specification . . . . . . . . . . . . . . . . . . . . . . 30 76 5.2.1. Transport . . . . . . . . . . . . . . . . . . . . . . 30 77 5.2.2. Link (Neighbor) Discovery (LIE Exchange) . . . . . . 31 78 5.2.3. Topology Exchange (TIE Exchange) . . . . . . . . . . 33 79 5.2.3.1. Topology Information Elements . . . . . . . . . . 33 80 5.2.3.2. South- and Northbound Representation . . . . . . 33 81 5.2.3.3. Flooding . . . . . . . . . . . . . . . . . . . . 36 82 5.2.3.4. TIE Flooding Scopes . . . . . . . . . . . . . . . 36 83 5.2.3.5. 'Flood Only Node TIEs' Bit . . . . . . . . . . . 39 84 5.2.3.6. Initial and Periodic Database Synchronization . . 40 85 5.2.3.7. Purging and Roll-Overs . . . . . . . . . . . . . 40 86 5.2.3.8. Southbound Default Route Origination . . . . . . 41 87 5.2.3.9. Northbound TIE Flooding Reduction . . . . . . . . 41 88 5.2.3.10. Special Considerations . . . . . . . . . . . . . 46 89 5.2.4. Reachability Computation . . . . . . . . . . . . . . 47 90 5.2.4.1. Northbound SPF . . . . . . . . . . . . . . . . . 47 91 5.2.4.2. Southbound SPF . . . . . . . . . . . . . . . . . 48 92 5.2.4.3. East-West Forwarding Within a Level . . . . . . . 48 93 5.2.5. Automatic Disaggregation on Link & Node Failures . . 48 94 5.2.5.1. Positive, Non-transitive Disaggregation . . . . . 48 95 5.2.5.2. Negative, Transitive Disaggregation for Fallen 96 Leafs . . . . . . . . . . . . . . . . . . . . . . 52 98 5.2.6. Attaching Prefixes . . . . . . . . . . . . . . . . . 54 99 5.2.7. Optional Zero Touch Provisioning (ZTP) . . . . . . . 63 100 5.2.7.1. Terminology . . . . . . . . . . . . . . . . . . . 64 101 5.2.7.2. Automatic SystemID Selection . . . . . . . . . . 65 102 5.2.7.3. Generic Fabric Example . . . . . . . . . . . . . 66 103 5.2.7.4. Level Determination Procedure . . . . . . . . . . 67 104 5.2.7.5. Resulting Topologies . . . . . . . . . . . . . . 68 105 5.2.8. Stability Considerations . . . . . . . . . . . . . . 70 106 5.3. Further Mechanisms . . . . . . . . . . . . . . . . . . . 70 107 5.3.1. Overload Bit . . . . . . . . . . . . . . . . . . . . 70 108 5.3.2. Optimized Route Computation on Leafs . . . . . . . . 70 109 5.3.3. Mobility . . . . . . . . . . . . . . . . . . . . . . 70 110 5.3.3.1. Clock Comparison . . . . . . . . . . . . . . . . 72 111 5.3.3.2. Interaction between Time Stamps and Sequence 112 Counters . . . . . . . . . . . . . . . . . . . . 72 113 5.3.3.3. Anycast vs. Unicast . . . . . . . . . . . . . . . 73 114 5.3.3.4. Overlays and Signaling . . . . . . . . . . . . . 73 115 5.3.4. Key/Value Store . . . . . . . . . . . . . . . . . . . 74 116 5.3.4.1. Southbound . . . . . . . . . . . . . . . . . . . 74 117 5.3.4.2. Northbound . . . . . . . . . . . . . . . . . . . 74 118 5.3.5. Interactions with BFD . . . . . . . . . . . . . . . . 74 119 5.3.6. Fabric Bandwidth Balancing . . . . . . . . . . . . . 75 120 5.3.6.1. Northbound Direction . . . . . . . . . . . . . . 75 121 5.3.6.2. Southbound Direction . . . . . . . . . . . . . . 77 122 5.3.7. Label Binding . . . . . . . . . . . . . . . . . . . . 78 123 5.3.8. Segment Routing Support with RIFT . . . . . . . . . . 78 124 5.3.8.1. Global Segment Identifiers Assignment . . . . . . 78 125 5.3.8.2. Distribution of Topology Information . . . . . . 78 126 5.3.9. Leaf to Leaf Procedures . . . . . . . . . . . . . . . 79 127 5.3.10. Address Family and Multi Topology Considerations . . 79 128 5.3.11. Reachability of Internal Nodes in the Fabric . . . . 79 129 5.3.12. One-Hop Healing of Levels with East-West Links . . . 80 130 5.4. Security . . . . . . . . . . . . . . . . . . . . . . . . 80 131 5.4.1. Security Model . . . . . . . . . . . . . . . . . . . 80 132 5.4.2. Security Mechanisms . . . . . . . . . . . . . . . . . 82 133 5.4.3. Security Envelope . . . . . . . . . . . . . . . . . . 82 134 5.4.4. Weak Nonces . . . . . . . . . . . . . . . . . . . . . 85 135 5.4.5. Lifetime . . . . . . . . . . . . . . . . . . . . . . 86 136 5.4.6. Key Management . . . . . . . . . . . . . . . . . . . 86 137 5.4.7. Security Association Changes . . . . . . . . . . . . 86 138 6. Examples . . . . . . . . . . . . . . . . . . . . . . . . . . 87 139 6.1. Normal Operation . . . . . . . . . . . . . . . . . . . . 87 140 6.2. Leaf Link Failure . . . . . . . . . . . . . . . . . . . . 88 141 6.3. Partitioned Fabric . . . . . . . . . . . . . . . . . . . 89 142 6.4. Northbound Partitioned Router and Optional East-West 143 Links . . . . . . . . . . . . . . . . . . . . . . . . . . 91 144 6.5. Multi-Plane Fabric and Negative Disaggregation . . . . . 92 145 7. Implementation and Operation: Further Details . . . . . . . . 92 146 7.1. Considerations for Leaf-Only Implementation . . . . . . . 92 147 7.2. Considerations for Spine Implementation . . . . . . . . . 93 148 7.3. Adaptations to Other Proposed Data Center Topologies . . 93 149 7.4. Originating Non-Default Route Southbound . . . . . . . . 94 150 8. Security Considerations . . . . . . . . . . . . . . . . . . . 95 151 8.1. General . . . . . . . . . . . . . . . . . . . . . . . . . 95 152 8.2. ZTP . . . . . . . . . . . . . . . . . . . . . . . . . . . 95 153 8.3. Lifetime . . . . . . . . . . . . . . . . . . . . . . . . 95 154 8.4. Packet Number . . . . . . . . . . . . . . . . . . . . . . 95 155 8.5. Outer Fingerprint Attacks . . . . . . . . . . . . . . . . 96 156 8.6. TIE Origin Fingerprint DoS Attacks . . . . . . . . . . . 96 157 9. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 96 158 10. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . 96 159 11. References . . . . . . . . . . . . . . . . . . . . . . . . . 97 160 11.1. Normative References . . . . . . . . . . . . . . . . . . 97 161 11.2. Informative References . . . . . . . . . . . . . . . . . 99 162 Appendix A. Sequence Number Binary Arithmetic . . . . . . . . . 102 163 Appendix B. Information Elements Schema . . . . . . . . . . . . 103 164 B.1. common.thrift . . . . . . . . . . . . . . . . . . . . . . 103 165 B.2. encoding.thrift . . . . . . . . . . . . . . . . . . . . . 109 166 Appendix C. Finite State Machines and Precise Operational 167 Specifications . . . . . . . . . . . . . . . . . . . 117 168 C.1. LIE FSM . . . . . . . . . . . . . . . . . . . . . . . . . 117 169 C.2. ZTP FSM . . . . . . . . . . . . . . . . . . . . . . . . . 123 170 C.3. Flooding Procedures . . . . . . . . . . . . . . . . . . . 131 171 C.3.1. FloodState Structure per Adjacency . . . . . . . . . 132 172 C.3.2. TIDEs . . . . . . . . . . . . . . . . . . . . . . . . 134 173 C.3.2.1. TIDE Generation . . . . . . . . . . . . . . . . . 134 174 C.3.2.2. TIDE Processing . . . . . . . . . . . . . . . . . 135 175 C.3.3. TIREs . . . . . . . . . . . . . . . . . . . . . . . . 136 176 C.3.3.1. TIRE Generation . . . . . . . . . . . . . . . . . 136 177 C.3.3.2. TIRE Processing . . . . . . . . . . . . . . . . . 136 178 C.3.4. TIEs Processing on Flood State Adjacency . . . . . . 137 179 C.3.5. TIEs Processing When LSDB Received Newer Version on 180 Other Adjacencies . . . . . . . . . . . . . . . . . . 138 181 C.3.6. Sending TIEs . . . . . . . . . . . . . . . . . . . . 138 182 Appendix D. Constants . . . . . . . . . . . . . . . . . . . . . 138 183 D.1. Configurable Protocol Constants . . . . . . . . . . . . . 138 184 Appendix E. TODO . . . . . . . . . . . . . . . . . . . . . . . . 140 185 Author's Address . . . . . . . . . . . . . . . . . . . . . . . . 140 187 1. Authors 189 This work is a product of a growing list of individuals. 191 Tony Przygienda, Ed | Alankar Sharma | Pascal Thubert 192 Juniper Networks | Comcast | Cisco 194 Bruno Rijsman | Ilya Vershkov | Dmitry Afanasiev 195 Individual | Mellanox | Yandex 197 Don Fedyk | Alia Atlas | John Drake 198 Individual | Individual | Juniper 200 Table 1: RIFT Authors 202 2. Introduction 204 Clos [CLOS] and Fat-Tree [FATTREE] topologies have gained prominence 205 in today's networking, primarily as result of the paradigm shift 206 towards a centralized data-center based architecture that is poised 207 to deliver a majority of computation and storage services in the 208 future. Today's current routing protocols were geared towards a 209 network with an irregular topology and low degree of connectivity 210 originally but given they were the only available options, 211 consequently several attempts to apply those protocols to Clos have 212 been made. Most successfully BGP [RFC4271] [RFC7938] has been 213 extended to this purpose, not as much due to its inherent suitability 214 but rather because the perceived capability to easily modify BGP and 215 the immanent difficulties with link-state [DIJKSTRA] based protocols 216 to optimize topology exchange and converge quickly in large scale 217 densely meshed topologies. The incumbent protocols precondition 218 normally extensive configuration or provisioning during bring up and 219 re-dimensioning which is only viable for a set of organizations with 220 according networking operation skills and budgets. For the majority 221 of data center consumers a preferable protocol would be one that 222 auto-configures itself and deals with failures and misconfigurations 223 with a minimum of human intervention only. Such a solution would 224 allow local IP fabric bandwidth to be consumed in a standardized 225 component fashion, i.e. provision it much faster and operate it at 226 much lower costs, much like compute or storage is consumed today. 228 In looking at the problem through the lens of data center 229 requirements, an optimal approach does not seem however to be a 230 simple modification of either a link-state (distributed computation) 231 or distance-vector (diffused computation) approach but rather a 232 mixture of both, colloquially best described as "link-state towards 233 the spine" and "distance vector towards the leafs". In other words, 234 "bottom" levels are flooding their link-state information in the 235 "northern" direction while each node generates under normal 236 conditions a default route and floods it in the "southern" direction. 237 This type of protocol allows naturally for highly desirable 238 aggregation. Alas, such aggregation could blackhole traffic in cases 239 of misconfiguration or while failures are being resolved or even 240 cause partial network partitioning and this has to be addressed. The 241 approach RIFT takes is described in Section 5.2.5 and is basically 242 based on automatic, sufficient disaggregation of prefixes. 244 For the visually oriented reader, Figure 1 presents a first level 245 simplified view of the resulting information and routes on a RIFT 246 fabric. The top of the fabric is holding in its link-state database 247 the nodes below it and the routes to them. In the second row of the 248 database table we indicate that partial information of other nodes in 249 the same level is available as well. The details of how this is 250 achieved will be postponed for the moment. When we look at the 251 "bottom" of the fabric, the leafs, we see that the topology is 252 basically empty and they only hold a load balanced default route to 253 the next level. 255 The balance of this document details the resulting protocol and fills 256 in the missing details. 258 . [A,B,C,D] 259 . [E] 260 . +-----+ +-----+ 261 . | E | | F | A/32 @ [C,D] 262 . +-+-+-+ +-+-+-+ B/32 @ [C,D] 263 . | | | | C/32 @ C 264 . | | +-----+ | D/32 @ D 265 . | | | | 266 . | +------+ | 267 . | | | | 268 . [A,B] +-+---+ | | +---+-+ [A,B] 269 . [D] | C +--+ +-+ D | [C] 270 . +-+-+-+ +-+-+-+ 271 . 0/0 @ [E,F] | | | | 0/0 @ [E,F] 272 . A/32 @ A | | +-----+ | A/32 @ A 273 . B/32 @ B | | | | B/32 @ B 274 . | +------+ | 275 . | | | | 276 . +-+---+ | | +---+-+ 277 . | A +--+ +-+ B | 278 . 0/0 @ [C,D] +-----+ +-----+ 0/0 @ [C,D] 280 Figure 1: RIFT information distribution 282 2.1. Requirements Language 284 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 285 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 286 document are to be interpreted as described in RFC 2119 [RFC2119]. 288 3. Reference Frame 290 3.1. Terminology 292 This section presents the terminology used in this document. It is 293 assumed that the reader is thoroughly familiar with the terms and 294 concepts used in OSPF [RFC2328] and IS-IS [ISO10589-Second-Edition], 295 [ISO10589] as well as the according graph theoretical concepts of 296 shortest path first (SPF) [DIJKSTRA] computation and directed acyclic 297 graphs (DAG). 299 Level: Clos and Fat Tree networks are topologically partially 300 ordered graphs and 'level' denotes the set of nodes at the same 301 height in such a network, where the bottom level (leaf) is the 302 level with lowest value. A node has links to nodes one level down 303 and/or one level up. Under some circumstances, a node may have 304 links to nodes at the same level. As footnote: Clos terminology 305 uses often the concept of "stage" but due to the folded nature of 306 the Fat Tree we do not use it to prevent misunderstandings. 308 Superspine/Aggregation or Spine/Edge Levels: Traditional names in 309 5-stages folded Clos for Level 2, 1 and 0 respectively. Level 0 310 is often called leaf as well. We normalize this language to talk 311 about leafs, spines and top-of-fabric (ToF). 313 Point of Delivery (PoD): A self-contained vertical slice or subset 314 of a Clos or Fat Tree network containing normally only level 0 and 315 level 1 nodes. A node in a PoD communicates with nodes in other 316 PoDs via the Top-of-Fabric. We number PoDs to distinguish them 317 and use PoD #0 to denote "undefined" PoD. 319 Top of PoD (ToP): The set of nodes that provide intra-PoD 320 communication and have northbound adjacencies outside of the PoD, 321 i.e. are at the "top" of the PoD. 323 Top of Fabric (ToF): The set of nodes that provide inter-PoD 324 communication and have no northbound adjacencies, i.e. are at the 325 "very top" of the fabric. ToF nodes do not belong to any PoD and 326 are assigned "undefined" PoD value to indicate the equivalent of 327 "any" PoD. 329 Spine: Any nodes north of leafs and south of top-of-fabric nodes. 330 Multiple layers of spines in a PoD are possible. 332 Leaf: A node without southbound adjacencies. Its level is 0 (except 333 cases where it is deriving its level via ZTP and is running 334 without LEAF_ONLY which will be explained in Section 5.2.7). 336 Top-of-fabric Plane or Partition: In large fabrics top-of-fabric 337 switches may not have enough ports to aggregate all switches south 338 of them and with that, the ToF is 'split' into multiple 339 independent planes. Introduction and Section 5.1.2 explains the 340 concept in more detail. A plane is subset of ToF nodes that see 341 each other through south reflection or E-W links. 343 Radix: A radix of a switch is basically number of switching ports it 344 provides. It's sometimes called fanout as well. 346 North Radix: Ports cabled northbound to higher level nodes. 348 South Radix: Ports cabled southbound to lower level nodes. 350 South/Southbound and North/Northbound (Direction): When describing 351 protocol elements and procedures, we will be using in different 352 situations the directionality of the compass. I.e., 'south' or 353 'southbound' mean moving towards the bottom of the Clos or Fat 354 Tree network and 'north' and 'northbound' mean moving towards the 355 top of the Clos or Fat Tree network. 357 Northbound Link: A link to a node one level up or in other words, 358 one level further north. 360 Southbound Link: A link to a node one level down or in other words, 361 one level further south. 363 East-West Link: A link between two nodes at the same level. East- 364 West links are normally not part of Clos or "fat-tree" topologies. 366 Leaf shortcuts (L2L): East-West links at leaf level will need to be 367 differentiated from East-West links at other levels. 369 Southbound representation: Subset of topology information sent 370 towards a lower level. 372 South Reflection: Often abbreviated just as "reflection" it defines 373 a mechanism where South Node TIEs are "reflected" back up north to 374 allow nodes in same level without E-W links to "see" each other. 376 TIE: This is an acronym for a "Topology Information Element". TIEs 377 are exchanged between RIFT nodes to describe parts of a network 378 such as links and address prefixes. A TIE can be thought of as 379 largely equivalent to ISIS LSPs or OSPF LSA. We will talk about 380 N-TIEs when talking about TIEs in the northbound representation 381 and S-TIEs for the southbound equivalent. 383 Node TIE: This is an acronym for a "Node Topology Information 384 Element", largely equivalent to OSPF Router LSA, i.e. it contains 385 all adjacencies the node discovered and information about node 386 itself. 388 Prefix TIE: This is an acronym for a "Prefix Topology Information 389 Element" and it contains all prefixes directly attached to this 390 node in case of a N-TIE and in case of S-TIE the necessary default 391 the node passes southbound. 393 Key Value TIE: A S-TIE that is carrying a set of key value pairs 394 [DYNAMO]. It can be used to distribute information in the 395 southbound direction within the protocol. 397 TIDE: Topology Information Description Element, equivalent to CSNP 398 in ISIS. 400 TIRE: Topology Information Request Element, equivalent to PSNP in 401 ISIS. It can both confirm received and request missing TIEs. 403 De-aggregation/Disaggregation: Process in which a node decides to 404 advertise certain prefixes it received in N-TIEs to prevent black- 405 holing and suboptimal routing upon link failures. 407 LIE: This is an acronym for a "Link Information Element", largely 408 equivalent to HELLOs in IGPs and exchanged over all the links 409 between systems running RIFT to form adjacencies. 411 Flood Repeater (FR): A node can designate one or more northbound 412 neighbor nodes to be flood repeaters. The flood repeaters are 413 responsible for flooding northbound TIEs further north. They are 414 similar to MPR in OSLR. The document sometimes calls them flood 415 leaders as well. 417 Bandwidth Adjusted Distance (BAD): This is an acronym for Bandwidth 418 Adjusted Distance. Each RIFT node calculates the amount of 419 northbound bandwidth available towards a node compared to other 420 nodes at the same level and modifies the default route distance 421 accordingly to allow for the lower level to adjust their load 422 balancing towards spines. 424 Overloaded: Applies to a node advertising `overload` attribute as 425 set. The semantics closely follow the meaning of the same 426 attribute in [ISO10589-Second-Edition]. 428 Interface: A layer 3 entity over which RIFT control packets are 429 exchanged. 431 Adjacency: RIFT tries to form a unique adjacency over an interface 432 and exchange local configuration and necessary ZTP information. 434 Neighbor: Once a three way adjacency has been formed a neighborship 435 relationship contains the neighbor's properties. Multiple 436 adjacencies can be formed to a neighbor via parallel interfaces 437 but such adjacencies are NOT sharing a neighbor structure. Saying 438 "neighbor" is thus equivalent to saying "a three way adjacency". 440 Cost: The term signifies the weighted distance between two 441 neighbors. 443 Distance: Sum of costs (bound by infinite distance) between two 444 nodes. 446 Metric: Without going deeper into the proper differentiation, a 447 metric is equivalent to distance. 449 3.2. Topology 450 . +--------+ +--------+ ^ N 451 . |ToF 21| |ToF 22| | 452 .Level 2 ++-+--+-++ ++-+--+-++ <-*-> E/W 453 . | | | | | | | | | 454 . P111/2| |P121 | | | | S v 455 . ^ ^ ^ ^ | | | | 456 . | | | | | | | | 457 . +--------------+ | +-----------+ | | | +---------------+ 458 . | | | | | | | | 459 . South +-----------------------------+ | | ^ 460 . | | | | | | | All TIEs 461 . 0/0 0/0 0/0 +-----------------------------+ | 462 . v v v | | | | | 463 . | | +-+ +<-0/0----------+ | | 464 . | | | | | | | | 465 .+-+----++ optional +-+----++ ++----+-+ ++-----++ 466 .| | E/W link | | | | | | 467 .|Spin111+----------+Spin112| |Spin121| |Spin122| 468 .+-+---+-+ ++----+-+ +-+---+-+ ++---+--+ 469 . | | | South | | | | 470 . | +---0/0--->-----+ 0/0 | +----------------+ | 471 . 0/0 | | | | | | | 472 . | +---<-0/0-----+ | v | +--------------+ | | 473 . v | | | | | | | 474 .+-+---+-+ +--+--+-+ +-+---+-+ +---+-+-+ 475 .| | (L2L) | | | | Level 0 | | 476 .|Leaf111~~~~~~~~~~~~Leaf112| |Leaf121| |Leaf122| 477 .+-+-----+ +-+---+-+ +--+--+-+ +-+-----+ 478 . + + \ / + + 479 . Prefix111 Prefix112 \ / Prefix121 Prefix122 480 . multi-homed 481 . Prefix 482 .+---------- Pod 1 ---------+ +---------- Pod 2 ---------+ 484 Figure 2: A three level spine-and-leaf topology 485 .+--------+ +--------+ +--------+ +--------+ 486 .|ToF A1| |ToF B1| |ToF B2| |ToF A2| 487 .++-+-----+ ++-+-----+ ++-+-----+ ++-+-----+ 488 . | | | | | | | | 489 . | | | | | +---------------+ 490 . | | | | | | | | 491 . | | | +-------------------------+ | 492 . | | | | | | | | 493 . | +-----------------------+ | | | | 494 . | | | | | | | | 495 . | | +---------+ | +---------+ | | 496 . | | | | | | | | 497 . | +---------------------------------+ | | 498 . | | | | | | | | 499 .++-+-----+ ++-+-----+ +--+-+---+ +----+-+-+ 500 .|Spine111| |Spine112| |Spine121| |Spine122| 501 .+-+---+--+ ++----+--+ +-+---+--+ ++---+---+ 502 . | | | | | | | | 503 . | +--------+ | | +--------+ | 504 . | | | | | | | | 505 . | -------+ | | | +------+ | | 506 . | | | | | | | | 507 .+-+---+-+ +--+--+-+ +-+---+-+ +---+-+-+ 508 .|Leaf111| |Leaf112| |Leaf121| |Leaf122| 509 .+-------+ +-------+ +-------+ +-------+ 511 Figure 3: Topology with multiple planes 513 We will use topology in Figure 2 (called commonly a fat tree/network 514 in modern IP fabric considerations [VAHDAT08] as homonym to the 515 original definition of the term [FATTREE]) in all further 516 considerations. This figure depicts a generic "single plane fat- 517 tree" and the concepts explained using three levels apply by 518 induction to further levels and higher degrees of connectivity. 519 Further, this document will deal also with designs that provide only 520 sparser connectivity and "partitioned spines" as shown in Figure 3 521 and explained further in Section 5.1.2. 523 4. Requirement Considerations 525 [RFC7938] gives the original set of requirements augmented here based 526 upon recent experience in the operation of fat-tree networks. 528 REQ1: The control protocol should discover the physical links 529 automatically and be able to detect cabling that violates 530 fat-tree topology constraints. It must react accordingly to 531 such mis-cabling attempts, at a minimum preventing 532 adjacencies between nodes from being formed and traffic from 533 being forwarded on those mis-cabled links. E.g. connecting 534 a leaf to a spine at level 2 should be detected and ideally 535 prevented. 537 REQ2: A node without any configuration beside default values 538 should come up at the correct level in any PoD it is 539 introduced into. Optionally, it must be possible to 540 configure nodes to restrict their participation to the 541 PoD(s) targeted at any level. 543 REQ3: Optionally, the protocol should allow to provision IP 544 fabrics where the individual switches carry no configuration 545 information and are all deriving their level from a "seed". 546 Observe that this requirement may collide with the desire to 547 detect cabling misconfiguration and with that only one of 548 the requirements can be fully met in a chosen configuration 549 mode. 551 REQ4: The solution should allow for minimum size routing 552 information base and forwarding tables at leaf level for 553 speed, cost and simplicity reasons. Holding excessive 554 amount of information away from leaf nodes simplifies 555 operation and lowers cost of the underlay and allows to 556 scale and introduce proper multi-homing down to the server 557 level. The routing solution should allow for easy 558 instantiation of multiple routing planes. Coupled with 559 mobility defined in Paragraph 17 this should allow for 560 "light-weight" overlays on an IP fabric with e.g. native 561 IPv6 mobility support. 563 REQ5: Very high degree of ECMP must be supported. Maximum ECMP is 564 currently understood as the most efficient routing approach 565 to maximize the throughput of switching fabrics 566 [MAKSIC2013]. 568 REQ6: Non equal cost anycast must be supported to allow for easy 569 and robust multi-homing of services without regressing to 570 careful balancing of link costs. 572 REQ7: Traffic engineering should be allowed by modification of 573 prefixes and/or their next-hops. 575 REQ8: The solution should allow for access to link states of the 576 whole topology to enable efficient support for modern 577 control architectures like SPRING [RFC7855] or PCE 578 [RFC4655]. 580 REQ9: The solution should easily accommodate opaque data to be 581 carried throughout the topology to subsets of nodes. This 582 can be used for many purposes, one of them being a key-value 583 store that allows bootstrapping of nodes based right at the 584 time of topology discovery. Another use is distributing MAC 585 to L3 address binding from the leafs up north in case of 586 e.g. DHCP. 588 REQ10: Nodes should be taken out and introduced into production 589 with minimum wait-times and minimum of "shaking" of the 590 network, i.e. radius of propagation (often called "blast 591 radius") of changed information should be as small as 592 feasible. 594 REQ11: The protocol should allow for maximum aggregation of carried 595 routing information while at the same time automatically de- 596 aggregating the prefixes to prevent black-holing in case of 597 failures. The de-aggregation should support maximum 598 possible ECMP/N-ECMP remaining after failure. 600 REQ12: Reducing the scope of communication needed throughout the 601 network on link and state failure, as well as reducing 602 advertisements of repeating or idiomatic information in 603 stable state is highly desirable since it leads to better 604 stability and faster convergence behavior. 606 REQ13: Once a packet traverses a link in a "southbound" direction, 607 it must not take any further "northbound" steps along its 608 path to delivery to its destination under normal, i.e. 609 fully converged, conditions. Taking a path through the 610 spine in cases where a shorter path is available is highly 611 undesirable. 613 REQ14: Parallel links between same set of nodes must be 614 distinguishable for SPF, failure and traffic engineering 615 purposes. 617 REQ15: The protocol must not rely on interfaces having discernible 618 unique addresses, i.e. it must operate in presence of 619 unnumbered links (even parallel ones) or links of a single 620 node having same addresses. 622 REQ16: It would be desirable to achieve fast re-balancing of flows 623 when links, especially towards the spines are lost or 624 provisioned without regressing to per flow traffic 625 engineering which introduces significant amount of 626 complexity while possibly not being reactive enough to 627 account for short-lived flows. 629 REQ17: The control plane should be able to unambiguously determine 630 the current point of attachment (which port on which leaf 631 node) of a prefix, even in a context of fast mobility, e.g., 632 when the prefix is a host address on a wireless node that 1) 633 may associate to any of multiple access points (APs) that 634 are attached to different ports on a same leaf node or to 635 different leaf nodes, and 2) may move and reassociate 636 several times to a different access point within a sub- 637 second period. 639 REQ18: The protocol should provide security mechanisms that allow 640 to restrict nodes, especially leafs without proper 641 credentials from forming three-way adjacencies. 643 Following list represents possible requirements and requirements 644 under discussion: 646 PEND1: Supporting anything but point-to-point links is a non- 647 requirement. Questions remain: for connecting to the 648 leaves, is there a case where multipoint is desirable? One 649 could still model it as point-to-point links; it seems there 650 is no need for anything more than a NBMA-type construct. 652 PEND2: What is the maximum scale of number leaf prefixes we need to 653 carry. 500'000 seems plenty even if we deploy RIFT down to 654 servers as leafs. 656 Finally, following are the non-requirements: 658 NONREQ1: Broadcast media support is unnecessary. However, 659 miscabling leading to multiple nodes on a broadcast 660 segment must be operationally easily recognizable and 661 detectable while not taxing the protocol excessively. 663 NONREQ2: Purging link state elements is unnecessary given its 664 fragility and complexity and today's large memory size on 665 even modest switches and routers. 667 NONREQ3: Special support for layer 3 multi-hop adjacencies is not 668 part of the protocol specification. Such support can be 669 easily provided by using tunneling technologies the same 670 way IGPs today are solving the problem. 672 5. RIFT: Routing in Fat Trees 674 Derived from the above requirements we present a detailed outline of 675 a protocol optimized for Routing in Fat Trees (RIFT) that in most 676 abstract terms has many properties of a modified link-state protocol 678 [RFC2328][ISO10589-Second-Edition] when "pointing north" and path- 679 vector [RFC4271] protocol when "pointing south". While this is an 680 unusual combination, it does quite naturally exhibit the desirable 681 properties we seek. 683 5.1. Overview 685 5.1.1. Properties 687 The most singular property of RIFT is that it floods flat link-state 688 information northbound only so that each level obtains the full 689 topology of levels south of it. That information is never flooded 690 East-West (we'll talk about exceptions later) or back South again. 691 In the southbound direction the protocol operates like a "fully 692 summarizing, unidirectional" path vector protocol or rather a 693 distance vector with implicit split horizon whereas the information 694 propagates one hop south and is 're-advertised' by nodes at next 695 lower level, normally just the default route. However, RIFT uses 696 flooding in the southern direction as well to avoid the necessity to 697 build an update per adjacency. We omit describing the East-West 698 direction out for the moment. 700 Those information flow constraints create not only an anisotropic 701 protocol (i.e. the information is not distributed "evenly" or 702 "clumped" but summarized along the N-S gradient) but also a "smooth" 703 information propagation where nodes do not receive the same 704 information from multiple fronts which would force them to perform a 705 diffused computation to tie-break the same reachability information 706 arriving on arbitrary links and ultimately force hop-by-hop 707 forwarding on shortest-paths only. The application of those 708 principle lead to RIFT having moreover the highly desirable 709 properties of being loop-free and guaranteeing valley-free forwarding 710 behavior. 712 To account for the "northern" and the "southern" information split 713 the link state database is partitioned into "north representation" 714 and "south representation" TIEs, whereas in simplest terms the N-TIEs 715 contain a link state topology description of lower levels and and 716 S-TIEs carry simply default routes. This oversimplified view will be 717 refined gradually in following sections while introducing protocol 718 procedures aimed to fulfill the described requirements. 720 5.1.2. Generalized Topology View 722 This section will dwell on the topologies addresses by RIFT including 723 multi plane fabrics and their related implications. Readers that are 724 only interested in single plane designs, i.e. all top-of-fabric nodes 725 being topologically equal and initially connected to all the switches 726 at the level below them can skip this section and resulting 727 Section 5.2.5.2 as well. 729 Given the difficulty of visualizing multi plane design which are 730 effectively multi-dimensional switching matrices we will introduce a 731 methodology allowing us to visualize the connectivity in a two- 732 dimensional document and leverage the fact that we are dealing 733 basically with crossbar fabrics stacked on top of each other where 734 ports also align "on top of each other" in a regular fashion. 736 The typical topology for which RIFT is defined is built of a number P 737 of PoDs, connected together by a number S of spine nodes. A PoD node 738 has a number of ports called Radix, with half of them (K=Radix/2) 739 used to connect host devices from the south, and half to connect to 740 interleaved PoD Top-Level switches to the north. Ratio K can be 741 chosen differently without loss of generality when port speeds differ 742 or fabric is oversubscribed but K=R/2 allows for more readable 743 representation whereby there are as many ports facing north as south 744 on any intermediate node. We represent a node hence in a schematic 745 fashion with ports "sticking out" to its north and south rather than 746 by the usual real-world front faceplate designs of the day. 748 Figure 4 provides a view of a leaf node as seen from the north, i.e. 749 showing ports that connect northbound and for lack of a better 750 symbol, we have chosen to use the "HH" symbol as ASCII visualisation 751 of a RJ45 jack. In that example, K_LEAF is chosen to be 6 ports. 752 Observe that the number of PoDs is not related to Radix unless the 753 ToF Nodes are constrained to be the same as the PoD nodes in a 754 particular deployment. 756 Top view 757 +----+ 758 | | 759 | HH | e.g., Radix = 12, K_LEAF = 6 760 | | 761 | HH | 762 | | ------------------------- 763 | HH ------- Physical Port (Ethernet) ----+ 764 | | ------------------------- | 765 | HH | | 766 | | | 767 | HH | | 768 | | | 769 | HH | | 770 | | | 771 +----+ | 773 || || || || || || || 774 +----+ +------------------------------------------------+ 775 | | | | 776 +----+ +------------------------------------------------+ 777 || || || || || || || 778 Side views 780 Figure 4: A Leaf Node, K_LEAF=6 782 The Radix of a node on top of a PoD may be different than that of the 783 leaf node, though more often than not a same type of node is used for 784 both, effectively forming a square (K*K). In the general case, we 785 could have switches with K_TOP southern ports on nodes at the top of 786 the PoD that is not necessarily the same as K_LEAF; for instance, in 787 the representations below, we pick a K_LEAF of 6 and a K_TOP of 8. 788 In order to form a crossbar, we need K_TOP Leaf Nodes as illustrated 789 in Figure 5. 791 +----+ +----+ +----+ +----+ +----+ +----+ +----+ +----+ 792 | | | | | | | | | | | | | | | | 793 | HH | | HH | | HH | | HH | | HH | | HH | | HH | | HH | 794 | | | | | | | | | | | | | | | | 795 | HH | | HH | | HH | | HH | | HH | | HH | | HH | | HH | 796 | | | | | | | | | | | | | | | | 797 | HH | | HH | | HH | | HH | | HH | | HH | | HH | | HH | 798 | | | | | | | | | | | | | | | | 799 | HH | | HH | | HH | | HH | | HH | | HH | | HH | | HH | 800 | | | | | | | | | | | | | | | | 801 | HH | | HH | | HH | | HH | | HH | | HH | | HH | | HH | 802 | | | | | | | | | | | | | | | | 803 | HH | | HH | | HH | | HH | | HH | | HH | | HH | | HH | 804 | | | | | | | | | | | | | | | | 805 +----+ +----+ +----+ +----+ +----+ +----+ +----+ +----+ 807 Figure 5: Southern View of a PoD, K_TOP=8 809 The K_TOP Leaf Nodes are fully interconnected with the K_LEAF PoD-top 810 nodes, providing a connectivity that can be represented as a crossbar 811 as seen from the north and illustrated in Figure 6. The result is 812 that, in the absence of a breakage, a packet entering the PoD from 813 North on any port can be routed to any port on the south of the PoD 814 and vice versa. 816 E<-*->W 818 +----+ +----+ +----+ +----+ +----+ +----+ +----+ +----+ 819 | | | | | | | | | | | | | | | | 820 +----------------------------------------------------------------+ 821 | HH HH HH HH HH HH HH HH | 822 +----------------------------------------------------------------+ 823 +----------------------------------------------------------------+ 824 | HH HH HH HH HH HH HH HH | 825 +----------------------------------------------------------------+ 826 +----------------------------------------------------------------+ 827 | HH HH HH HH HH HH HH HH | 828 +----------------------------------------------------------------+ 829 +----------------------------------------------------------------+ 830 | HH HH HH HH HH HH HH HH | 831 +----------------------------------------------------------------+ 832 +----------------------------------------------------------------+ 833 | HH HH HH HH HH HH HH HH |<-+ 834 +----------------------------------------------------------------+ | 835 +----------------------------------------------------------------+ | 836 | HH HH HH HH HH HH HH HH | | 837 +----------------------------------------------------------------+ | 838 | | | | | | | | | | | | | | | | | 839 +----+ +----+ +----+ +----+ +----+ +----+ +----+ +----+ | 840 ^ | 841 | | 842 | ---------- --------------------- | 843 +----- Leaf Node PoD top Node (Spine) --+ 844 ---------- --------------------- 846 Figure 6: Northern View of a PoD's Spines, K_TOP=8 848 Side views of this PoD is illustrated in Figure 7 and Figure 8. 850 Connecting to Spine 852 || || || || || || || || 853 +----------------------------------------------------------------+ N 854 | PoD top Node seen sideways | ^ 855 +----------------------------------------------------------------+ | 856 || || || || || || || || * 857 +----+ +----+ +----+ +----+ +----+ +----+ +----+ +----+ | 858 | | | | | | | | | | | | | | | | v 859 +----+ +----+ +----+ +----+ +----+ +----+ +----+ +----+ S 860 || || || || || || || || 862 Connecting to Client nodes 864 Figure 7: Side View of a PoD, K_TOP=8, K_LEAF=6 866 Connecting to Spine 868 || || || || || || 869 +----+ +----+ +----+ +----+ +----+ +----+ N 870 | | | | | | | | | | | PoD top Nodes ^ 871 +----+ +----+ +----+ +----+ +----+ +----+ | 872 || || || || || || * 873 +------------------------------------------------+ | 874 | Leaf seen sideways | v 875 +------------------------------------------------+ S 876 || || || || || || 878 Connecting to Client nodes 880 Figure 8: Other side View of a PoD, K_TOP=8, K_LEAF=6, 90o turn in 881 E-W Plane 883 Note that a resulting PoD can be abstracted as a bigger node with a 884 number K of K_POD= K_TOP * K_LEAF, and the design can recurse. 886 It is critical at this junction that the concept and the picture of 887 those "crossed crossbars" is clear before progressing further, 888 otherwise following considerations will be difficult to comprehend. 890 Further, the PoDs are interconnected with one another through a Top- 891 of-Fabric at the very top or the north edge of the fabric. The 892 resulting ToF is NOT partitioned if and only if (IIF) every PoD top 893 level node (spine) is connected to every ToF Node. This is also 894 referred to as a single plane configuration. In order to reach a 895 1::1 connectivity ratio between the ToF and the Leaves, it results 896 that there are K_TOP ToF nodes, because each port of a ToP node 897 connects to a different ToF node, and K_LEAF ToP nodes for the same 898 reason. Consequently, it takes (P * K_LEAF) ports on a ToF node to 899 connect to each of the K_LEAF ToP nodes of the P PoDs, as illustrated 900 in Figure 9. 902 [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] <-----+ 903 | | | | | | | | | 904 [=================================] | ----------- 905 | | | | | | | | +----- Top-of-Fabric 906 [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] +----- Node -------+ 907 | ----------- | 908 | v 909 +-+ +-+ +-+ +-+ +-+ +-+ +-+ +-+ <-----+ +-+ 910 | | | | | | | | | | | | | | | | | | 911 [ |H| |H| |H| |H| |H| |H| |H| |H| ] | | 912 [ |H| |H| |H| |H| |H| |H| |H| |H| ] ------------------------- | | 913 [ |H| |H| |H| |H| |H| |H| |H| |H<--- Physical Port (Ethernet) | | 914 [ |H| |H| |H| |H| |H| |H| |H| |H| ] ------------------------- | | 915 [ |H| |H| |H| |H| |H| |H| |H| |H| ] | | 916 [ |H| |H| |H| |H| |H| |H| |H| |H| ] | | 917 | | | | | | | | | | | | | | | | | | 918 [ |H| |H| |H| |H| |H| |H| |H| |H| ] | | 919 [ |H| |H| |H| |H| |H| |H| |H| |H| ] -------------- | | 920 [ |H| |H| |H| |H| |H| |H| |H| |H| ] <--- PoD top level | | 921 [ |H| |H| |H| |H| |H| |H| |H| |H| ] node (Spine) ---+ | | 922 [ |H| |H| |H| |H| |H| |H| |H| |H| ] -------------- | | | 923 [ |H| |H| |H| |H| |H| |H| |H| |H| ] | | | 924 | | | | | | | | | | | | | | | | -+ +- +-+ v | | 925 [ |H| |H| |H| |H| |H| |H| |H| |H| ] | | --| |--[ ]--| | 926 [ |H| |H| |H| |H| |H| |H| |H| |H| ] | ----- | --| |--[ ]--| | 927 [ |H| |H| |H| |H| |H| |H| |H| |H| ] +--- PoD ---+ --| |--[ ]--| | 928 [ |H| |H| |H| |H| |H| |H| |H| |H| ] | ----- | --| |--[ ]--| | 929 [ |H| |H| |H| |H| |H| |H| |H| |H| ] | | --| |--[ ]--| | 930 [ |H| |H| |H| |H| |H| |H| |H| |H| ] | | --| |--[ ]--| | 931 | | | | | | | | | | | | | | | | -+ +- +-+ | | 932 +-+ +-+ +-+ +-+ +-+ +-+ +-+ +-+ +-+ 934 Figure 9: Fabric Spines and TOFs in Single Plane Design, 3 PoDs 936 The top view can be collapsed into a third dimension where the hidden 937 depth index is representing the PoD number. So we can show one PoD 938 as a class of PoDs and hence save one dimension in our 939 representation. The Spine Node expands in the depth and the vertical 940 dimensions whereas the PoD top level Nodes are constrained in 941 horizontal dimension. A port in the 2-D representation represents 942 effectively the class of all the ports at the same position in all 943 the PoDs that are projected in its position along the depth axis. 944 This is shown in Figure 10. 946 / / / / / / / / / / / / / / / / 947 / / / / / / / / / / / / / / / / 948 / / / / / / / / / / / / / / / / 949 / / / / / / / / / / / / / / / / ] 950 +-+ +-+ +-+ +-+ +-+ +-+ +-+ +-+ ]] 951 | | | | | | | | | | | | | | | | ] --------------------------- 952 [ |H| |H| |H| |H| |H| |H| |H| |H| ] <-- PoD top level node (Spine) 953 [ |H| |H| |H| |H| |H| |H| |H| |H| ] --------------------------- 954 [ |H| |H| |H| |H| |H| |H| |H| |H| ]]]] 955 [ |H| |H| |H| |H| |H| |H| |H| |H| ]]] ^^ 956 [ |H| |H| |H| |H| |H| |H| |H| |H| ]] // PoDs 957 [ |H| |H| |H| |H| |H| |H| |H| |H| ] // (in depth) 958 | |/| |/| |/| |/| |/| |/| |/| |/ // 959 +-+ +-+ +-+/+-+/+-+ +-+ +-+ +-+ // 960 ^ 961 | ---------------- 962 +----- Top-of-Fabric Node 963 ---------------- 965 Figure 10: Collapsed Northern View of a Fabric for Any Number of PoDs 967 This type of deployment introduces a "single plane limit" where the 968 bound is the available radix of the ToF nodes, which limits (P * 969 K_LEAF). Nevertheless, a distinct advantage of a connected or 970 unpartitioned Top-of-Fabric is that all failures can be resolved by 971 simple, non-transitive, positive disaggregation described in 972 Section 5.2.5.1 that propagates only within one level of the fabric. 973 In other words unpartitoned ToF nodes can always reach nodes below or 974 withdraw the routes from PoDs they cannot reach unambiguously. To be 975 more precise, all failures which still allow all the ToF nodes to see 976 each other via south reflection as explained in Section 5.2.5. 978 In order to scale beyond the "single plane limit", the Top-of-Fabric 979 can be partitioned by a number N of identically wired planes, N being 980 an integer divider of K_LEAF. The 1::1 ratio and the desired 981 symmetry are still served, this time with (K_TOP * N) ToF nodes, each 982 of (P * K_LEAF / N) ports. N=1 represents a non-partitioned Spine 983 and N=K_LEAF is a maximally partitioned Spine. Further, if R is any 984 divisor of K_LEAF, then (N=K_LEAF/R) is a feasible number of planes 985 and R a redundancy factor. If proves convenient for deployments to 986 use a radix for the leaf nodes that is a power of 2 so they can pick 987 a number of planes that is a lower power of 2. The example in 988 Figure 11 splits the Spine in 2 planes with a redundancy factor R=3, 989 meaning that there are 3 non-intersecting paths between any leaf node 990 and any ToF node. A ToF node must have in this case at least 3*P 991 ports, and be directly connected to 3 of the 6 PoD-ToP nodes (spines) 992 in each PoD. 994 +----+ +----+ +----+ +----+ +----+ +----+ +----+ +----+ 995 +-| |--| |--| |--| |--| |--| |--| |--| |-+ 996 | | HH | | HH | | HH | | HH | | HH | | HH | | HH | | HH | | 997 +-| |--| |--| |--| |--| |--| |--| |--| |-+ 998 +-| |--| |--| |--| |--| |--| |--| |--| |-+ 999 | | HH | | HH | | HH | | HH | | HH | | HH | | HH | | HH | | 1000 +-| |--| |--| |--| |--| |--| |--| |--| |-+ 1001 +-| |--| |--| |--| |--| |--| |--| |--| |-+ 1002 | | HH | | HH | | HH | | HH | | HH | | HH | | HH | | HH | | 1003 +-| |--| |--| |--| |--| |--| |--| |--| |-+ 1004 +----+ +----+ +----+ +----+ +----+ +----+ +----+ +----+ 1006 Plane 1 1007 ----------- . ------------ . ------------ . ------------ . -------- 1008 Plane 2 1010 +----+ +----+ +----+ +----+ +----+ +----+ +----+ +----+ 1011 +-| |--| |--| |--| |--| |--| |--| |--| |-+ 1012 | | HH | | HH | | HH | | HH | | HH | | HH | | HH | | HH | | 1013 +-| |--| |--| |--| |--| |--| |--| |--| |-+ 1014 +-| |--| |--| |--| |--| |--| |--| |--| |-+ 1015 | | HH | | HH | | HH | | HH | | HH | | HH | | HH | | HH | | 1016 +-| |--| |--| |--| |--| |--| |--| |--| |-+ 1017 +-| |--| |--| |--| |--| |--| |--| |--| |-+ 1018 | | HH | | HH | | HH | | HH | | HH | | HH | | HH | | HH | | 1019 +-| |--| |--| |--| |--| |--| |--| |--| |-+ 1020 +----+ +----+ +----+ +----+ +----+ +----+ +----+ +----+ 1021 ^ 1022 | 1023 | ---------------- 1024 +----- Top-of-Fabric node 1025 "across" depth 1026 ---------------- 1028 Figure 11: Northern View of a Multi-Plane ToF Level, K_LEAF=6, N=2 1030 At the extreme end of the spectrum, it is even possible to fully 1031 partition the spine with N = K_LEAF and R=1, while maintaining 1032 connectivity between each leaf node and each Top-of-Fabric node. In 1033 that case the ToF node connects to a single Port per PoD, so it 1034 appears as a single port in the projected view represented in 1035 Figure 12 and the number of ports required on the Spine Node is more 1036 or equal to P, the number of PoDs. 1038 Plane 1 1039 +----+ +----+ +----+ +----+ +----+ +----+ +----+ +----+ -+ 1040 +-| |--| |--| |--| |--| |--| |--| |--| |-+ | 1041 | | HH | | HH | | HH | | HH | | HH | | HH | | HH | | HH | | | 1042 +-| |--| |--| |--| |--| |--| |--| |--| |-+ | 1043 +----+ +----+ +----+ +----+ +----+ +----+ +----+ +----+ | 1044 ----------- . ------------ . ------------ . ------------ . -------- | 1045 +----+ +----+ +----+ +----+ +----+ +----+ +----+ +----+ | 1046 +-| |--| |--| |--| |--| |--| |--| |--| |-+ | 1047 | | HH | | HH | | HH | | HH | | HH | | HH | | HH | | HH | | | 1048 +-| |--| |--| |--| |--| |--| |--| |--| |-+ | 1049 +----+ +----+ +----+ +----+ +----+ +----+ +----+ +----+ | 1050 ----------- . ------------ . ------------ . ------------ . -------- | 1051 +----+ +----+ +----+ +----+ +----+ +----+ +----+ +----+ | 1052 +-| |--| |--| |--| |--| |--| |--| |--| |-+ | 1053 | | HH | | HH | | HH | | HH | | HH | | HH | | HH | | HH | | | 1054 +-| |--| |--| |--| |--| |--| |--| |--| |-+ | 1055 +----+ +----+ +----+ +----+ +----+ +----+ +----+ +----+ | 1056 ----------- . ------------ . ------------ . ------------ . -------- +<-+ 1057 +----+ +----+ +----+ +----+ +----+ +----+ +----+ +----+ | | 1058 +-| |--| |--| |--| |--| |--| |--| |--| |-+ | | 1059 | | HH | | HH | | HH | | HH | | HH | | HH | | HH | | HH | | | | 1060 +-| |--| |--| |--| |--| |--| |--| |--| |-+ | | 1061 +----+ +----+ +----+ +----+ +----+ +----+ +----+ +----+ | | 1062 ----------- . ------------ . ------------ . ------------ . -------- | | 1063 +----+ +----+ +----+ +----+ +----+ +----+ +----+ +----+ | | 1064 +-| |--| |--| |--| |--| |--| |--| |--| |-+ | | 1065 | | HH | | HH | | HH | | HH | | HH | | HH | | HH | | HH | | | | 1066 +-| |--| |--| |--| |--| |--| |--| |--| |-+ | | 1067 +----+ +----+ +----+ +----+ +----+ +----+ +----+ +----+ | | 1068 ----------- . ------------ . ------------ . ------------ . -------- | | 1069 +----+ +----+ +----+ +----+ +----+ +----+ +----+ +----+ | | 1070 +-| |--| |--| |--| |--| |--| |--| |--| |-+ | | 1071 | | HH | | HH | | HH | | HH | | HH | | HH | | HH | | HH | | | | 1072 +-| |--| |--| |--| |--| |--| |--| |--| |-+ | | 1073 +----+ +----+ +----+ +----+ +----+ +----+ +----+ +----+ -+ | 1074 Plane 6 ^ | 1075 | | 1076 | ---------------- -------------- | 1077 +----- ToF Node Class of PoDs ---+ 1078 ---------------- ------------- 1080 Figure 12: Northern View of a Maximally Partitioned ToF Level, R=1 1082 5.1.3. Fallen Leaf Problem 1084 As mentioned earlier, RIFT exhibits an anisotropic behavior tailored 1085 for fabrics with a North / South orientation and a high level of 1086 interleaving paths. A non-partitioned fabric makes a total loss of 1087 connectivity between a Top-of-Fabric node at the north and a leaf 1088 node at the south a very rare but yet possible occasion that is fully 1089 healed by positive disaggregation described in Section 5.2.5.1. In 1090 large fabrics or fabrics built from switches with low radix, the ToF 1091 ends often being partioned in planes which makes the occurrence of 1092 having a given leaf being only reachable from a subset of the ToF 1093 nodes more likely to happen. This makes some further considerations 1094 necessary. 1096 We define a "Fallen Leaf" as a leaf that can be reached by only a 1097 subset of Top-of-Fabric nodes but cannot be reached by all due to 1098 missing connectivity. If R is the redundancy factor, then it takes 1099 at least R breakages to reach a "Fallen Leaf" situation. 1101 In a general manner, the mechanism of non-transitive positive 1102 disaggregation is sufficient when the disaggregating ToF nodes 1103 collectively connect to all the ToP nodes in the broken plane. This 1104 happens in the following case: 1106 If the breakage is the last northern link from a ToP node to a ToF 1107 node going down, then the fallen leaf problem affects only The ToF 1108 node, and the connectivity to all the nodes in the PoD is lost 1109 from that ToF node. This can be observed by other ToF nodes 1110 within the plane where the ToP node is located and positively 1111 disaggregated within that plane. 1113 On the other hand, there is a need to disaggregate the routes to 1114 Fallen Leaves in a transitive fashion all the way to the other leaves 1115 in the following cases: 1117 If the breakage is the last northern link from a Leaf node within 1118 a plane - there is only one such link in a maximally partitioned 1119 fabric - that goes down, then connectivity to all unicast prefixes 1120 attached to the Leaf node is lost within the plane where the link 1121 is located. Southern Reflection by a Leaf Node - e.g., between 1122 ToP nodes if the PoD has only 2 levels - happens in between 1123 planes, allowing the ToP nodes to detect the problem within the 1124 PoD where it occurs and positively disaggregate. The breakage can 1125 be observed by the ToF nodes in the same plane through the 1126 flooding of N-TIEs from the ToP nodes, but the ToF nodes need to 1127 be aware of all the affected prefixes for the negative 1128 disaggregation to be fully effective. The problem can also be 1129 observed by the ToF nodes in the other planes through the flooding 1130 of N-TIEs from the affected Leaf nodes, together with non-node 1131 N-TIEs which indicate the affected prefixes. To be effective in 1132 that case, the positive disaggregation must reach down to the 1133 nodes that make the plane selection, which are typically the 1134 ingress Leaf nodes, and the information is not useful for routing 1135 in the intermediate levels. 1137 If the breakage is a ToP node in a maximally partitioned fabric - 1138 in which case it is the only ToP node serving that plane in that 1139 PoD - that goes down, then the connectivity to all the nodes in 1140 the PoD is lost within the plane where the ToP node is located - 1141 all leaves fall. Since the Southern Reflection between the ToF 1142 nodes happens only within a plane, ToF nodes in other planes 1143 cannot discover the case of fallen leaves in a different plane, 1144 and cannot determine beyond their local plane whether a Leaf node 1145 that was initially reachable has become unreachable. As above, 1146 the breakage can be observed by the ToF nodes in the plane where 1147 the breakage happened, and then again, the ToF nodes in the plane 1148 need to be aware of all the affected prefixes for the negative 1149 disaggregation to be fully effective. The problem can also be 1150 observed by the ToF nodes in the other planes through the flooding 1151 of N-TIEs from the affected Leaf nodes, if there are only 3 levels 1152 and the ToP nodes are directly connected to the Leaf nodes, and 1153 then again it can only be effective it is propagated transitively 1154 to the Leaf, and useless above that level. 1156 For the sake of easy comprehension let us roll the abstractions back 1157 to a simple example and observe that in Figure 3 the loss of link 1158 Spine 122 to Leaf 122 will make Leaf 122 a fallen leaf for Top-of- 1159 Fabric plane B. Worse, if the cabling was never present in first 1160 place, plane B will not even be able to know that such a fallen leaf 1161 exists. Hence partitioning without further treatment results in two 1162 grave problems: 1164 o Leaf111 trying to route to Leaf122 MUST choose Spine 111 in plane 1165 A as its next hop since plane B will inevitably blackhole the 1166 packet when forwarding using default routes or do excessive bow 1167 tie'ing, i.e. this information must be in its routing table. 1169 o any kind of "flooding" or distance vector trying to deal with the 1170 problem by distributing host routes will be able to converge only 1171 using paths through leafs, i.e. the flooding of information on 1172 Leaf122 will go up to Top-of-Fabric A and then "loopback" over 1173 other leafs to ToF B leading in extreme cases to traffic for 1174 Leaf122 when presented to plane B taking an "inverted fabric" path 1175 where leafs start to serve as TOFs. 1177 5.1.4. Discovering Fallen Leaves 1179 As we illustrate later and without further proof here, to deal with 1180 fallen leafs in multi-plane designs RIFT requires all the ToF nodes 1181 to share the same topology database. This happens naturally in 1182 single plane design but needs additional considerations in multi- 1183 plane fabrics. To satisfy this RIFT in multi-plane designs relies at 1184 the ToF Level on ring interconnection of switches in multiple planes. 1185 Other solutions are possible but they either need more cabling or end 1186 up having much longer flooding path and/or single points of failure. 1188 In more detail, by reserving two ports on each Top-of-Fabric node it 1189 is possible to connect them together in an interplane bi-directional 1190 ring as illustrated in Figure 13 (where we show a bi-directional ring 1191 connecting switches across planes). The rings will exchange full 1192 topology information between planes and with that allow consequently 1193 by the means of transitive, negative disaggregation described in 1194 Section 5.2.5.2 to efficiently fix any possible fallen leaf scenario. 1195 Somewhat as a side-effect, the exchange of information fulfills the 1196 requirement to present full view of the fabric topology at the Top- 1197 of-Fabric level without the need to collate it from multiple points 1198 by additional complexity of technologies like [RFC7752]. 1200 +----+ +----+ +----+ +----+ +----+ +----+ +--------+ 1201 | | | | | | | | | | | | | | 1202 | | | | | | | | 1203 +-o--+ +-o--+ +-o--+ +-o--+ +-o--+ +-o--+ +-o--+ | 1204 +-| |--| |--| |--| |--| |--| |--| |-+ | 1205 | | HH | | HH | | HH | | HH | | HH | | HH | | HH | | | Plane A 1206 +-| |--| |--| |--| |--| |--| |--| |-+ | 1207 +-o--+ +-o--+ +-o--+ +-o--+ +-o--+ +-o--+ +-o--+ | 1208 | | | | | | | | 1209 +-o--+ +-o--+ +-o--+ +-o--+ +-o--+ +-o--+ +-o--+ | 1210 +-| |--| |--| |--| |--| |--| |--| |-+ | 1211 | | HH | | HH | | HH | | HH | | HH | | HH | | HH | | | Plane B 1212 +-| |--| |--| |--| |--| |--| |--| |-+ | 1213 +-o--+ +-o--+ +-o--+ +-o--+ +-o--+ +-o--+ +-o--+ | 1214 | | | | | | | | 1215 ... | 1216 | | | | | | | | 1217 +-o--+ +-o--+ +-o--+ +-o--+ +-o--+ +-o--+ +-o--+ | 1218 +-| |--| |--| |--| |--| |--| |--| |-+ | 1219 | | HH | | HH | | HH | | HH | | HH | | HH | | HH | | | Plane X 1220 +-| |--| |--| |--| |--| |--| |--| |-+ | 1221 +-o--+ +-o--+ +-o--+ +-o--+ +-o--+ +-o--+ +-o--+ | 1222 | | | | | | | | 1223 | | | | | | | | | | | | | | 1224 +----+ +----+ +----+ +----+ +----+ +----+ +--------+ 1226 Figure 13: Connecting Top-of-Fabric Nodes Across Planes by Two Rings 1228 5.1.5. Addressing the Fallen Leaves Problem 1230 One consequence of the Fallen Leaf problem is that some prefixes 1231 attached to the fallen leaf become unreachable from some of the ToF 1232 nodes. RIFT proposes two methods to address this issue, the positive 1233 and the negative disaggregation. Both methods flood S-TIEs to 1234 advertise the impacted prefix(es). 1236 When used for the operation of disaggregation, a positive S-TIE, as 1237 usual, indicates reachability to a prefix of given length and all 1238 addresses subsumed by it. In contrast, a negative route 1239 advertisement indicates that the origin cannot route to the 1240 advertised prefix. 1242 The positive disaggregation is originated by a router that can still 1243 reach the advertised prefix, and the operation is not transitive, 1244 meaning that the receiver does not generate its own flooding south as 1245 a consequence of receiving positive disaggregation advertisements 1246 from an higher level node. The effect of a positive disaggregation 1247 is that the traffic to the impacted prefix will follow the prefix 1248 longest match and will be limited to the northbound routers that 1249 advertised the more specific route. 1251 In contrast, the negative disaggregation is transitive, and is 1252 propagated south when all the possible routes northwards are barred. 1253 A negative route advertisement is only actionable when the negative 1254 prefix is aggregated by a positive route advertisement for a shorter 1255 prefix. In that case, the negative advertisement carves an exception 1256 to the positive route in the routing table (one could think of 1257 "punching a hole"), making the positive prefix reachable through the 1258 originator with the special consideration of the negative prefix 1259 removing certain next hop neighbors. 1261 When the ToF is not partitioned, the collective southern flooding of 1262 the positive disaggregation by the ToF nodes that can still reach the 1263 impacted prefix is in general enough to cover all the switches at the 1264 next level south, typically the ToP nodes. If all those switches are 1265 aware of the disaggregation, they collectively create a ceiling that 1266 intercepts all the traffic north and forwards it to the ToF nodes 1267 that advertised the more specific route. In that case, the positive 1268 disaggregation alone is sufficient to solve the fallen leaf problem. 1270 On the other hand, when the fabric is partitioned in planes, the 1271 positive disaggregation from ToF nodes in different planes do not 1272 reach the ToP switches in the affected plane and cannot solve the 1273 fallen leaves problem. In other words, a breakage in a plane can 1274 only be solved in that plane. Also, the selection of the plane for a 1275 packet typically occurs at the leaf level and the disaggregation must 1276 be transitive and reach all the leaves. In that case, the negative 1277 disaggregation is necessary. The details on the RIFT approach to 1278 deal with fallen leafs in an optimal way is specified in 1279 Section 5.2.5.2. 1281 5.2. Specification 1283 5.2.1. Transport 1285 All packet formats are defined in Thrift models in Appendix B. 1287 The serialized model is carried in an envelope within a UDP frame 1288 that provides security and allows validation/modification of several 1289 important fields without de-serialization for performance and 1290 security reasons. 1292 5.2.2. Link (Neighbor) Discovery (LIE Exchange) 1294 LIE exchange happens over well-known administratively locally scoped 1295 and configured or otherwise well-known IPv4 multicast address 1296 [RFC2365] or link-local multicast scope [RFC4291] for IPv6 [RFC8200] 1297 using a configured or otherwise a well-known destination UDP port 1298 defined in Appendix D.1. LIEs SHOULD be sent with a TTL of 1 to 1299 prevent RIFT information reaching beyond a single L3 next-hop in the 1300 topology. LIEs SHOULD be sent with network control precedence. 1301 Originating port of the LIE has no further significance other than 1302 identifying the origination point. LIEs are exchanged over all links 1303 running RIFT. An implementation MAY listen and send LIEs on IPv4 1304 and/or IPv6 multicast addresses. LIEs on same link are considered 1305 part of the same negotiation independent on the address family they 1306 arrive on. Observe further that the LIE source address may not 1307 identify the peer uniquely in unnumbered or link-local address cases 1308 so the response transmission MUST occur over the same interface the 1309 LIEs have been received on. A node CAN use any of the adjacency's 1310 source addresses it saw in LIEs on the specific interface during 1311 adjacency formation to send TIEs. That implies that an 1312 implementation MUST be ready to accept TIEs on all addresses it used 1313 as source of LIE frames. 1315 Observe further that the protocol does NOT support selective 1316 disabling of address families or any local address changes in three 1317 way state, i.e. if a link has entered three way IPv4 and/or IPv6 with 1318 a neighbor on an adjacency and it wants to stop supporting one of the 1319 families or change any of its local addresses, it has to tear down 1320 and rebuild the adjacency. It also has to remove any information it 1321 stored about adjacency's' LIE source addresses seen. 1323 All RIFT routers MUST support IPv4 forwarding and MAY support IPv6 1324 forwarding. A three way adjacency over IPv6 addresses implies 1325 support for IPv4 forwarding. 1327 Unless Section 5.2.7 is used, each node is provisioned with the level 1328 at which it is operating and its PoD (or otherwise a default level 1329 and "undefined" PoD are assumed; meaning that leafs do not need to be 1330 configured at all if initial configuration values are all left at 0). 1331 Nodes in the spine are configured with "any" PoD which has the same 1332 value "undefined" PoD hence we will talk about "undefined/any" PoD. 1333 This information is propagated in the LIEs exchanged. 1335 Further definitions of leaf flags are found in Section 5.2.7 given 1336 they have implications in terms of level and adjacency forming here. 1338 A node tries to form a three way adjacency if and only if 1339 1. the node is in the same PoD or either the node or the neighbor 1340 advertises "undefined/any" PoD membership (PoD# = 0) AND 1342 2. the neighboring node is running the same MAJOR schema version AND 1344 3. the neighbor is not member of some PoD while the node has a 1345 northbound adjacency already joining another PoD AND 1347 4. the neighboring node uses a valid System ID AND 1349 5. the neighboring node uses a different System ID than the node 1350 itself 1352 6. the advertised MTUs match on both sides AND 1354 7. both nodes advertise defined level values AND 1356 8. [ 1358 i) the node is at level 0 and has no three way adjacencies 1359 already to HAT nodes with level different than the adjacent 1360 node OR 1362 ii) the node is not at level 0 and the neighboring node is at 1363 level 0 OR 1365 iii) both nodes are at level 0 AND both indicate support for 1366 Section 5.3.9 OR 1368 iv) neither node is at level 0 and the neighboring node is at 1369 most one level away 1371 ]. 1373 The rule in Paragraph 3 MAY be optionally disregarded by a node if 1374 PoD detection is undesirable or has to be ignored. 1376 A node configured with "undefined" PoD membership MUST, after 1377 building first northbound three way adjacencies to a node being in a 1378 defined PoD, advertise that PoD as part of its LIEs. In case that 1379 adjacency is lost, from all available northbound three way 1380 adjacencies the node with the highest System ID and defined PoD is 1381 chosen. That way the northmost defined PoD value (normally the top 1382 spines in a PoD) can diffuse southbound towards the leafs "forcing" 1383 the PoD value on any node with "undefined" PoD. 1385 LIEs arriving with a TTL larger than 1 MUST be ignored. 1387 A node SHOULD NOT send out LIEs without defined level in the header 1388 but in certain scenarios it may be beneficial for trouble-shooting 1389 purposes. 1391 LIE exchange uses three way handshake mechanism which is a cleaned up 1392 version of [RFC5303]. Observe that for easier comprehension the 1393 terminology of one/two and three-way states does NOT align with OSPF 1394 or ISIS FSMs albeit they use roughly same mechanisms. 1396 5.2.3. Topology Exchange (TIE Exchange) 1398 5.2.3.1. Topology Information Elements 1400 Topology and reachability information in RIFT is conveyed by the 1401 means of TIEs which have good amount of commonalities with LSAs in 1402 OSPF. 1404 The TIE exchange mechanism uses the port indicated by each node in 1405 the LIE exchange and the interface on which the adjacency has been 1406 formed as destination. It SHOULD use TTL of 1 as well and set inter- 1407 network control precedence on according packets. 1409 TIEs contain sequence numbers, lifetimes and a type. Each type has 1410 ample identifying number space and information is spread across 1411 possibly many TIEs of a certain type by the means of a hash function 1412 that a node or deployment can individually determine. One extreme 1413 design choice is a prefix per TIE which leads to more BGP-like 1414 behavior where small increments are only advertised on route changes 1415 vs. deploying with dense prefix packing into few TIEs leading to more 1416 traditional IGP trade-off with fewer TIEs. An implementation may 1417 even rehash prefix to TIE mapping at any time at the cost of 1418 significant amount of re-advertisements of TIEs. 1420 More information about the TIE structure can be found in the schema 1421 in Appendix B. 1423 5.2.3.2. South- and Northbound Representation 1425 A central concept of RIFT is that each node represents itself 1426 differently depending on the direction in which it is advertising 1427 information. More precisely, a spine node represents two different 1428 databases over its adjacencies depending whether it advertises TIEs 1429 to the north or to the south/sideways. We call those differing TIE 1430 databases either south- or northbound (S-TIEs and N-TIEs) depending 1431 on the direction of distribution. 1433 The N-TIEs hold all of the node's adjacencies and local prefixes 1434 while the S-TIEs hold only all of the node's adjacencies, the default 1435 prefix with necessary disaggregated prefixes and local prefixes. We 1436 will explain this in detail further in Section 5.2.5. 1438 The TIE types are mostly symmetric in both directions and Table 2 1439 provides a quick reference to main TIE types including direction and 1440 their function. 1442 +-------------------+-----------------------------------------------+ 1443 | TIE-Type | Content | 1444 +-------------------+-----------------------------------------------+ 1445 | Node N-TIE | node properties and adjacencies | 1446 +-------------------+-----------------------------------------------+ 1447 | Node S-TIE | same content as node N-TIE | 1448 +-------------------+-----------------------------------------------+ 1449 | Prefix N-TIE | contains nodes' directly reachable prefixes | 1450 +-------------------+-----------------------------------------------+ 1451 | Prefix S-TIE | contains originated defaults and directly | 1452 | | reachable prefixes | 1453 +-------------------+-----------------------------------------------+ 1454 | Positive | contains disaggregated prefixes | 1455 | Disaggregation | | 1456 | S-TIE | | 1457 +-------------------+-----------------------------------------------+ 1458 | Negative | contains special, negatively disaggreagted | 1459 | Disaggregation | prefixes to support multi-plane designs | 1460 | S-TIE | | 1461 +-------------------+-----------------------------------------------+ 1462 | External Prefix | contains external prefixes | 1463 | N-TIE | | 1464 +-------------------+-----------------------------------------------+ 1465 | Key-Value N-TIE | contains nodes northbound KVs | 1466 +-------------------+-----------------------------------------------+ 1467 | Key-Value S-TIE | contains nodes southbound KVs | 1468 +-------------------+-----------------------------------------------+ 1470 Table 2: TIE Types 1472 As an example illustrating a databases holding both representations, 1473 consider the topology in Figure 2 with the optional link between 1474 spine 111 and spine 112 (so that the flooding on an East-West link 1475 can be shown). This example assumes unnumbered interfaces. First, 1476 here are the TIEs generated by some nodes. For simplicity, the key 1477 value elements which may be included in their S-TIEs or N-TIEs are 1478 not shown. 1480 Spine21 S-TIEs: 1481 Node S-TIE: 1483 NodeElement(level=2, neighbors((Spine 111, level 1, cost 1), 1484 (Spine 112, level 1, cost 1), (Spine 121, level 1, cost 1), 1485 (Spine 122, level 1, cost 1))) 1486 Prefix S-TIE: 1487 SouthPrefixesElement(prefixes(0/0, cost 1), (::/0, cost 1)) 1489 Spine 111 S-TIEs: 1490 Node S-TIE: 1491 NodeElement(level=1, neighbors((Spine21, level 2, cost 1, links(...)), 1492 (Spine22, level 2, cost 1, links(...)), 1493 (Spine 112, level 1, cost 1, links(...)), 1494 (Leaf111, level 0, cost 1, links(...)), 1495 (Leaf112, level 0, cost 1, links(...)))) 1496 Prefix S-TIE: 1497 SouthPrefixesElement(prefixes(0/0, cost 1), (::/0, cost 1)) 1499 Spine 111 N-TIEs: 1500 Node N-TIE: 1501 NodeElement(level=1, 1502 neighbors((Spine21, level 2, cost 1, links(...)), 1503 (Spine22, level 2, cost 1, links(...)), 1504 (Spine 112, level 1, cost 1, links(...)), 1505 (Leaf111, level 0, cost 1, links(...)), 1506 (Leaf112, level 0, cost 1, links(...)))) 1507 Prefix N-TIE: 1508 NorthPrefixesElement(prefixes(Spine 111.loopback) 1510 Spine 121 S-TIEs: 1511 Node S-TIE: 1512 NodeElement(level=1, neighbors((Spine21,level 2,cost 1), 1513 (Spine22, level 2, cost 1), (Leaf121, level 0, cost 1), 1514 (Leaf122, level 0, cost 1))) 1515 Prefix S-TIE: 1516 SouthPrefixesElement(prefixes(0/0, cost 1), (::/0, cost 1)) 1518 Spine 121 N-TIEs: 1519 Node N-TIE: 1520 NodeElement(level=1, 1521 neighbors((Spine21, level 2, cost 1, links(...)), 1522 (Spine22, level 2, cost 1, links(...)), 1523 (Leaf121, level 0, cost 1, links(...)), 1524 (Leaf122, level 0, cost 1, links(...)))) 1525 Prefix N-TIE: 1526 NorthPrefixesElement(prefixes(Spine 121.loopback) 1528 Leaf112 N-TIEs: 1529 Node N-TIE: 1530 NodeElement(level=0, 1531 neighbors((Spine 111, level 1, cost 1, links(...)), 1532 (Spine 112, level 1, cost 1, links(...)))) 1533 Prefix N-TIE: 1534 NorthPrefixesElement(prefixes(Leaf112.loopback, Prefix112, 1535 Prefix_MH)) 1537 Figure 14: example TIES generated in a 2 level spine-and-leaf 1538 topology 1540 5.2.3.3. Flooding 1542 The mechanism used to distribute TIEs is the well-known (albeit 1543 modified in several respects to address fat tree requirements) 1544 flooding mechanism used by today's link-state protocols. Although 1545 flooding is initially more demanding to implement it avoids many 1546 problems with update style used in diffused computation such as path 1547 vector protocols. Since flooding tends to present an unscalable 1548 burden in large, densely meshed topologies (fat trees being 1549 unfortunately such a topology) we provide as solution a close to 1550 optimal global flood reduction and load balancing optimization in 1551 Section 5.2.3.9. 1553 As described before, TIEs themselves are transported over UDP with 1554 the ports indicated in the LIE exchanges and using the destination 1555 address on which the LIE adjacency has been formed. For unnumbered 1556 IPv4 interfaces same considerations apply as in equivalent OSPF case. 1558 On reception of a TIE with an undefined level value in the packet 1559 header the node SHOULD issue a warning and indiscriminately discard 1560 the packet. 1562 Precise finite state machines and procedures can be found in 1563 Appendix C.3. 1565 5.2.3.4. TIE Flooding Scopes 1567 In a somewhat analogous fashion to link-local, area and domain 1568 flooding scopes, RIFT defines several complex "flooding scopes" 1569 depending on the direction and type of TIE propagated. 1571 Every N-TIE is flooded northbound, providing a node at a given level 1572 with the complete topology of the Clos or Fat Tree network underneath 1573 it, including all specific prefixes. This means that a packet 1574 received from a node at the same or lower level whose destination is 1575 covered by one of those specific prefixes may be routed directly 1576 towards the node advertising that prefix rather than sending the 1577 packet to a node at a higher level. 1579 A node's Node S-TIEs, consisting of all node's adjacencies and prefix 1580 S-TIEs limited to those related to default IP prefix and 1581 disaggregated prefixes, are flooded southbound in order to allow the 1582 nodes one level down to see connectivity of the higher level as well 1583 as reachability to the rest of the fabric. In order to allow an E-W 1584 disconnected node in a given level to receive the S-TIEs of other 1585 nodes at its level, every *NODE* S-TIE is "reflected" northbound to 1586 level from which it was received. It should be noted that East-West 1587 links are included in South TIE flooding (except at ToF level); those 1588 TIEs need to be flooded to satisfy algorithms in Section 5.2.4. In 1589 that way nodes at same level can learn about each other without a 1590 lower level, e.g. in case of leaf level. The precise flooding scopes 1591 are given in Table 3. Those rules govern as well what SHOULD be 1592 included in TIDEs on the adjacency. Again, East-West flooding scopes 1593 are identical to South flooding scopes except in case of ToF East- 1594 West links (rings). 1596 Node S-TIE "south reflection" allows to support positive 1597 disaggregation on failures describes in Section 5.2.5 and flooding 1598 reduction in Section 5.2.3.9. 1600 +-----------+---------------------+---------------+-----------------+ 1601 | Type / | South | North | East-West | 1602 | Direction | | | | 1603 +-----------+---------------------+---------------+-----------------+ 1604 | node | flood if level of | flood if | flood only if | 1605 | S-TIE | originator is equal | level of | this node is | 1606 | | to this node | originator is | not ToF | 1607 | | | higher than | | 1608 | | | this node | | 1609 +-----------+---------------------+---------------+-----------------+ 1610 | non-node | flood self- | flood only if | flood only if | 1611 | S-TIE | originated only | neighbor is | self-originated | 1612 | | | originator of | and this node | 1613 | | | TIE | is not ToF | 1614 +-----------+---------------------+---------------+-----------------+ 1615 | all | never flood | flood always | flood only if | 1616 | N-TIEs | | | this node is | 1617 | | | | ToF | 1618 +-----------+---------------------+---------------+-----------------+ 1619 | TIDE | include at least | include at | if this node is | 1620 | | all non-self | least all | ToF then | 1621 | | originated N-TIE | node S-TIEs | include all | 1622 | | headers and self- | and all | N-TIEs, | 1623 | | originated S-TIE | S-TIEs | otherwise only | 1624 | | headers and node | originated by | self-originated | 1625 | | S-TIEs of nodes at | peer and all | TIEs | 1626 | | same level | N-TIEs | | 1627 +-----------+---------------------+---------------+-----------------+ 1628 | TIRE as | request all N-TIEs | request all | if this node is | 1629 | Request | and all peer's | S-TIEs | ToF then apply | 1630 | | self-originated | | North scope | 1631 | | TIEs and all node | | rules, | 1632 | | S-TIEs | | otherwise South | 1633 | | | | scope rules | 1634 +-----------+---------------------+---------------+-----------------+ 1635 | TIRE as | Ack all received | Ack all | Ack all | 1636 | Ack | TIEs | received TIEs | received TIEs | 1637 +-----------+---------------------+---------------+-----------------+ 1639 Table 3: Flooding Scopes 1641 If the TIDE includes additional TIE headers beside the ones 1642 specified, the receiving neighbor must apply according filter to the 1643 received TIDE strictly and MUST NOT request the extra TIE headers 1644 that were not allowed by the flooding scope rules in its direction. 1646 As an example to illustrate these rules, consider using the topology 1647 in Figure 2, with the optional link between spine 111 and spine 112, 1648 and the associated TIEs given in Figure 14. The flooding from 1649 particular nodes of the TIEs is given in Table 4. 1651 +-------------+----------+------------------------------------------+ 1652 | Router | Neighbor | TIEs | 1653 | floods to | | | 1654 +-------------+----------+------------------------------------------+ 1655 | Leaf111 | Spine | Leaf111 N-TIEs, Spine 111 node S-TIE | 1656 | | 112 | | 1657 | Leaf111 | Spine | Leaf111 N-TIEs, Spine 112 node S-TIE | 1658 | | 111 | | 1659 | | | | 1660 | Spine 111 | Leaf111 | Spine 111 S-TIEs | 1661 | Spine 111 | Leaf112 | Spine 111 S-TIEs | 1662 | Spine 111 | Spine | Spine 111 S-TIEs | 1663 | | 112 | | 1664 | Spine 111 | Spine21 | Spine 111 N-TIEs, Leaf111 N-TIEs, | 1665 | | | Leaf112 N-TIEs, Spine22 node S-TIE | 1666 | Spine 111 | Spine22 | Spine 111 N-TIEs, Leaf111 N-TIEs, | 1667 | | | Leaf112 N-TIEs, Spine21 node S-TIE | 1668 | | | | 1669 | ... | ... | ... | 1670 | Spine21 | Spine | Spine21 S-TIEs | 1671 | | 111 | | 1672 | Spine21 | Spine | Spine21 S-TIEs | 1673 | | 112 | | 1674 | Spine21 | Spine | Spine21 S-TIEs | 1675 | | 121 | | 1676 | Spine21 | Spine | Spine21 S-TIEs | 1677 | | 122 | | 1678 | ... | ... | ... | 1679 +-------------+----------+------------------------------------------+ 1681 Table 4: Flooding some TIEs from example topology 1683 5.2.3.5. 'Flood Only Node TIEs' Bit 1685 RIFT includes an optional ECN mechanism to prevent "flooding inrush" 1686 on restart or bring-up with many southbound neighbors. A node MAY 1687 set on its LIEs the according bit to indicate to the neighbor that it 1688 should temporarily flood node TIEs only to it. It should only set it 1689 in the southbound direction. The receiving node SHOULD accomodate 1690 the request to lessen the flooding load on the affected node if south 1691 of the sender and SHOULD ignore the bit if northbound. 1693 Obviously this mechanism is most useful in southbound direction. The 1694 distribution of node TIEs guarantees correct behavior of algorithms 1695 like disaggregation or default route origination. Furthermore 1696 though, the use of this bit presents an inherent trade-off between 1697 processing load and convergence speed since suppressing flooding of 1698 northbound prefixes from neighbors will lead to blackholes. 1700 5.2.3.6. Initial and Periodic Database Synchronization 1702 The initial exchange of RIFT is modeled after ISIS with TIDE being 1703 equivalent to CSNP and TIRE playing the role of PSNP. The content of 1704 TIDEs and TIREs is governed by Table 3. 1706 5.2.3.7. Purging and Roll-Overs 1708 RIFT does not purge information that has been distributed by the 1709 protocol. Purging mechanisms in other routing protocols have proven 1710 to be complex and fragile over many years of experience. Abundant 1711 amounts of memory are available today even on low-end platforms. The 1712 information will age out and all computations will deliver correct 1713 results if a node leaves the network due to the new information 1714 distributed by its adjacent nodes. 1716 Once a RIFT node issues a TIE with an ID, it MUST preserve the ID as 1717 long as feasible (also when the protocol restarts), even if the TIE 1718 looses all content. The re-advertisement of empty TIE fulfills the 1719 purpose of purging any information advertised in previous versions. 1720 The originator is free to not re-originate the according empty TIE 1721 again or originate an empty TIE with relatively short lifetime to 1722 prevent large number of long-lived empty stubs polluting the network. 1723 Each node MUST timeout and clean up the according empty TIEs 1724 independently. 1726 Upon restart a node MUST, as any link-state implementation, be 1727 prepared to receive TIEs with its own system ID and supersede them 1728 with equivalent, newly generated, empty TIEs with a higher sequence 1729 number. As above, the lifetime can be relatively short since it only 1730 needs to exceed the necessary propagation and processing delay by all 1731 the nodes that are within the TIE's flooding scope. 1733 TIE sequence numbers are rolled over using the method described in 1734 Appendix A. First sequence number of any spontaneously originated 1735 TIE (i.e. not originated to override a detected older copy in the 1736 network) MUST be a reasonbly unpredictable random number in the 1737 interval [0, 2^10-1] which will prevent otherwise identical TIE 1738 headers to remain "stuck" in the network with content different from 1739 TIE originated after reboot. 1741 5.2.3.8. Southbound Default Route Origination 1743 Under certain conditions nodes issue a default route in their South 1744 Prefix TIEs with costs as computed in Section 5.3.6.1. 1746 A node X that 1748 1. is NOT overloaded AND 1750 2. has southbound or East-West adjacencies 1752 originates in its south prefix TIE such a default route IIF 1754 1. all other nodes at X's' level are overloaded OR 1756 2. all other nodes at X's' level have NO northbound adjacencies OR 1758 3. X has computed reachability to a default route during N-SPF. 1760 The term "all other nodes at X's' level" describes obviously just the 1761 nodes at the same level in the PoD with a viable lower level 1762 (otherwise the node S-TIEs cannot be reflected and the nodes in e.g. 1763 PoD 1 and PoD 2 are "invisible" to each other). 1765 A node originating a southbound default route MUST install a default 1766 discard route if it did not compute a default route during N-SPF. 1768 5.2.3.9. Northbound TIE Flooding Reduction 1770 Section 1.4 of the Optimized Link State Routing Protocol [RFC3626] 1771 (OLSR) introduces the concept of a "multipoint relay" (MPR) that 1772 minimize the overhead of flooding messages in the network by reducing 1773 redundant retransmissions in the same region. 1775 A similar technique is applied to RIFT to control northbound 1776 flooding. Important observations first: 1778 1. a node MUST flood self-originated N-TIEs to all the reachable 1779 nodes at the level above which we call the node's "parents"; 1781 2. it is typically not necessary that all parents reflood the N-TIEs 1782 to achieve a complete flooding of all the reachable nodes two 1783 levels above which we choose to call the node's "grandparents"; 1785 3. to control the volume of its flooding two hops North and yet keep 1786 it robust enough, it is advantageous for a node to select a 1787 subset of its parents as "Flood Repeaters" (FRs), which combined 1788 together deliver two or more copies of its flooding to all of its 1789 parents, i.e. the originating node's grandparents; 1791 4. nodes at the same level do NOT have to agree on a specific 1792 algorithm to select the FRs, but overall load balancing should be 1793 achieved so that different nodes at the same level should tend to 1794 select different parents as FRs; 1796 5. there are usually many solutions to the problem of finding a set 1797 of FRs for a given node; the problem of finding the minimal set 1798 is (similar to) a NP-Complete problem and a globally optimal set 1799 may not be the minimal one if load-balancing with other nodes is 1800 an important consideration; 1802 6. it is expected that there will be often sets of equivalent nodes 1803 at a level L, defined as having a common set of parents at L+1. 1804 Applying this observation at both L and L+1, an algorithm may 1805 attempt to split the larger problem in a sum of smaller separate 1806 problems; 1808 7. it is another expectation that there will be from time to time a 1809 broken link between a parent and a grandparent, and in that case 1810 the parent is probably a poor FR due to its lower reliability. 1811 An algorithm may attempt to eliminate parents with broken 1812 northbound adjacencies first in order to reduce the number of 1813 FRs. Albeit it could be argued that relying on higher fanout FRs 1814 will slow flooding due to higher replication load reliability of 1815 FR's links seems to be a more pressing concern. 1817 In a fully connected Clos Network, this means that a node selects one 1818 arbitrary parent as FR and then a second one for redundancy. The 1819 computation can be kept relatively simple and completely distributed 1820 without any need for synchronization amongst nodes. In a "PoD" 1821 structure, where the Level L+2 is partitioned in silos of equivalent 1822 grandparents that are only reachable from respective parents, this 1823 means treating each silo as a fully connected Clos Network and solve 1824 the problem within the silo. 1826 In terms of signaling, a node has enough information to select its 1827 set of FRs; this information is derived from the node's parents' Node 1828 S-TIEs, which indicate the parent's reachable northbound adjacencies 1829 to its own parents, i.e. the node's grandparents. A node may send a 1830 LIE to a northbound neighbor with the optional boolean field 1831 `you_are_flood_repeater` set to false, to indicate that the 1832 northbound neighbor is not a flood repeater for the node that sent 1833 the LIE. In that case the northbound neighbor SHOULD NOT reflood 1834 northbound TIEs received from the node that sent the LIE. If the 1835 `you_are_flood_repeater` is absent or if `you_are_flood_repeater` is 1836 set to true, then the northbound neighbor is a flood repeater for the 1837 node that sent the LIE and MUST reflood northbound TIEs received from 1838 that node. 1840 This specification proposes a simple default algorithm that SHOULD be 1841 implemented and used by default on every RIFT node. 1843 o let |NA(Node) be the set of Northbound adjacencies of node Node 1844 and CN(Node) be the cardinality of |NA(Node); 1846 o let |SA(Node) be the set of Southbound adjacencies of node Node 1847 and CS(Node) be the cardinality of |SA(Node); 1849 o let |P(Node) be the set of node Node's parents; 1851 o let |G(Node) be the set of node Node's grandparents. Observe 1852 that |G(Node) = |P(|P(Node)); 1854 o let N be the child node at level L computing a set of FR; 1856 o let P be a node at level L+1 and a parent node of N, i.e. bi- 1857 directionally reachable over adjacency A(N, P); 1859 o let G be a grandparent node of N, reachable transitively via a 1860 parent P over adjacencies ADJ(N, P) and ADJ(P, G). Observe that N 1861 does not have enough information to check bidirectional 1862 reachability of A(P, G); 1864 o let R be a redundancy constant integer; a value of 2 or higher for 1865 R is RECOMMENDED; 1867 o let S be a similarity constant integer; a value in range 0 .. 2 1868 for S is RECOMMENDED, the value of 1 SHOULD be used. Two 1869 cardinalities are considered as equivalent if their absolute 1870 difference is less than or equal to S, i.e. |a-b|<=S. 1872 o let RND be a 64-bit random number generated by the system once on 1873 startup. 1875 The algorithm consists of the following steps: 1877 1. Derive a 64-bits number by XOR'ing 'N's system ID with RND. 1879 2. Derive a 16-bits pseudo-random unsigned integer PR(N) from the 1880 resulting 64-bits number by splitting it in 16-bits-long words 1881 W1, W2, W3, W4 (where W1 are the least significant 16 bits of the 1882 64-bits number, and W4 are the most significant 16 bits) and then 1883 XOR'ing the circularly shifted resulting words together: 1885 (W1<<1) xor (W2<<2) xor (W3<<3) xor (W4<<4); 1887 where << is the circular shift operator. 1889 3. Sort the parents by decreasing number of northbound adjacencies 1890 (using decreasing system id of the parent as tie-breaker): 1891 sort |P(N) by decreasing CN(P), for all P in |P(N), as ordered 1892 array |A(N) 1894 4. Partition |A(N) in subarrays |A_k(N) of parents with equivalent 1895 cardinality of northbound adjacencies (in other words with 1896 equivalent number of grandparents they can reach): 1898 1. set k=0; // k is the ID of the subarrray 1900 2. set i=0; 1902 3. while i < CN(N) do 1904 1. set j=i; 1906 2. while i < CN(N) and CN(|A(N)[j]) - CN(|A(N)[i]) <= S 1908 1. place |A(N)[i] in |A_k(N) // abstract action, maybe 1909 noop 1911 2. set i=i+1; 1913 3. /* At this point j is the index in |A(N) of the first 1914 member of |A_k(N) and (i-j) is C_k(N) defined as the 1915 cardinality of |A_k(N) */ 1917 4. set k=k+1; 1919 4. /* At this point k is the total number of subarrays, 1920 initialized for the shuffling operation below */ 1922 5. shuffle individually each subarrays |A_k(N) of cardinality C_k(N) 1923 within |A(N) using the Durstenfeld variation of Fisher-Yates 1924 algorithm that depends on N's System ID: 1926 1. while k > 0 do 1928 1. for i from C_k(N)-1 to 1 decrementing by 1 do 1930 1. set j to PR(N) modulo i; 1932 2. exchange |A_k[j] and |A_k[i]; 1934 2. set k=k-1; 1936 6. For each grandparent G, initialize a counter c(G) with the number 1937 of its south-bound adjacencies to elected flood repeaters (which 1938 is initially zero): 1940 1. for each G in |G(N) set c(G) = 0; 1942 7. Finally keep as FRs only parents that are needed to maintain the 1943 number of adjacencies between the FRs and any grandparent G equal 1944 or above the redundancy constant R: 1946 1. for each P in reshuffled |A(N); 1948 1. if there exists an adjacency ADJ(P, G) in |NA(P) such 1949 that c(G) < R then 1951 1. place P in FR set; 1953 2. for all adjacencies ADJ(P, G') in |NA(P) increment 1954 c(G') 1956 2. If any c(G) is still < R, it was not possible to elect a set 1957 of FRs that covers all grandparents with redundancy R 1959 Additional rules for flooding reduction: 1961 1. The algorithm MUST be re-evaluated by a node on every change of 1962 local adjacencies or reception of a parent S-TIE with changed 1963 adjacencies. A node MAY apply a hysteresis to prevent excessive 1964 amount of computation during periods of network instability just 1965 like in case of reachability computation. 1967 2. A node SHOULD send out LIEs that grant flood repeater status 1968 before LIEs that revoke it on flood repeater set changes to 1969 prevent transient behavior where the full coverage of grand 1970 parents is not guaranteed. Albeit the condition will correct in 1971 positively stable manner due to LIE retransmission and periodic 1972 TIDEs, it can slow down flooding convergence on flood repeater 1973 status changes. 1975 3. A node always floods its self-originated TIEs. 1977 4. A node receiving a TIE originated by a node for which it is not a 1978 flood repeater does NOT re-flood such TIEs to its neighbors 1979 except for rules in Paragraph 6. 1981 5. The indication of flood reduction capability is carried in the 1982 node TIEs and can be used to optimize the algorithm to account 1983 for nodes that will flood regardless. 1985 6. A node generates TIDEs as usual but when receiving TIREs or TIDEs 1986 resulting in requests for a TIE of which the newest received copy 1987 came on an adjacency where the node was not flood repeater it 1988 SHOULD ignore such requests on first and first request ONLY. 1989 Normally, the nodes that received the TIEs as flooding repeaters 1990 should satisfy the requesting node and with that no further TIREs 1991 for such TIEs will be generated. Otherwise, the next set of 1992 TIDEs and TIREs MUST lead to flooding independent of the flood 1993 repeater status. This solves a very difficult incast problem on 1994 nodes restarting with a very wide fanout, especially northbound. 1995 To retrieve the full database they often end up processing many 1996 in-rushing copies whereas this approach should load-balance the 1997 incoming database between adjacent nodes and flood repeaters 1998 should guarantee that two copies are sent by different nodes to 1999 ensure against any losses. 2001 7. Obviously sine flooding reduction does NOT apply to self 2002 originated TIEs and since all policy-guided information consists 2003 of self-originated TIEs those are unaffected. 2005 5.2.3.10. Special Considerations 2007 First, due to the distributed, asynchronous nature of ZTP, it can 2008 create temporary convergence anomalies where nodes at higher levels 2009 of the fabric temporarily see themselves lower than they belong. 2010 Since flooding can begin before ZTP is "finished" and in fact must do 2011 so given there is no global termination criteria, information may end 2012 up in wrong layers. A special clause when changing level takes care 2013 of that. 2015 More difficult is a condition where a node floods a TIE north towards 2016 a super-spine, then its spine reboots, in fact partitioning 2017 superspine from it directly and then the node itself reboots. That 2018 leaves in a sense the super-spine holding the "primary copy" of the 2019 node's TIE. Normally this condition is resolved easily by the node 2020 re-originating its TIE with a higher sequence number than it sees in 2021 northbound TIEs, here however, when spine comes back it won't be able 2022 to obtain a N-TIE from its superspine easily and with that the node 2023 below may issue the same version of the TIE with a lower sequence 2024 number. Flooding procedures are are extended to deal with the 2025 problem by the means of special clauses that override the database of 2026 a lower level with headers of newer TIEs seen in TIDEs coming from 2027 the north. 2029 5.2.4. Reachability Computation 2031 A node has three sources of relevant information. A node knows the 2032 full topology south from the received N-TIEs. A node has the set of 2033 prefixes with associated distances and bandwidths from received 2034 S-TIEs. 2036 To compute reachability, a node runs conceptually a northbound and a 2037 southbound SPF. We call that N-SPF and S-SPF. 2039 Since neither computation can "loop", it is possible to compute non- 2040 equal-cost or even k-shortest paths [EPPSTEIN] and "saturate" the 2041 fabric to the extent desired. 2043 5.2.4.1. Northbound SPF 2045 N-SPF uses northbound and East-West adjacencies in the computing 2046 node's node N-TIEs (since if the node is a leaf it may not have 2047 generated a node S-TIE) when starting Dijkstra. Observe that N-SPF 2048 is really just a one hop variety since Node S-TIEs are not re-flooded 2049 southbound beyond a single level (or East-West) and with that the 2050 computation cannot progress beyond adjacent nodes. 2052 Once progressing, we are using the next level's node S-TIEs to find 2053 according adjacencies to verify backlink connectivity. Just as in 2054 case of IS-IS or OSPF, two unidirectional links are associated 2055 together to confirm bidirectional connectivity. Particular care MUST 2056 be paid that the Node TIEs do not only contain the correct system IDs 2057 but matching levels as well. 2059 Default route found when crossing an E-W link is used IIF 2061 1. the node itself does NOT have any northbound adjacencies AND 2063 2. the adjacent node has one or more northbound adjacencies 2065 This rule forms a "one-hop default route split-horizon" and prevents 2066 looping over default routes while allowing for "one-hop protection" 2067 of nodes that lost all northbound adjacencies except at Top-of-Fabric 2068 where the links are used exclusively to flood topology information in 2069 multi-plane designs. 2071 Other south prefixes found when crossing E-W link MAY be used IIF 2073 1. no north neighbors are advertising same or supersuming non- 2074 default prefix AND 2076 2. the node does not originate a non-default supersuming prefix 2077 itself. 2079 i.e. the E-W link can be used as the gateway of last resort for a 2080 specific prefix only. Using south prefixes across E-W link can be 2081 beneficial e.g. on automatic de-aggregation in pathological fabric 2082 partitioning scenarios. 2084 A detailed example can be found in Section 6.4. 2086 5.2.4.2. Southbound SPF 2088 S-SPF uses only the southbound adjacencies in the node S-TIEs, i.e. 2089 progresses towards nodes at lower levels. Observe that E-W 2090 adjacencies are NEVER used in the computation. This enforces the 2091 requirement that a packet traversing in a southbound direction must 2092 never change its direction. 2094 S-SPF uses northbound adjacencies in node N-TIEs to verify backlink 2095 connectivity. 2097 5.2.4.3. East-West Forwarding Within a Level 2099 Ultimately, it should be observed that in presence of a "ring" of E-W 2100 links in a level neither SPF will provide a "ring protection" scheme 2101 since such a computation would have to deal necessarily with breaking 2102 of "loops" in generic Dijkstra sense; an application for which RIFT 2103 is not intended. It is outside the scope of this document how an 2104 underlay can be used to provide a full-mesh connectivity between 2105 nodes in the same level that would allow for N-SPF to provide 2106 protection for a single node loosing all its northbound adjacencies 2107 (as long as any of the other nodes in the level are northbound 2108 connected). 2110 Using south prefixes over horizontal links is optional and can 2111 protect against pathological fabric partitioning cases that leave 2112 only paths to destinations that would necessitate multiple changes of 2113 forwarding direction between north and south. 2115 5.2.5. Automatic Disaggregation on Link & Node Failures 2117 5.2.5.1. Positive, Non-transitive Disaggregation 2119 Under normal circumstances, node's S-TIEs contain just the 2120 adjacencies and a default route. However, if a node detects that its 2121 default IP prefix covers one or more prefixes that are reachable 2122 through it but not through one or more other nodes at the same level, 2123 then it MUST explicitly advertise those prefixes in an S-TIE. 2125 Otherwise, some percentage of the northbound traffic for those 2126 prefixes would be sent to nodes without according reachability, 2127 causing it to be black-holed. Even when not black-holing, the 2128 resulting forwarding could 'backhaul' packets through the higher 2129 level spines, clearly an undesirable condition affecting the blocking 2130 probabilities of the fabric. 2132 We refer to the process of advertising additional prefixes southbound 2133 as 'positive de-aggregation' or 'positive dis-aggregation'. 2135 A node determines the set of prefixes needing de-aggregation using 2136 the following steps: 2138 1. A DAG computation in the southern direction is performed first, 2139 i.e. the N-TIEs are used to find all of prefixes it can reach and 2140 the set of next-hops in the lower level for each of them. Such a 2141 computation can be easily performed on a fat tree by e.g. setting 2142 all link costs in the southern direction to 1 and all northern 2143 directions to infinity. We term set of those prefixes |R, and 2144 for each prefix, r, in |R, we define its set of next-hops to 2145 be |H(r). 2147 2. The node uses reflected S-TIEs to find all nodes at the same 2148 level in the same PoD and the set of southbound adjacencies for 2149 each. The set of nodes at the same level is termed |N and for 2150 each node, n, in |N, we define its set of southbound adjacencies 2151 to be |A(n). 2153 3. For a given r, if the intersection of |H(r) and |A(n), for any n, 2154 is null then that prefix r must be explicitly advertised by the 2155 node in an S-TIE. 2157 4. Identical set of de-aggregated prefixes is flooded on each of the 2158 node's southbound adjacencies. In accordance with the normal 2159 flooding rules for an S-TIE, a node at the lower level that 2160 receives this S-TIE will not propagate it south-bound. Neither 2161 is it necessary for the receiving node to reflect the 2162 disaggregated prefixes back over its adjacencies to nodes at the 2163 level from which it was received. 2165 To summarize the above in simplest terms: if a node detects that its 2166 default route encompasses prefixes for which one of the other nodes 2167 in its level has no possible next-hops in the level below, it has to 2168 disaggregate it to prevent black-holing or suboptimal routing through 2169 such nodes. Hence a node X needs to determine if it can reach a 2170 different set of south neighbors than other nodes at the same level, 2171 which are connected to it via at least one common south neighbor. If 2172 it can, then prefix disaggregation may be required. If it can't, 2173 then no prefix disaggregation is needed. An example of 2174 disaggregation is provided in Section 6.3. 2176 A possible algorithm is described last: 2178 1. Create partial_neighbors = (empty), a set of neighbors with 2179 partial connectivity to the node X's level from X's perspective. 2180 Each entry is a list of south neighbor of X and a list of nodes 2181 of X.level that can't reach that neighbor. 2183 2. A node X determines its set of southbound neighbors 2184 X.south_neighbors. 2186 3. For each S-TIE originated from a node Y that X has which is at 2187 X.level, if Y.south_neighbors is not the same as 2188 X.south_neighbors but the nodes share at least one southern 2189 neighbor, for each neighbor N in X.south_neighbors but not in 2190 Y.south_neighbors, add (N, (Y)) to partial_neighbors if N isn't 2191 there or add Y to the list for N. 2193 4. If partial_neighbors is empty, then node X does not to 2194 disaggregate any prefixes. If node X is advertising 2195 disaggregated prefixes in its S-TIE, X SHOULD remove them and re- 2196 advertise its according S-TIEs. 2198 A node X computes reachability to all nodes below it based upon the 2199 received N-TIEs first. This results in a set of routes, each 2200 categorized by (prefix, path_distance, next-hop-set). Alternately, 2201 for clarity in the following procedure, these can be organized by 2202 next-hop-set as ( (next-hops), {(prefix, path_distance)}). If 2203 partial_neighbors isn't empty, then the following procedure describes 2204 how to identify prefixes to disaggregate. 2206 disaggregated_prefixes = { empty } 2207 nodes_same_level = { empty } 2208 for each S-TIE 2209 if (S-TIE.level == X.level and 2210 X shares at least one S-neighbor with X) 2211 add S-TIE.originator to nodes_same_level 2212 end if 2213 end for 2215 for each next-hop-set NHS 2216 isolated_nodes = nodes_same_level 2217 for each NH in NHS 2218 if NH in partial_neighbors 2219 isolated_nodes = intersection(isolated_nodes, 2220 partial_neighbors[NH].nodes) 2221 end if 2222 end for 2224 if isolated_nodes is not empty 2225 for each prefix using NHS 2226 add (prefix, distance) to disaggregated_prefixes 2227 end for 2228 end if 2229 end for 2231 copy disaggregated_prefixes to X's S-TIE 2232 if X's S-TIE is different 2233 schedule S-TIE for flooding 2234 end if 2236 Figure 15: Computation of Disaggregated Prefixes 2238 Each disaggregated prefix is sent with the according path_distance. 2239 This allows a node to send the same S-TIE to each south neighbor. 2240 The south neighbor which is connected to that prefix will thus have a 2241 shorter path. 2243 Finally, to summarize the less obvious points partially omitted in 2244 the algorithms to keep them more tractable: 2246 1. all neighbor relationships MUST perform backlink checks. 2248 2. overload bits as introduced in Section 5.3.1 have to be respected 2249 during the computation. 2251 3. all the lower level nodes are flooded the same disaggregated 2252 prefixes since we don't want to build an S-TIE per node and 2253 complicate things unnecessarily. The PoD containing the prefix 2254 will prefer southbound anyway. 2256 4. positively disaggregated prefixes do NOT have to propagate to 2257 lower levels. With that the disturbance in terms of new flooding 2258 is contained to a single level experiencing failures. 2260 5. disaggregated prefix S-TIEs are not "reflected" by the lower 2261 level, i.e. nodes within same level do NOT need to be aware 2262 which node computed the need for disaggregation. 2264 6. The fabric is still supporting maximum load balancing properties 2265 while not trying to send traffic northbound unless necessary. 2267 In case positive disaggregation is triggered and due to the very 2268 stable but un-synchronized nature of the algorithm the nodes may 2269 issue the necessary disaggregated prefixes at different points in 2270 time. This can lead for a short time to an "incast" behavior where 2271 the first advertising router based on the nature of longest prefix 2272 match will attract all the traffic. An implementation MAY hence 2273 choose different strategies to address this behavior if needed. 2275 To close this section it is worth to observe that in a single plane 2276 ToF this disaggregation prevents blackholing up to (K_LEAF * P) link 2277 failures in terms of Section 5.1.2 or in other terms, it takes at 2278 minimum that many link failures to partition the ToF into multiple 2279 planes. 2281 5.2.5.2. Negative, Transitive Disaggregation for Fallen Leafs 2283 As explained in Section 5.1.3 failures in multi-plane Top-of-Fabric 2284 or more than (K_LEAF * P) links failing in single plane design can 2285 generate fallen leafs. Such scenario cannot be addressed by positive 2286 disaggregation only and needs a further mechanism. 2288 5.2.5.2.1. Cabling of Multiple Top-of-Fabric Planes 2290 Let us return in this section to designs with multiple planes as 2291 shown in Figure 3. Figure 16 highlights how the ToF is cabled in 2292 case of two planes by the means of dual-rings to distribute all the 2293 N-TIEs within both planes. For people familiar with traditional 2294 link-state routing protocols ToF level can be considered equivalent 2295 to area 0 in OSPF or level-2 in ISIS which need to be "connected" as 2296 well for the protocol to operate correctly. 2298 . ++==========++ ++==========++ 2299 . II II II II 2300 .+----++--+ +----++--+ +----++--+ +----++--+ 2301 .|ToF A1| |ToF B1| |ToF B2| |ToF A2| 2302 .++-+-++--+ ++-+-++--+ ++-+-++--+ ++-+-++--+ 2303 . | | II | | II | | II | | II 2304 . | | ++==========++ | | ++==========++ 2305 . | | | | | | | | 2306 . 2307 . ~~~ Highlighted ToF of the previous multi-plane figure ~~ 2309 Figure 16: Topologically connected planes 2311 As described in Section 5.1.3 failures in multi-plane fabrics can 2312 lead to blackholes which normal positive disaggregation cannot fix. 2313 The mechanism of negative, transitive disaggregation incorporated in 2314 RIFT provides the according solution. 2316 5.2.5.2.2. Transitive Advertisement of Negative Disaggregates 2318 A ToF node that discovers that it cannot reach a fallen leaf 2319 disaggregates all the prefixes of such leafs. It uses for that 2320 purpose negative prefix S-TIEs that are, as usual, flooded southwards 2321 with the scope defined in Section 5.2.3.4. 2323 Transitively, a node explicitly loses connectivity to a prefix when 2324 none of its children advertises it and when the prefix is negatively 2325 disaggregated by all of its parents. When that happens, the node 2326 originates the negative prefix further down south. Since the 2327 mechanism applies recursively south the negative prefix may propagate 2328 transitively all the way down to the leaf. This is necessary since 2329 leafs connected to multiple planes by means of disjoint paths may 2330 have to choose the correct plane already at the very bottom of the 2331 fabric to make sure that they don't send traffic towards another leaf 2332 using a plane where it is "fallen" at which in point a blackhole is 2333 unavoidable. 2335 When the connectivity is restored, a node that disaggregated a prefix 2336 withdraws the negative disaggregation by the usual mechanism of re- 2337 advertising TIEs omitting the negative prefix. 2339 5.2.5.2.3. Computation of Negative Disaggregates 2341 The document omitted so far the description of the computation 2342 necessary to generate the correct set of negative prefixes. Negative 2343 prefixes can in fact be advertised due to two different triggers. We 2344 describe them consecutively. 2346 The first origination reason is a computation that uses all the node 2347 N-TIEs to build the set of all reachable nodes by reachability 2348 computation over the complete graph and including ToF links. The 2349 computation uses the node itself as root. This is compared with the 2350 result of the normal southbound SPF as described in Section 5.2.4.2. 2351 The difference are the fallen leafs and all their attached prefixes 2352 are advertised as negative prefixes southbound if the node does not 2353 see the prefix being reachable within southbound SPF. 2355 The second mechanism hinges on the understanding how the negative 2356 prefixes are used within the computation as described in Figure 17. 2357 When attaching the negative prefixes at certain point in time the 2358 negative prefix may find itself with all the viable nodes from the 2359 shorter match nexthop being pruned. In other words, all its 2360 northbound neighbors provided a negative prefix advertisement. This 2361 is the trigger to advertise this negative prefix transitively south 2362 and normally caused by the node being in a plane where the prefix 2363 belongs to a fabric leaf that has "fallen" in this plane. Obviously, 2364 when one of the northbound switches withdraws its negative 2365 advertisement, the node has to withdraw its transitively provided 2366 negative prefix as well. 2368 5.2.6. Attaching Prefixes 2370 After the SPF is run, it is necessary to attach according prefixes. 2371 For S-SPF, prefixes from an N-TIE are attached to the originating 2372 node with that node's next-hop set and a distance equal to the 2373 prefix's cost plus the node's minimized path distance. The RIFT 2374 route database, a set of (prefix, type=spf, path_distance, next-hop 2375 set), accumulates these results. Obviously, the prefix retains its 2376 type which is used to tie-break between the same prefix advertised 2377 with different types. 2379 In case of N-SPF prefixes from each S-TIE need to also be added to 2380 the RIFT route database. The N-SPF is really just a stub so the 2381 computing node needs simply to determine, for each prefix in an S-TIE 2382 that originated from adjacent node, what next-hops to use to reach 2383 that node. Since there may be parallel links, the next-hops to use 2384 can be a set; presence of the computing node in the associated Node 2385 S-TIE is sufficient to verify that at least one link has 2386 bidirectional connectivity. The set of minimum cost next-hops from 2387 the computing node X to the originating adjacent node is determined. 2389 Each prefix has its cost adjusted before being added into the RIFT 2390 route database. The cost of the prefix is set to the cost received 2391 plus the cost of the minimum distance next-hop to that neighbor while 2392 taking into account its attributes such as mobility per Section 5.3.3 2393 necessary. Then each prefix can be added into the RIFT route 2394 database with the next_hop_set; ties are broken based upon type first 2395 and then distance and further attributes. RIFT route preferences are 2396 normalized by the according thrift model type. 2398 An exemplary implementation for node X follows: 2400 for each S-TIE 2401 if S-TIE.level > X.level 2402 next_hop_set = set of minimum cost links to the S-TIE.originator 2403 next_hop_cost = minimum cost link to S-TIE.originator 2404 end if 2405 for each prefix P in the S-TIE 2406 P.cost = P.cost + next_hop_cost 2407 if P not in route_database: 2408 add (P, type=DistVector, P.cost, next_hop_set) to route_database 2409 end if 2410 if (P in route_database): 2411 if route_database[P].cost > P.cost or route_database[P].type > P.type: 2412 update route_database[P] with (P, DistVector, P.cost, P.type, next_hop_set) 2413 else if route_database[P].cost == P.cost and route_database[P].type == P.type: 2414 update route_database[P] with (P, DistVector, P.cost, P.type, 2415 merge(next_hop_set, route_database[P].next_hop_set)) 2416 else 2417 // Not preferred route so ignore 2418 end if 2419 end if 2420 end for 2421 end for 2423 Figure 17: Adding Routes from S-TIE Positive and Negative Prefixes 2425 After the positive prefixes are attached and tie-broken, negative 2426 prefixes are attached and used in case of northbound computation, 2427 ideally from the shortest length to the longest. The nexthop 2428 adjacencies for a negative prefix are inherited from the longest 2429 prefix that aggregates it, and subsequently adjacencies to nodes that 2430 advertised negative for this prefix are removed. 2432 The rule of inheritance MUST be maintained when the nexthop list for 2433 a prefix is modified, as the modification may affect the entries for 2434 matching negative prefixes of immediate longer prefix length. For 2435 instance, if a nexthop is added, then by inheritance it must be added 2436 to all the negative routes of immediate longer prefixes length unless 2437 it is pruned due to a negative advertisement for the same next hop. 2438 Similarily, if a nexthop is deleted for a given prefix, then it is 2439 deleted for all the immediately aggregated negative routes. This 2440 will recurse in the case of nested negative prefix aggregations. 2442 The rule of inheritance must also be maintained when a new prefix of 2443 intermediate length is inserted, or when the immediately aggregating 2444 prefix is deleted from the routing table, making an even shorter 2445 aggregating prefix the one from which the negative routes now inherit 2446 their adjacencies. As the aggregating prefix changes, all the 2447 negative routes must be recomputed, and then again the process may 2448 recurse in case of nested negative prefix aggregations. 2450 Observe that despite seeming quite computationally expensive the 2451 operations are only necessary if the only available advertisements 2452 for a prefix are negative since tie-breaking always prefers positive 2453 information for the prefix which stops any kind of recursion since 2454 positive information never inherits next hops. 2456 To make the negative disaggregation less abstract and provide an 2457 example let us consider a ToP node T1 with 4 ToF parents S1..S4 as 2458 represented in Figure 18: 2460 +----+ +----+ +----+ +----+ N 2461 | S1 | | S1 | | S1 | | S1 | ^ 2462 +----+ +----+ +----+ +----+ W< + >E 2463 | | | | v 2464 |+--------+ | | S 2465 ||+-----------------+ | 2466 |||+----------------+---------+ 2467 |||| 2468 +----+ 2469 | T1 | 2470 +----+ 2472 Figure 18: A ToP node with 4 parents 2474 If all ToF nodes can reach all the prefixes in the network; with 2475 RIFT, they will normally advertise a default route south. An 2476 abstract Routing Information Base (RIB), more commonly known as a 2477 routing table, stores all types of maintained routes including the 2478 negative ones and "tie-breaks" for the best one, whereas an abstract 2479 Forwarding table (FIB) retains only the ultimately computed 2480 "positive" routing instructions. In T1, those tables would look as 2481 illustrated in Figure 19: 2483 +---------+ 2484 | Default | 2485 +---------+ 2486 | 2487 | +--------+ 2488 +---> | Via S1 | 2489 | +--------+ 2490 | 2491 | +--------+ 2492 +---> | Via S2 | 2493 | +--------+ 2494 | 2495 | +--------+ 2496 +---> | Via S3 | 2497 | +---------+ 2498 | 2499 | +--------+ 2500 +---> | Via S4 | 2501 +--------+ 2503 Figure 19: Abstract RIB 2505 In case T1 receives a negative advertisement for prefix 2001:db8::/32 2506 from S1 a negative route is stored in the RIB (indicated by a ~ 2507 sign), while the more specific routes to the complementing ToF nodes 2508 are installed in FIB. RIB and FIB in T1 now look as illustrated in 2509 Figure 20 and Figure 21, respectively: 2511 +---------+ +-----------------+ 2512 | Default | <-------------- | ~2001:db8::/32 | 2513 +---------+ +-----------------+ 2514 | | 2515 | +--------+ | +--------+ 2516 +---> | Via S1 | +---> | Via S1 | 2517 | +--------+ +--------+ 2518 | 2519 | +--------+ 2520 +---> | Via S2 | 2521 | +--------+ 2522 | 2523 | +--------+ 2524 +---> | Via S3 | 2525 | +---------+ 2526 | 2527 | +--------+ 2528 +---> | Via S4 | 2529 +--------+ 2531 Figure 20: Abstract RIB after negative 2001:db8::/32 from S1 2533 Negative 2001:db8::/32 entry inherits from ::/0, so the positive more 2534 specific routes are the complements to S1 in the set of next-hops for 2535 the default route. That entry is composed of S2, S3, and S4, or, in 2536 other words, it uses all entries of the default route with a "hole 2537 punched" for S1 into them. These are the next hops that are still 2538 available to reach 2001:db8::/32, now that S1 advertised that it will 2539 not forward 2001:db8::/32 anymore. Ultimately, those resulting next- 2540 hops are installed in FIB for the more specific route to 2541 2001:db8::/32 as illustrated below: 2543 +---------+ +---------------+ 2544 | Default | | 2001:db8::/32 | 2545 +---------+ +---------------+ 2546 | | 2547 | +--------+ | 2548 +---> | Via S1 | | 2549 | +--------+ | 2550 | | 2551 | +--------+ | +--------+ 2552 +---> | Via S2 | +---> | Via S2 | 2553 | +--------+ | +--------+ 2554 | | 2555 | +--------+ | +--------+ 2556 +---> | Via S3 | +---> | Via S3 | 2557 | +--------+ | +--------+ 2558 | | 2559 | +--------+ | +--------+ 2560 +---> | Via S4 | +---> | Via S4 | 2561 +--------+ +--------+ 2563 Figure 21: Abstract FIB after negative 2001:db8::/32 from S1 2565 To illustrate matters further let us consider T1 receiving a negative 2566 advertisement for prefix 2001:db8:1::/48 from S2, which is stored in 2567 RIB again. After the update, the RIB in T1 is illustrated in 2568 Figure 22: 2570 +---------+ +----------------+ +------------------+ 2571 | Default | <----- | ~2001:db8::/32 | <------ | ~2001:db8:1::/48 | 2572 +---------+ +----------------+ +------------------+ 2573 | | | 2574 | +--------+ | +--------+ | 2575 +---> | Via S1 | +---> | Via S1 | | 2576 | +--------+ +--------+ | 2577 | | 2578 | +--------+ | +--------+ 2579 +---> | Via S2 | +---> | Via S2 | 2580 | +--------+ +--------+ 2581 | 2582 | +--------+ 2583 +---> | Via S3 | 2584 | +---------+ 2585 | 2586 | +--------+ 2587 +---> | Via S4 | 2588 +--------+ 2590 Figure 22: Abstract RIB after negative 2001:db8:1::/48 from S2 2592 Negative 2001:db8:1::/48 inherits from 2001:db8::/32 now, so the 2593 positive more specific routes are the complements to S2 in the set of 2594 next hops for 2001:db8::/32, which are S3 and S4, or, in other words, 2595 all entries of the father with the negative holes "punched in" again. 2596 After the update, the FIB in T1 shows as illustrated in Figure 23: 2598 +---------+ +---------------+ +-----------------+ 2599 | Default | | 2001:db8::/32 | | 2001:db8:1::/48 | 2600 +---------+ +---------------+ +-----------------+ 2601 | | | 2602 | +--------+ | | 2603 +---> | Via S1 | | | 2604 | +--------+ | | 2605 | | | 2606 | +--------+ | +--------+ | 2607 +---> | Via S2 | +---> | Via S2 | | 2608 | +--------+ | +--------+ | 2609 | | | 2610 | +--------+ | +--------+ | +--------+ 2611 +---> | Via S3 | +---> | Via S3 | +---> | Via S3 | 2612 | +--------+ | +--------+ | +--------+ 2613 | | | 2614 | +--------+ | +--------+ | +--------+ 2615 +---> | Via S4 | +---> | Via S4 | +---> | Via S4 | 2616 +--------+ +--------+ +--------+ 2618 Figure 23: Abstract FIB after negative 2001:db8:1::/48 from S2 2620 Further, let us say that S3 stops advertising its service as default 2621 gateway. The entry is removed from RIB as usual. In order to update 2622 the FIB, it is necessary to eliminate the FIB entry for the default 2623 route, as well as all the FIB entries that were created for negative 2624 routes pointing to the RIB entry being removed (::/0). This is done 2625 recursively for 2001:db8::/32 and then for, 2001:db8:1::/48. The 2626 related FIB entries via S3 are removed, as illustrated in Figure 24. 2628 +---------+ +---------------+ +-----------------+ 2629 | Default | | 2001:db8::/32 | | 2001:db8:1::/48 | 2630 +---------+ +---------------+ +-----------------+ 2631 | | | 2632 | +--------+ | | 2633 +---> | Via S1 | | | 2634 | +--------+ | | 2635 | | | 2636 | +--------+ | +--------+ | 2637 +---> | Via S2 | +---> | Via S2 | | 2638 | +--------+ | +--------+ | 2639 | | | 2640 | | | 2641 | | | 2642 | | | 2643 | | | 2644 | +--------+ | +--------+ | +--------+ 2645 +---> | Via S4 | +---> | Via S4 | +---> | Via S4 | 2646 +--------+ +--------+ +--------+ 2648 Figure 24: Abstract FIB after loss of S3 2650 Say that at that time, S4 would also disaggregate prefix 2651 2001:db8:1::/48. This would mean that the FIB entry for 2652 2001:db8:1::/48 becomes a discard route, and that would be the signal 2653 for T1 to disaggregate prefix 2001:db8:1::/48 negatively in a 2654 transitive fashion with its own children. 2656 Finally, let us look at the case where S3 becomes available again as 2657 default gateway, and a negative advertisement is received from S4 2658 about prefix 2001:db8:2::/48 as opposed to 2001:db8:1::/48. Again, a 2659 negative route is stored in the RIB, and the more specific route to 2660 the complementing ToF nodes are installed in FIB. Since 2661 2001:db8:2::/48 inherits from 2001:db8::/32, the positive FIB routes 2662 are chosen by removing S4 from S2, S3, S4. The abstract FIB in T1 2663 now shows as illustrated in Figure 25: 2665 +-----------------+ 2666 | 2001:db8:2::/48 | 2667 +-----------------+ 2668 | 2669 +---------+ +---------------+ +-----------------+ 2670 | Default | | 2001:db8::/32 | | 2001:db8:1::/48 | 2671 +---------+ +---------------+ +-----------------+ 2672 | | | | 2673 | +--------+ | | | +--------+ 2674 +---> | Via S1 | | | +---> | Via S2 | 2675 | +--------+ | | | +--------+ 2676 | | | | 2677 | +--------+ | +--------+ | | +--------+ 2678 +---> | Via S2 | +---> | Via S2 | | +---> | Via S3 | 2679 | +--------+ | +--------+ | +--------+ 2680 | | | 2681 | +--------+ | +--------+ | +--------+ 2682 +---> | Via S3 | +---> | Via S3 | +---> | Via S3 | 2683 | +--------+ | +--------+ | +--------+ 2684 | | | 2685 | +--------+ | +--------+ | +--------+ 2686 +---> | Via S4 | +---> | Via S4 | +---> | Via S4 | 2687 +--------+ +--------+ +--------+ 2689 Figure 25: Abstract FIB after negative 2001:db8:2::/48 from S4 2691 5.2.7. Optional Zero Touch Provisioning (ZTP) 2693 Each RIFT node can optionally operate in zero touch provisioning 2694 (ZTP) mode, i.e. it has no configuration (unless it is a Top-of- 2695 Fabric at the top of the topology or the must operate in the topology 2696 as leaf and/or support leaf-2-leaf procedures) and it will fully 2697 configure itself after being attached to the topology. Configured 2698 nodes and nodes operating in ZTP can be mixed and will form a valid 2699 topology if achievable. 2701 The derivation of the level of each node happens based on offers 2702 received from its neighbors whereas each node (with possibly 2703 exceptions of configured leafs) tries to attach at the highest 2704 possible point in the fabric. This guarantees that even if the 2705 diffusion front reaches a node from "below" faster than from "above", 2706 it will greedily abandon already negotiated level derived from nodes 2707 topologically below it and properly peers with nodes above. 2709 The fabric is very conciously numbered from the top to allow for PoDs 2710 of different heights and minimize number of provisioning necessary, 2711 in this case just a TOP_OF_FABRIC flag on every node at the top of 2712 the fabric. 2714 This section describes the necessary concepts and procedures for ZTP 2715 operation. 2717 5.2.7.1. Terminology 2719 The interdependencies between the different flags and the configured 2720 level can be somewhat vexing at first and it may take multiple reads 2721 of the glossary to comprehend them. 2723 Automatic Level Derivation: Procedures which allow nodes without 2724 level configured to derive it automatically. Only applied if 2725 CONFIGURED_LEVEL is undefined. 2727 UNDEFINED_LEVEL: A "null" value that indicates that the level has 2728 not beeen determined and has not been configured. Schemas 2729 normally indicate that by a missing optional value without an 2730 available defined default. 2732 LEAF_ONLY: An optional configuration flag that can be configured on 2733 a node to make sure it never leaves the "bottom of the hierarchy". 2734 TOP_OF_FABRIC flag and CONFIGURED_LEVEL cannot be defined at the 2735 same time as this flag. It implies CONFIGURED_LEVEL value of 0. 2737 TOP_OF_FABRIC flag: Configuration flag that MUST be provided to all 2738 Top-of-Fabric nodes. LEAF_FLAG and CONFIGURED_LEVEL cannot be 2739 defined at the same time as this flag. It implies a 2740 CONFIGURED_LEVEL value. In fact, it is basically a shortcut for 2741 configuring same level at all Top-of-Fabric nodes which is 2742 unavoidable since an initial 'seed' is needed for other ZTP nodes 2743 to derive their level in the topology. The flag plays an 2744 important role in fabrics with multiple planes to enable 2745 successful negative disaggregation (Section 5.2.5.2). 2747 CONFIGURED_LEVEL: A level value provided manually. When this is 2748 defined (i.e. it is not an UNDEFINED_LEVEL) the node is not 2749 participating in ZTP. TOP_OF_FABRIC flag is ignored when this 2750 value is defined. LEAF_ONLY can be set only if this value is 2751 undefined or set to 0. 2753 DERIVED_LEVEL: Level value computed via automatic level derivation 2754 when CONFIGURED_LEVEL is equal to UNDEFINED_LEVEL. 2756 LEAF_2_LEAF: An optional flag that can be configured on a node to 2757 make sure it supports procedures defined in Section 5.3.9. In a 2758 strict sense it is a capability that implies LEAF_ONLY and the 2759 according restrictions. TOP_OF_FABRIC flag is ignored when set at 2760 the same time as this flag. 2762 LEVEL_VALUE: In ZTP case the original definition of "level" in 2763 Section 3.1 is both extended and relaxed. First, level is defined 2764 now as LEVEL_VALUE and is the first defined value of 2765 CONFIGURED_LEVEL followed by DERIVED_LEVEL. Second, it is 2766 possible for nodes to be more than one level apart to form 2767 adjacencies if any of the nodes is at least LEAF_ONLY. 2769 Valid Offered Level (VOL): A neighbor's level received on a valid 2770 LIE (i.e. passing all checks for adjacency formation while 2771 disregarding all clauses involving level values) persisting for 2772 the duration of the holdtime interval on the LIE. Observe that 2773 offers from nodes offering level value of 0 do not constitute VOLs 2774 (since no valid DERIVED_LEVEL can be obtained from those and 2775 consequently `not_a_ztp_offer` MUST be ignored). Offers from LIEs 2776 with `not_a_ztp_offer` being true are not VOLs either. If a node 2777 maintains parallel adjacencies to the neighbor, VOL on each 2778 adjacency is considered as equivalent, i.e. the newest VOL from 2779 any such adjacency updates the VOL received from the same node. 2781 Highest Available Level (HAL): Highest defined level value seen from 2782 all VOLs received. 2784 Highest Available Level Systems (HALS): Set of nodes offering HAL 2785 VOLs. 2787 Highest Adjacency Three Way (HAT): Highest neigbhor level of all the 2788 formed three way adjacencies for the node. 2790 5.2.7.2. Automatic SystemID Selection 2792 RIFT identifies each node via a SystemID which is a 64 bits wide 2793 integer. It is relatively simple to derive a, for all practical 2794 purposes collision free, value for each node on startup. For that 2795 purpose, a node MUST use as system ID EUI-64 MA-L format [EUI64] 2796 where the organizationally governed 24 bits can be used to generate 2797 system IDs for multiple RIFT instances running on the system. 2799 As matter of operational concern, the router MUST ensure that such 2800 identifier is not changing very frequently (or at least not without 2801 sending all its TIEs with fairly short lifetimes) since otherwise the 2802 network may be left with large amounts of stale TIEs in other nodes 2803 (though this is not necessarily a serious problem if the procedures 2804 described in Section 8 are implemented). 2806 5.2.7.3. Generic Fabric Example 2808 ZTP forces us to think about miscabled or unusually cabled fabric and 2809 how such a topology can be forced into a "lattice" structure which a 2810 fabric represents (with further restrictions). Let us consider a 2811 necessary and sufficient physical cabling in Figure 26. We assume 2812 all nodes being in the same PoD. 2814 . +---+ 2815 . | A | s = TOP_OF_FABRIC 2816 . | s | l = LEAF_ONLY 2817 . ++-++ l2l = LEAF_2_LEAF 2818 . | | 2819 . +--+ +--+ 2820 . | | 2821 . +--++ ++--+ 2822 . | E | | F | 2823 . | +-+ | +-----------+ 2824 . ++--+ | ++-++ | 2825 . | | | | | 2826 . | +-------+ | | 2827 . | | | | | 2828 . | | +----+ | | 2829 . | | | | | 2830 . ++-++ ++-++ | 2831 . | I +-----+ J | | 2832 . | | | +-+ | 2833 . ++-++ +--++ | | 2834 . | | | | | 2835 . +---------+ | +------+ | 2836 . | | | | | 2837 . +-----------------+ | | 2838 . | | | | | 2839 . ++-++ ++-++ | 2840 . | X +-----+ Y +-+ 2841 . |l2l| | l | 2842 . +---+ +---+ 2844 Figure 26: Generic ZTP Cabling Considerations 2846 First, we must anchor the "top" of the cabling and that's what the 2847 TOP_OF_FABRIC flag at node A is for. Then things look smooth until 2848 we have to decide whether node Y is at the same level as I, J or at 2849 the same level as Y and consequently, X is south of it. This is 2850 unresolvable here until we "nail down the bottom" of the topology. 2851 To achieve that we choose to use in this example the leaf flags. We 2852 will see further then whether Y chooses to form adjacencies to F or 2853 I, J successively. 2855 5.2.7.4. Level Determination Procedure 2857 A node starting up with UNDEFINED_VALUE (i.e. without a 2858 CONFIGURED_LEVEL or any leaf or TOP_OF_FABRIC flag) MUST follow those 2859 additional procedures: 2861 1. It advertises its LEVEL_VALUE on all LIEs (observe that this can 2862 be UNDEFINED_LEVEL which in terms of the schema is simply an 2863 omitted optional value). 2865 2. It computes HAL as numerically highest available level in all 2866 VOLs. 2868 3. It chooses then MAX(HAL-1,0) as its DERIVED_LEVEL. The node then 2869 starts to advertise this derived level. 2871 4. A node that lost all adjacencies with HAL value MUST hold down 2872 computation of new DERIVED_LEVEL for a short period of time 2873 unless it has no VOLs from southbound adjacencies. After the 2874 holddown expired, it MUST discard all received offers, recompute 2875 DERIVED_LEVEL and announce it to all neighbors. 2877 5. A node MUST reset any adjacency that has changed the level it is 2878 offering and is in three way state. 2880 6. A node that changed its defined level value MUST readvertise its 2881 own TIEs (since the new `PacketHeader` will contain a different 2882 level than before). Sequence number of each TIE MUST be 2883 increased. 2885 7. After a level has been derived the node MUST set the 2886 `not_a_ztp_offer` on LIEs towards all systems offering a VOL for 2887 HAL. 2889 8. A node that changed its level SHOULD flush from its link state 2890 database TIEs of all other nodes, otherwise stale information may 2891 persist on "direction reversal", i.e. nodes that seemed south 2892 are now north or east-west. This will not prevent the correct 2893 operation of the protocol but could be slightly confusing 2894 operationally. 2896 A node starting with LEVEL_VALUE being 0 (i.e. it assumes a leaf 2897 function by being configured with the appropriate flags or has a 2898 CONFIGURED_LEVEL of 0) MUST follow those additional procedures: 2900 1. It computes HAT per procedures above but does NOT use it to 2901 compute DERIVED_LEVEL. HAT is used to limit adjacency formation 2902 per Section 5.2.2. 2904 It MAY also follow modified procedures: 2906 1. It may pick a different strategy to choose VOL, e.g. use the VOL 2907 value with highest number of VOLs. Such strategies are only 2908 possible since the node always remains "at the bottom of the 2909 fabric" while another layer could "invert" the fabric by picking 2910 its prefered VOL in a different fashion than always trying to 2911 achieve the highest viable level. 2913 5.2.7.5. Resulting Topologies 2915 The procedures defined in Section 5.2.7.4 will lead to the RIFT 2916 topology and levels depicted in Figure 27. 2918 . +---+ 2919 . | As| 2920 . | 24| 2921 . ++-++ 2922 . | | 2923 . +--+ +--+ 2924 . | | 2925 . +--++ ++--+ 2926 . | E | | F | 2927 . | 23+-+ | 23+-----------+ 2928 . ++--+ | ++-++ | 2929 . | | | | | 2930 . | +-------+ | | 2931 . | | | | | 2932 . | | +----+ | | 2933 . | | | | | 2934 . ++-++ ++-++ | 2935 . | I +-----+ J | | 2936 . | 22| | 22| | 2937 . ++--+ +--++ | 2938 . | | | 2939 . +---------+ | | 2940 . | | | 2941 . ++-++ +---+ | 2942 . | X | | Y +-+ 2943 . | 0 | | 0 | 2944 . +---+ +---+ 2946 Figure 27: Generic ZTP Topology Autoconfigured 2948 In case we imagine the LEAF_ONLY restriction on Y is removed the 2949 outcome would be very different however and result in Figure 28. 2950 This demonstrates basically that auto configuration makes miscabling 2951 detection hard and with that can lead to undesirable effects in cases 2952 where leafs are not "nailed" by the accordingly configured flags and 2953 arbitrarily cabled. 2955 A node MAY analyze the outstanding level offers on its interfaces and 2956 generate warnings when its internal ruleset flags a possible 2957 miscabling. As an example, when a node's sees ZTP level offers that 2958 differ by more than one level from its chosen level (with proper 2959 accounting for leaf's being at level 0) this can indicate miscabling. 2961 . +---+ 2962 . | As| 2963 . | 24| 2964 . ++-++ 2965 . | | 2966 . +--+ +--+ 2967 . | | 2968 . +--++ ++--+ 2969 . | E | | F | 2970 . | 23+-+ | 23+-------+ 2971 . ++--+ | ++-++ | 2972 . | | | | | 2973 . | +-------+ | | 2974 . | | | | | 2975 . | | +----+ | | 2976 . | | | | | 2977 . ++-++ ++-++ +-+-+ 2978 . | I +-----+ J +-----+ Y | 2979 . | 22| | 22| | 22| 2980 . ++-++ +--++ ++-++ 2981 . | | | | | 2982 . | +-----------------+ | 2983 . | | | 2984 . +---------+ | | 2985 . | | | 2986 . ++-++ | 2987 . | X +--------+ 2988 . | 0 | 2989 . +---+ 2991 Figure 28: Generic ZTP Topology Autoconfigured 2993 5.2.8. Stability Considerations 2995 The autoconfiguration mechanism computes a global maximum of levels 2996 by diffusion. The achieved equilibrium can be disturbed massively by 2997 all nodes with highest level either leaving or entering the domain 2998 (with some finer distinctions not explained further). It is 2999 therefore recommended that each node is multi-homed towards nodes 3000 with respective HAL offerings. Fortuntately, this is the natural 3001 state of things for the topology variants considered in RIFT. 3003 5.3. Further Mechanisms 3005 5.3.1. Overload Bit 3007 Overload Bit MUST be respected in all according reachability 3008 computations. A node with overload bit set SHOULD NOT advertise any 3009 reachability prefixes southbound except locally hosted ones. A node 3010 in overload SHOULD advertise all its locally hosted prefixes north 3011 and southbound. 3013 The leaf node SHOULD set the 'overload' bit on its node TIEs, since 3014 if the spine nodes were to forward traffic not meant for the local 3015 node, the leaf node does not have the topology information to prevent 3016 a routing/forwarding loop. 3018 5.3.2. Optimized Route Computation on Leafs 3020 Since the leafs do see only "one hop away" they do not need to run a 3021 full SPF but can simply gather prefix candidates from their neighbors 3022 and build the according routing table. 3024 A leaf will have no N-TIEs except its own and optionally from its 3025 East-West neighbors. A leaf will have S-TIEs from its neighbors. 3027 Instead of creating a network graph from its N-TIEs and neighbor's 3028 S-TIEs and then running an SPF, a leaf node can simply compute the 3029 minimum cost and next_hop_set to each leaf neighbor by examining its 3030 local adjacencies, determining bi-directionality from the associated 3031 N-TIE, and specifying the neighbor's next_hop_set set and cost from 3032 the minimum cost local adjacency to that neighbor. 3034 Then a leaf attaches prefixes as described in Section 5.2.6. 3036 5.3.3. Mobility 3038 It is a requirement for RIFT to maintain at the control plane a real 3039 time status of which prefix is attached to which port of which leaf, 3040 even in a context of mobility where the point of attachement may 3041 change several times in a subsecond period of time. 3043 There are two classical approaches to maintain such knowledge in an 3044 unambiguous fashion: 3046 time stamp: With this method, the infrastructure memorizes the 3047 precise time at which the movement is observed. One key advantage 3048 of this technique is that it has no dependency on the mobile 3049 device. One drawback is that the infrastructure must be precisely 3050 synchronized to be able to compare time stamps as observed by the 3051 various points of attachment, e.g., using the variation of the 3052 Precision Time Protocol (PTP) IEEE Std. 1588 [IEEEstd1588], 3053 [IEEEstd8021AS] designed for bridged LANs IEEE Std. 802.1AS 3054 [IEEEstd8021AS]. Both the precision of the synchronisation 3055 protocol and the resolution of the time stamp must beat the 3056 highest possible roaming time on the fabric. Another drawback is 3057 that the presence of the mobile device may be observed only 3058 asynchronously, e.g., after it starts using an IP protocol such as 3059 ARP [RFC0826], IPv6 Neighbor Discovery [RFC4861][RFC4862], or DHCP 3060 [RFC2131][RFC8415]. 3062 sequence counter: With this method, a mobile node notifies its point 3063 of attachment on arrival with a sequence counter that is 3064 incremented upon each movement. On the positive side, this method 3065 does not have a dependency on a precise sense of time, since the 3066 sequence of movements is kept in order by the device. The 3067 disadvantage of this approach is the lack of support for protocols 3068 that may be used by the mobile node to register its presence to 3069 the leaf node with the capability to provide a sequence counter. 3070 Well-known issues with wrapping sequence counters must be 3071 addressed properly, and many forms of sequence counters that vary 3072 in both wrapping rules and comparison rules. A particular 3073 knowledge of the source of the sequence counter is required to 3074 operate it, and the comparison between sequence counters from 3075 heterogeneous sources can be hard to impossible. 3077 RIFT supports a hybrid approach contained in an optional 3078 `PrefixSequenceType` prefix attribute that we call a `monotonic 3079 clock` consisting of a timestamp and optional sequence number. In 3080 case of presence of the attribute: 3082 o The leaf node MUST advertise a time stamp of the latest sighting 3083 of a prefix, e.g., by snooping IP protocols or the node using the 3084 time at which it advertised the prefix. RIFT transports the time 3085 stamp within the desired prefix N-TIEs as 802.1AS timestamp. 3087 o RIFT may interoperate with the "update to 6LoWPAN Neighbor 3088 Discovery" [RFC8505], which provides a method for registering a 3089 prefix with a sequence counter called a Transaction ID (TID). 3090 RIFT transports in such case the TID in its native form. 3092 o RIFT also defines an abstract negative clock (ANSC) that compares 3093 as less than any other clock. By default, the lack of a 3094 `PrefixSequenceType` in a Prefix N-TIE is interpreted as ANSC. We 3095 call this also an `undefined` clock. 3097 o Any prefix present on the fabric in multiple nodes that has the 3098 `same` clock is considered as anycast. ASNC is always considered 3099 smaller than any defined clock. 3101 o RIFT implementation assumes by default that all nodes are being 3102 synchronized to 200 milliseconds precision which is easily 3103 achievable even in very large fabrics using [RFC5905]. An 3104 implementation MAY provide a way to reconfigure a domain to a 3105 different value. We call this variable MAXIMUM_CLOCK_DELTA. 3107 5.3.3.1. Clock Comparison 3109 All monotonic clock values are comparable to each other using the 3110 following rules: 3112 1. ASNC is older than any other value except ASNC AND 3114 2. Clock with timestamp differing by more than MAXIMUM_CLOCK_DELTA 3115 are comparable by using the timestamps only AND 3117 3. Clocks with timestamps differing by less than MAXIMUM_CLOCK_DELTA 3118 are comparable by using their TIDs only AND 3120 4. An undefined TID is always older than any other TID AND 3122 5. TIDs are compared using rules of [RFC8505]. 3124 5.3.3.2. Interaction between Time Stamps and Sequence Counters 3126 For slow movements that occur less frequently than e.g. once per 3127 second, the time stamp that the RIFT infrastruture captures is enough 3128 to determine the freshest discovery. If the point of attachement 3129 changes faster than the maximum drift of the time stamping mechanism 3130 (i.e. MAXIMUM_CLOCK_DELTA), then a sequence counter is required to 3131 add resolution to the freshness evaluation, and it must be sized so 3132 that the counters stay comparable within the resolution of the time 3133 stampling mechanism. 3135 The sequence counter in [RFC8505] is encoded as one octet and wraps 3136 around using Appendix A. 3138 Within the resolution of MAXIMUM_CLOCK_DELTA the sequence counters 3139 captured during 2 sequential values of the time stamp SHOULD be 3140 comparable. This means with default values that a node may move up 3141 to 127 times during a 200 milliseconds period and the clocks remain 3142 still comparable thus allowing the infrastructure to assert the 3143 freshest advertisement with no ambiguity. 3145 5.3.3.3. Anycast vs. Unicast 3147 A unicast prefix can be attached to at most one leaf, whereas an 3148 anycast prefix may be reachable via more than one leaf. 3150 If a monotonic clock attribute is provided on the prefix, then the 3151 prefix with the `newest` clock value is strictly prefered. An 3152 anycast prefix does not carry a clock or all clock attributes MUST be 3153 the same under the rules of Section 5.3.3.1. 3155 Observe that it is important that in mobility events the leaf is re- 3156 flooding as quickly as possible the absence of the prefix that moved 3157 away. 3159 Observe further that without support for [RFC8505] movements on the 3160 fabric within intervals smaller than 100msec will be seen as anycast. 3162 5.3.3.4. Overlays and Signaling 3164 RIFT is agnostic whether any overlay technology like [MIP, LISP, 3165 VxLAN, NVO3] and the associated signaling is deployed over it. But 3166 it is expected that leaf nodes, and possibly Top-of-Fabric nodes can 3167 perform the according encapsulation. 3169 In the context of mobility, overlays provide a classical solution to 3170 avoid injecting mobile prefixes in the fabric and improve the 3171 scalability of the solution. It makes sense on a data center that 3172 already uses overlays to consider their applicability to the mobility 3173 solution; as an example, a mobility protocol such as LISP may inform 3174 the ingress leaf of the location of the egress leaf in real time. 3176 Another possibility is to consider that mobility as an underlay 3177 service and support it in RIFT to an extent. The load on the fabric 3178 augments with the amount of mobility obviously since a move forces 3179 flooding and computation on all nodes in the scope of the move so 3180 tunneling from leaf to the Top-of-Fabric may be desired. Future 3181 versions of this document may describe support for such tunneling in 3182 RIFT. 3184 5.3.4. Key/Value Store 3186 5.3.4.1. Southbound 3188 The protocol supports a southbound distribution of key-value pairs 3189 that can be used to e.g. distribute configuration information during 3190 topology bring-up. The KV S-TIEs can arrive from multiple nodes and 3191 hence need tie-breaking per key. We use the following rules 3193 1. Only KV TIEs originated by nodes to which the receiver has a bi- 3194 directional adjacency are considered. 3196 2. Within all such valid KV S-TIEs containing the key, the value of 3197 the KV S-TIE for which the according node S-TIE is present, has 3198 the highest level and within the same level has highest 3199 originating system ID is preferred. If keys in the most 3200 preferred TIEs are overlapping, the behavior is undefined. 3202 Observe that if a node goes down, the node south of it looses 3203 adjacencies to it and with that the KVs will be disregarded and on 3204 tie-break changes new KV re-advertised to prevent stale information 3205 being used by nodes further south. KV information in southbound 3206 direction is not result of independent computation of every node over 3207 same set of TIEs but a diffused computation. 3209 5.3.4.2. Northbound 3211 Certain use cases seem to necessitate distribution of essentialy KV 3212 information that is generated in the leafs in the northbound 3213 direction. Such information is flooded in KV N-TIEs. Since the 3214 originator of northbound KV is preserved during northbound flooding, 3215 overlapping keys could be used. However, to omit further protocol 3216 complexity, only the value of the key in TIE tie-broken in same 3217 fashion as southbound KV TIEs is used. 3219 5.3.5. Interactions with BFD 3221 RIFT MAY incorporate BFD [RFC5881] to react quickly to link failures. 3222 In such case following procedures are introduced: 3224 After RIFT three way hello adjacency convergence a BFD session MAY 3225 be formed automatically between the RIFT endpoints without further 3226 configuration using the exchanged discriminators. The capability 3227 of the remote side to support BFD is carried on the LIEs. 3229 In case established BFD session goes Down after it was Up, RIFT 3230 adjacency should be re-initialized started from Init. 3232 In case of parallel links between nodes each link may run its own 3233 independent BFD session or they may share a session. 3235 In case RIFT changes link identifiers or BFD capability indication 3236 both the LIE as well as the BFD sessions SHOULD be brought down 3237 and back up again. 3239 Multiple RIFT instances MAY choose to share a single BFD session 3240 (in such case it is undefined what discriminators are used albeit 3241 RIFT CAN advertise the same link ID for the same interface in 3242 multiple instances and with that "share" the discriminators). 3244 BFD TTL follows [RFC5082]. 3246 5.3.6. Fabric Bandwidth Balancing 3248 A well understood problem in fabrics is that in case of link losses 3249 it would be ideal to rebalance how much traffic is offered to 3250 switches in the next level based on the ingress and egress bandwidth 3251 they have. Current attempts rely mostly on specialized traffic 3252 engineering via controller or leafs being aware of complete topology 3253 with according cost and complexity. 3255 RIFT can support a very light weight mechanism that can deal with the 3256 problem in an approximate way based on the fact that RIFT is loop- 3257 free. 3259 5.3.6.1. Northbound Direction 3261 Every RIFT node SHOULD compute the amount of northbound bandwith 3262 available through neighbors at higher level and modify distance 3263 received on default route from this neighbor. Those different 3264 distances SHOULD be used to support weighted ECMP forwarding towards 3265 higher level when using default route. We call such a distance 3266 Bandwidth Adjusted Distance or BAD. This is best illustrated by a 3267 simple example. 3269 . 100 x 100 100 MBits 3270 . | x | | 3271 . +-+---+-+ +-+---+-+ 3272 . | | | | 3273 . |Spin111| |Spin112| 3274 . +-+---+++ ++----+++ 3275 . |x || || || 3276 . || |+---------------+ || 3277 . || +---------------+| || 3278 . || || || || 3279 . || || || || 3280 . -----All Links 10 MBit------- 3281 . || || || || 3282 . || || || || 3283 . || +------------+| || || 3284 . || |+------------+ || || 3285 . |x || || || 3286 . +-+---+++ +--++-+++ 3287 . | | | | 3288 . |Leaf111| |Leaf112| 3289 . +-------+ +-------+ 3291 Figure 29: Balancing Bandwidth 3293 All links from Leafs in Figure 29 are assumed to 10 MBit/s bandwidth 3294 while the uplinks one level further up are assumed to be 100 MBit/s. 3295 Further, in Figure 29 we assume that Leaf111 lost one of the parallel 3296 links to Spine 111 and with that wants to possibly push more traffic 3297 onto Spine 112. Leaf 112 has equal bandwidth to Spine 111 and Spine 3298 112 but Spine 111 lost one of its uplinks. 3300 The local modification of the received default route distance from 3301 upper level is achieved by running a relatively simple algorithm 3302 where the bandwidth is weighted exponentially while the distance on 3303 the default route represents a multiplier for the bandwidth weight 3304 for easy operational adjustements. 3306 On a node L use Node TIEs to compute for each non-overloaded 3307 northbound neighbor N three values: 3309 L_N_u: as sum of the bandwidth available to N 3311 N_u: as sum of the uplink bandwidth available on N 3313 T_N_u: as sum of L_N_u * OVERSUBSCRIPTION_CONSTANT + N_u 3315 For all T_N_u determine the according M_N_u as 3316 log_2(next_power_2(T_N_u)) and determine MAX_M_N_u as maximum value 3317 of all M_N_u. 3319 For each advertised default route from a node N modify the advertised 3320 distance D to BAD = D * (1 + MAX_M_N_u - M_N_u) and use BAD instead 3321 of distance D to weight balance default forwarding towards N. 3323 For the example above a simple table of values will help the 3324 understanding. We assume the default route distance is advertised 3325 with D=1 everywhere and OVERSUBSCRIPTION_CONSTANT = 1. 3327 +---------+-----------+-------+-------+-----+ 3328 | Node | N | T_N_u | M_N_u | BAD | 3329 +---------+-----------+-------+-------+-----+ 3330 | Leaf111 | Spine 111 | 110 | 7 | 2 | 3331 +---------+-----------+-------+-------+-----+ 3332 | Leaf111 | Spine 112 | 220 | 8 | 1 | 3333 +---------+-----------+-------+-------+-----+ 3334 | Leaf112 | Spine 111 | 120 | 7 | 2 | 3335 +---------+-----------+-------+-------+-----+ 3336 | Leaf112 | Spine 112 | 220 | 8 | 1 | 3337 +---------+-----------+-------+-------+-----+ 3339 Table 5: BAD Computation 3341 All the multiplications and additions are saturating, i.e. when 3342 exceeding range of the bandwidth type are set to highest possible 3343 value of the type. 3345 BAD is only computed for default routes. A node MAY compute and use 3346 BAD for any disaggregated prefixes or other RIFT routes. A node MAY 3347 use another algorithm than BAD to weight northbound traffic based on 3348 bandwidth given that the algorithm is distributed and un-synchronized 3349 and ultimately, its correct behavior does not depend on uniformity of 3350 balancing algorithms used in the fabric. E.g. it is conceivable that 3351 leafs could use real time link loads gathered by analytics to change 3352 the amount of traffic assigned to each default route next hop. 3354 Observe further that a change in available bandwidth will only affect 3355 at maximum two levels down in the fabric, i.e. blast radius of 3356 bandwidth changes is contained no matter its height. 3358 5.3.6.2. Southbound Direction 3360 Due to its loop free properties a node CAN take during S-SPF into 3361 account the available bandwidth on the nodes in lower levels and 3362 modify the amount of traffic offered to next level's "southbound" 3363 nodes based as what it sees is the total achievable maximum flow 3364 through those nodes. It is worth observing that such computations 3365 may work better if standardized but does not have to be necessarily. 3366 As long the packet keeps on heading south it will take one of the 3367 available paths and arrive at the intended destination. 3369 5.3.7. Label Binding 3371 A node MAY advertise on its TIEs a locally significant, downstream 3372 assigned label for the according interface. One use of such label is 3373 a hop-by-hop encapsulation allowing to easily distinguish forwarding 3374 planes served by a multiplicity of RIFT instances. 3376 5.3.8. Segment Routing Support with RIFT 3378 Recently, alternative architecture to reuse labels as segment 3379 identifiers [RFC8402] has gained traction and may present use cases 3380 in IP fabric that would justify its deployment. Such use cases will 3381 either precondition an assignment of a label per node (or other 3382 entities where the mechanisms are equivalent) or a global assignment 3383 and a knowledge of topology everywhere to compute segment stacks of 3384 interest. We deal with the two issues separately. 3386 5.3.8.1. Global Segment Identifiers Assignment 3388 Global segment identifiers are normally assumed to be provided by 3389 some kind of a centralized "controller" instance and distributed to 3390 other entities. This can be performed in RIFT by attaching a 3391 controller to the Top-of-Fabric nodes at the top of the fabric where 3392 the whole topology is always visible, assign such identifiers and 3393 then distribute those via the KV mechanism towards all nodes so they 3394 can perform things like probing the fabric for failures using a stack 3395 of segments. 3397 5.3.8.2. Distribution of Topology Information 3399 Some segment routing use cases seem to precondition full knowledge of 3400 fabric topology in all nodes which can be performed albeit at the 3401 loss of one of highly desirable properties of RIFT, namely minimal 3402 blast radius. Basically, RIFT can function as a flat IGP by 3403 switching off its flooding scopes. All nodes will end up with full 3404 topology view and albeit the N-SPF and S-SPF are still performed 3405 based on RIFT rules, any computation with segment identifiers that 3406 needs full topology can use it. 3408 Beside blast radius problem, excessive flooding may present 3409 significant load on implementations. 3411 5.3.9. Leaf to Leaf Procedures 3413 RIFT can optionally allow special leaf East-West adjacencies under 3414 additional set of rules. The leaf supporting those procedures MUST: 3416 advertise the LEAF_2_LEAF flag in node capabilities AND 3418 set the overload bit on all leaf's node TIEs AND 3420 flood only node's own north and south TIEs over E-W leaf 3421 adjacencies AND 3423 always use E-W leaf adjacency in both north as well as south 3424 computation AND 3426 install a discard route for any advertised aggregate in leaf's 3427 TIEs AND 3429 never form southbound adjacencies. 3431 This will allow the E-W leaf nodes to exchange traffic strictly for 3432 the prefixes advertised in each other's north prefix TIEs (since the 3433 southbound computation will find the reverse direction in the other 3434 node's TIE and install its north prefixes). 3436 5.3.10. Address Family and Multi Topology Considerations 3438 Multi-Topology (MT)[RFC5120] and Multi-Instance (MI)[RFC8202] is used 3439 today in link-state routing protocols to support several domains on 3440 the same physical topology. RIFT supports this capability by 3441 carrying transport ports in the LIE protocol exchanges. Multiplexing 3442 of LIEs can be achieved by either choosing varying multicast 3443 addresses or ports on the same address. 3445 BFD interactions in Section 5.3.5 are implementation dependent when 3446 multiple RIFT instances run on the same link. 3448 5.3.11. Reachability of Internal Nodes in the Fabric 3450 RIFT does not precondition that its nodes have reachable addresses 3451 albeit for operational purposes this is clearly desirable. Under 3452 normal operating conditions this can be easily achieved by e.g. 3453 injecting the node's loopback address into North and South Prefix 3454 TIEs or other implementation specific mechanisms. 3456 Things get more interesting in case a node looses all its northbound 3457 adjacencies but is not at the top of the fabric. That is outside the 3458 scope of this document and may be covered in a separate document 3459 about policy guided prefixes [PGP reference]. 3461 5.3.12. One-Hop Healing of Levels with East-West Links 3463 Based on the rules defined in Section 5.2.4, Section 5.2.3.8 and 3464 given presence of E-W links, RIFT can provide a one-hop protection of 3465 nodes that lost all their northbound links or in other complex link 3466 set failure scenarios except at Top-of-Fabric where the links are 3467 used exclusively to flood topology information in multi-plane 3468 designs. Section 6.4 explains the resulting behavior based on one 3469 such example. 3471 5.4. Security 3473 5.4.1. Security Model 3475 An inherent property of any security and ZTP architecture is the 3476 resulting trade-off in regard to integrity verification of the 3477 information distributed through the fabric vs. necessary provisioning 3478 and auto-configuration. At a minimum, in all approaches, the 3479 security of an established adjacency can be ensured. The stricter 3480 the security model the more provisioning must take over the role of 3481 ZTP. 3483 The most security conscious operators will want to have full control 3484 over which port on which router/switch is connected to the respective 3485 port on the "other side", which we will call the "port-association 3486 model" (PAM) achievable e.g. by configuring on each port pair a 3487 designated shared key or pair of private/public keys. In secure data 3488 center locations, operators may want to control which router/switch 3489 is connected to which other router/switch only or choose a "node- 3490 association model" (NAM) which allows, for example, simplified port 3491 sparing. In an even more relaxed environment, an operator may only 3492 be concerned that the router/switches share credentials ensuring that 3493 they belong to this particular data center network hence allowing the 3494 flexible sparing of whole routers/switches. We will define that case 3495 as the "fabric-association model" (FAM), equivalent to using a shared 3496 secret for the whole fabric. Such flexibility may make sense for 3497 leaf nodes such as servers where the addition and swapping of servers 3498 is more frequent than the rest of the data center network. 3499 Generally, leafs of the fabric tend to be less trusted than switches. 3500 The different models could be mixed throughout the fabric if the 3501 benefits outweigh the cost of increased complexity in provisioning. 3503 In each of the above cases, some configuration mechanism is needed to 3504 allow the operator to specify which connections are allowed, and some 3505 mechanism is needed to: 3507 a. specify the according level in the fabric, 3509 b. discover and report missing connections, 3511 c. discover and report unexpected connections, and prevent such 3512 adjacencies from forming. 3514 On the more relaxed configuration side of the spectrum, operators 3515 might only configure the level of each switch, but don't explicitly 3516 configure which connections are allowed. In this case, RIFT will 3517 only allow adjacencies to come up between nodes are that in adjacent 3518 levels. The operators with lowest security requirements may not use 3519 any configuration to specify which connections are allowed. Such 3520 fabrics could rely fully on ZTP for each router/switch to discover 3521 its level and would only allow adjacencies between adjacent levels to 3522 come up. Figure 30 illustrates the tradeoffs inherent in the 3523 different security models. 3525 Ultimately, some level of verification of the link quality may be 3526 required before an adjacency is allowed to be used for forwarding. 3527 For example, an implementation may require that a BFD session comes 3528 up before advertising the adjacency. 3530 For the above outlined cases, RIFT has two approaches to enforce that 3531 a local port is connected to the correct port on the correct remote 3532 router/switch. One approach is to piggy-back on RIFT's 3533 authentication mechanism. Assuming the provisioning model (e.g. the 3534 YANG model) is flexible enough, operators can choose to provision a 3535 unique authentication key for: 3537 a. each pair of ports in "port-association model" or 3539 b. each pair of switches in "node-association model" or 3541 c. each pair of levels or 3543 d. the entire fabric in "fabric-association model". 3545 The other approach is to rely on the system-id, port-id and level 3546 fields in the LIE message to validate an adjacency against the 3547 configured expected cabling topology, and optionally introduce some 3548 new rules in the FSM to allow the adjacency to come up if the 3549 expectations are met. 3551 ^ /\ | 3552 /|\ / \ | 3553 | / \ | 3554 | / PAM \ | 3555 Increasing / \ Increasing 3556 Integrity +----------+ Flexibility 3557 & / NAM \ & 3558 Increasing +--------------+ Less 3559 Provisioning / FAM \ Configuration 3560 | +------------------+ | 3561 | / Level Provisioning \ | 3562 | +----------------------+ \|/ 3563 | / Zero Configuration \ v 3564 +--------------------------+ 3566 Figure 30: Security Model 3568 5.4.2. Security Mechanisms 3570 RIFT Security goals are to ensure authentication, message integrity 3571 and prevention of replay attacks. Low processing overhead and 3572 efficient messaging are also a goal. Message confidentiality is a 3573 non-goal. 3575 The model in the previous section allows a range of security key 3576 types that are analogous to the various security association models. 3577 PAM and NAM allow security associations at the port or node level 3578 using symmetric or asymmetric keys that are pre-installed. FAM 3579 argues for security associations to be applied only at a group level 3580 or to be refined once the topology has been established. RIFT does 3581 not specify how security keys are installed or updated it specifies 3582 how the key can be used to achieve goals. 3584 The protocol has provisions for "weak" nonces to prevent replay 3585 attacks and includes authentication mechanisms comparable to 3586 [RFC5709] and [RFC7987]. 3588 5.4.3. Security Envelope 3590 RIFT MUST be carried in a mandatory secure envelope illustrated in 3591 Figure 31. Any value in the packet following a security fingerprint 3592 MUST be used only after the according fingerprint has been validated. 3594 Local configuration MAY allow to skip the checking of the envelope's 3595 integrity. 3597 0 1 2 3 3598 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 3600 UDP Header: 3601 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 3602 | Source Port | RIFT destination port | 3603 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 3604 | UDP Length | UDP Checksum | 3605 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 3607 Outer Security Envelope Header: 3608 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 3609 | RIFT MAGIC | Packet Number | 3610 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 3611 | Reserved | RIFT Major | Outer Key ID | Fingerprint | 3612 | | Version | | Length | 3613 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 3614 | | 3615 ~ Security Fingerprint covers all following content ~ 3616 | | 3617 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 3618 | Weak Nonce Local | Weak Nonce Remote | 3619 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 3620 | Remaining TIE Lifetime (all 1s in case of LIE) | 3621 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 3623 TIE Origin Security Envelope Header: 3624 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 3625 | TIE Origin Key ID | Fingerprint | 3626 | | Length | 3627 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 3628 | | 3629 ~ Security Fingerprint covers all following content ~ 3630 | | 3631 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 3633 Serialized RIFT Model Object 3634 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 3635 | | 3636 ~ Serialized RIFT Model Object ~ 3637 | | 3638 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 3640 Figure 31: Security Envelope 3642 RIFT MAGIC: 16 bits. Constant value of 0xA1F7 that allows to 3643 classify RIFT packets independent of used UDP port. 3645 Packet Number: 16 bits. An optional, per packet type monotonically 3646 growing number rolling over using sequence number arithmetic 3647 defined inAppendix A. A node SHOULD correctly set the number on 3648 subsequent packets or otherwise MUST set the value to 3649 `undefined_packet_number` as provided in the schema. This number 3650 can be used to detect losses and misordering in flooding for 3651 either operational purposes or in implementation to adjust 3652 flooding behavior to current link or buffer quality. This number 3653 MUST NOT be used to discard or validate the correctness of 3654 packets. 3656 RIFT Major Version: 8 bits. It allows to check whether protocol 3657 versions are compatible, i.e. the serialized object can be decoded 3658 at all. An implementation MUST drop packets with unexpected value 3659 and MAY report a problem. Must be same as in encoded model 3660 object, otherwise packet is dropped. 3662 Outer Key ID: 8 bits to allow key rollovers. This implies key type 3663 and used algorithm. Value 0 means that no valid fingerprint was 3664 computed. This key ID scope is local to the nodes on both ends of 3665 the adjacency. 3667 TIE Origin Key ID: 24 bits. This implies key type and used 3668 algorithm. Value 0 means that no valid fingerprint was computed. 3669 This key ID scope is global to the RIFT instance since it implies 3670 the originator of the TIE so the contained object does not have to 3671 be de-serialized to obtain it. 3673 Length of Fingerprint: 8 bits. Length in 32-bit multiples of the 3674 following fingerprint not including lifetime or weak nonces. It 3675 allows to navigate the structure when an unknown key type is 3676 present. To clarify a common cornercase when this value is set to 3677 0 it signifies an empty (0 bytes long) security fingerprint. 3679 Security Fingerprint: 32 bits * Length of Fingerprint. This is a 3680 signature that is computed over all data following after it. If 3681 the signficant bits of fingerprint are fewer than the 32 bits 3682 padded length than the signficant bits MUST be left aligned and 3683 remaining bits on the right padded with 0s. When using PKI the 3684 Security fingerprint originating node uses its private key to 3685 create the signature. The original packet can then be verified 3686 provided the public key is shared and current. 3688 Remaining TIE Lifetime: 32 bits. In case of anything but TIEs this 3689 field MUST be set to all ones and Origin Security Envelope Header 3690 MUST NOT be present in the packet. For TIEs this field represents 3691 the remaining lifetime of the TIE and Origin Security Envelope 3692 Header MUST be present in the packet. The value in the serialized 3693 model object MUST be ignored. 3695 Weak Nonce Local: 16 bits. Local Weak Nonce of the adjacency as 3696 advertised in LIEs. 3698 Weak Nonce Remote: 16 bits. Remote Weak Nonce of the adjacency as 3699 received in LIEs. 3701 TIE Origin Security Envelope Header: It MUST be present if and only 3702 if the Remaining TIE Lifetime field is NOT all ones. It carries 3703 through the originators key ID and according fingerprint of the 3704 object to protect TIE from modification during flooding. This 3705 ensures origin validation and integrity (but does not provide 3706 validation of a chain of trust). 3708 Observe that due to the schema migration rules per Appendix B the 3709 contained model can be always decoded if the major version matches 3710 and the envelope integrity has been validated. Consequently, 3711 description of the TIE is available to flood it properly including 3712 unknown TIE types. 3714 5.4.4. Weak Nonces 3716 The protocol uses two 16 bit nonces to salt generated signatures. We 3717 use the term "nonce" a bit loosely since RIFT nonces are not being 3718 changed on every packet as common in cryptography. For efficiency 3719 purposes they are changed at a frequency high enough to dwarf replay 3720 attacks attempts for all practical purposes. Therefore, we call them 3721 "weak" nonces. 3723 Any implementation including RIFT security MUST generate and wrap 3724 around local nonces properly. When a nonce increment leads to 3725 `undefined_nonce` value the value SHOULD be incremented again 3726 immediately. All implementation MUST reflect the neighbor's nonces. 3727 An implementation SHOULD increment a chosen nonce on every LIE FSM 3728 transition that ends up in a different state from the previous and 3729 MUST increment its nonce at least every 5 minutes (such 3730 considerations allow for efficient implementations without opening a 3731 significant security risk). When flooding TIEs, the implementation 3732 MUST use recent (i.e. within allowed difference) nonces reflected in 3733 the LIE exchange. The schema specifies maximum allowable nonce value 3734 difference on a packet compared to reflected nonces in the LIEs. Any 3735 packet received with nonces deviating more than the allowed delta 3736 MUST be discarded without further computation of signatures to 3737 prevent computation load attacks. 3739 In case where a secure implementation does not receive signatures or 3740 receives undefined nonces from neighbor indicating that it does not 3741 support or verify signatures, it is a matter of local policy how such 3742 packets are treated. Any secure implementation MUST discard packets 3743 where its local nonce is not correctly mirrored but it may choose to 3744 either refuse forming an adjacency with an implementation not 3745 advertising signatures or valid nonces or simply keep on signing 3746 local packets while accepting neighbor's packets without further 3747 verification beside checking for proper nonce reflection. 3749 As a necessary exception, an implementation MUST advertise 3750 `undefined_nonce` for remote nonce value when the FSM is not in 2-way 3751 or 3-way state and accept an `undefined_nonce` for its local nonce 3752 value on packets in any other state than 3-way. 3754 As optional optimization, an implemenation MAY send one LIE with 3755 previously negotiated neighbor's nonce to try to speed up a 3756 neighbor's transition from 3-way to 1-way and MUST revert to sending 3757 `undefined_nonce` after that. 3759 5.4.5. Lifetime 3761 Protecting lifetime on flooding may lead to excessive number of 3762 security fingerprint computation and hence an application generating 3763 such fingerprints on TIEs MAY round the value down to the next 3764 `rounddown_lifetime_interval` defined in the schema when sending TIEs 3765 albeit such optimization in presence of security hashes over 3766 advancing weak nonces may not be feasible. 3768 5.4.6. Key Management 3770 As outlined in the Security Model a private shared key or a public/ 3771 private key pair is used to Authenticate the adjacency. The actual 3772 method of key distribution and key synchronization is assumed to be 3773 out of band from RIFT's perspective. Both nodes in the adjacency 3774 must share the same keys and configuration of key type and algorithm 3775 for a key ID. Mismatched keys will obviously not inter-operate due 3776 to unverifiable security envelope. 3778 Key roll-over while the adjacency is active is allowed and the 3779 technique is well known and described in e.g. [RFC6518]. Key 3780 distribution procedures are out of scope for RIFT. 3782 5.4.7. Security Association Changes 3784 There in no mechanism to convert a security envelope for the same key 3785 ID from one algorithm to another once the envelope is operational. 3786 The recommended procedure to change to a new algorithm is to take the 3787 adjacency down and make the changes and then bring the adjacency up. 3788 Obviously, an implementation may choose to stop verifying security 3789 envelope for the duration of key change to keep the adjacency up but 3790 since this introduces a security vulnerability window, such roll-over 3791 is not recommended. 3793 6. Examples 3795 6.1. Normal Operation 3797 This section describes RIFT deployment in the example topology 3798 without any node or link failures. We disregard flooding reduction 3799 for simplicity's sake. 3801 As first step, the following bi-directional adjacencies will be 3802 created (and any other links that do not fulfill LIE rules in 3803 Section 5.2.2 disregarded): 3805 1. Spine 21 (PoD 0) to Spine 111, Spine 112, Spine 121, and Spine 3806 122 3808 2. Spine 22 (PoD 0) to Spine 111, Spine 112, Spine 121, and Spine 3809 122 3811 3. Spine 111 to Leaf 111, Leaf 112 3813 4. Spine 112 to Leaf 111, Leaf 112 3815 5. Spine 121 to Leaf 121, Leaf 122 3817 6. Spine 122 to Leaf 121, Leaf 122 3819 Consequently, N-TIEs would be originated by Spine 111 and Spine 112 3820 and each set would be sent to both Spine 21 and Spine 22. N-TIEs 3821 also would be originated by Leaf 111 (w/ Prefix 111) and Leaf 112 (w/ 3822 Prefix 112 and the multi-homed prefix) and each set would be sent to 3823 Spine 111 and Spine 112. Spine 111 and Spine 112 would then flood 3824 these N-TIEs to Spine 21 and Spine 22. 3826 Similarly, N-TIEs would be originated by Spine 121 and Spine 122 and 3827 each set would be sent to both Spine 21 and Spine 22. N-TIEs also 3828 would be originated by Leaf 121 (w/ Prefix 121 and the multi-homed 3829 prefix) and Leaf 122 (w/ Prefix 122) and each set would be sent to 3830 Spine 121 and Spine 122. Spine 121 and Spine 122 would then flood 3831 these N-TIEs to Spine 21 and Spine 22. 3833 At this point both Spine 21 and Spine 22, as well as any controller 3834 to which they are connected, would have the complete network 3835 topology. At the same time, Spine 111/112/121/122 hold only the 3836 N-ties of level 0 of their respective PoD. Leafs hold only their own 3837 N-TIEs. 3839 S-TIEs with adjacencies and a default IP prefix would then be 3840 originated by Spine 21 and Spine 22 and each would be flooded to 3841 Spine 111, Spine 112, Spine 121, and Spine 122. Spine 111, Spine 3842 112, Spine 121, and Spine 122 would each send the S-TIE from Spine 21 3843 to Spine 22 and the S-TIE from Spine 22 to Spine 21. (S-TIEs are 3844 reflected up to level from which they are received but they are NOT 3845 propagated southbound.) 3847 A S-TIE with a default IP prefix would be originated by Node 111 and 3848 Spine 112 and each would be sent to Leaf 111 and Leaf 112. 3850 Similarly, an S-TIE with a default IP prefix would be originated by 3851 Node 121 and Spine 122 and each would be sent to Leaf 121 and Leaf 3852 122. At this point IP connectivity with maximum possible ECMP has 3853 been established between the leafs while constraining the amount of 3854 information held by each node to the minimum necessary for normal 3855 operation and dealing with failures. 3857 6.2. Leaf Link Failure 3859 . | | | | 3860 .+-+---+-+ +-+---+-+ 3861 .| | | | 3862 .|Spin111| |Spin112| 3863 .+-+---+-+ ++----+-+ 3864 . | | | | 3865 . | +---------------+ X 3866 . | | | X Failure 3867 . | +-------------+ | X 3868 . | | | | 3869 .+-+---+-+ +--+--+-+ 3870 .| | | | 3871 .|Leaf111| |Leaf112| 3872 .+-------+ +-------+ 3873 . + + 3874 . Prefix111 Prefix112 3876 Figure 32: Single Leaf link failure 3878 In case of a failing leaf link between spine 112 and leaf 112 the 3879 link-state information will cause re-computation of the necessary SPF 3880 and the higher levels will stop forwarding towards prefix 112 through 3881 spine 112. Only spines 111 and 112, as well as both spines will see 3882 control traffic. Leaf 111 will receive a new S-TIE from spine 112 3883 and reflect back to spine 111. Spine 111 will de-aggregate prefix 3884 111 and prefix 112 but we will not describe it further here since de- 3885 aggregation is emphasized in the next example. It is worth observing 3886 however in this example that if leaf 111 would keep on forwarding 3887 traffic towards prefix 112 using the advertised south-bound default 3888 of spine 112 the traffic would end up on Top-of-Fabric 21 and ToF 22 3889 and cross back into pod 1 using spine 111. This is arguably not as 3890 bad as black-holing present in the next example but clearly 3891 undesirable. Fortunately, de-aggregation prevents this type of 3892 behavior except for a transitory period of time. 3894 6.3. Partitioned Fabric 3895 . +--------+ +--------+ S-TIE of Spine21 3896 . | | | | received by 3897 . |ToF 21| |ToF 22| south reflection of 3898 . ++-+--+-++ ++-+--+-++ spines 112 and 111 3899 . | | | | | | | | 3900 . | | | | | | | 0/0 3901 . | | | | | | | | 3902 . | | | | | | | | 3903 . +--------------+ | +--- XXXXXX + | | | +---------------+ 3904 . | | | | | | | | 3905 . | +-----------------------------+ | | | 3906 . 0/0 | | | | | | | 3907 . | 0/0 0/0 +- XXXXXXXXXXXXXXXXXXXXXXXXX -+ | 3908 . | 1.1/16 | | | | | | 3909 . | | +-+ +-0/0-----------+ | | 3910 . | | | 1.1./16 | | | | 3911 .+-+----++ +-+-----+ ++-----0/0 ++----0/0 3912 .| | | | | 1.1/16 | 1.1/16 3913 .|Spin111| |Spin112| |Spin121| |Spin122| 3914 .+-+---+-+ ++----+-+ +-+---+-+ ++---+--+ 3915 . | | | | | | | | 3916 . | +---------------+ | | +----------------+ | 3917 . | | | | | | | | 3918 . | +-------------+ | | | +--------------+ | | 3919 . | | | | | | | | 3920 .+-+---+-+ +--+--+-+ +-+---+-+ +---+-+-+ 3921 .| | | | | | | | 3922 .|Leaf111| |Leaf112| |Leaf121| |Leaf122| 3923 .+-+-----+ ++------+ +-----+-+ +-+-----+ 3924 . + + + + 3925 . Prefix111 Prefix112 Prefix121 Prefix122 3926 . 1.1/16 3928 Figure 33: Fabric partition 3930 Figure 33 shows the arguably a more catastrophic but also a more 3931 interesting case. Spine 21 is completely severed from access to 3932 Prefix 121 (we use in the figure 1.1/16 as example) by double link 3933 failure. However unlikely, if left unresolved, forwarding from leaf 3934 111 and leaf 112 to prefix 121 would suffer 50% black-holing based on 3935 pure default route advertisements by Top-of-Fabric 21 and ToF 22. 3937 The mechanism used to resolve this scenario is hinging on the 3938 distribution of southbound representation by Top-of-Fabric 21 that is 3939 reflected by spine 111 and spine 112 to ToF 22. Spine 22, having 3940 computed reachability to all prefixes in the network, advertises with 3941 the default route the ones that are reachable only via lower level 3942 neighbors that ToF 21 does not show an adjacency to. That results in 3943 spine 111 and spine 112 obtaining a longest-prefix match to prefix 3944 121 which leads through ToF 22 and prevents black-holing through ToF 3945 21 still advertising the 0/0 aggregate only. 3947 The prefix 121 advertised by Top-of-Fabric 22 does not have to be 3948 propagated further towards leafs since they do no benefit from this 3949 information. Hence the amount of flooding is restricted to ToF 21 3950 reissuing its S-TIEs and south reflection of those by spine 111 and 3951 spine 112. The resulting SPF in ToF 22 issues a new prefix S-TIEs 3952 containing 1.1/16. None of the leafs become aware of the changes and 3953 the failure is constrained strictly to the level that became 3954 partitioned. 3956 To finish with an example of the resulting sets computed using 3957 notation introduced in Section 5.2.5, Top-of-Fabric 22 constructs the 3958 following sets: 3960 |R = Prefix 111, Prefix 112, Prefix 121, Prefix 122 3962 |H (for r=Prefix 111) = Spine 111, Spine 112 3964 |H (for r=Prefix 112) = Spine 111, Spine 112 3966 |H (for r=Prefix 121) = Spine 121, Spine 122 3968 |H (for r=Prefix 122) = Spine 121, Spine 122 3970 |A (for Spine 21) = Spine 111, Spine 112 3972 With that and |H (for r=prefix 121) and |H (for r=prefix 122) being 3973 disjoint from |A (for Top-of-Fabric 21), ToF 22 will originate an 3974 S-TIE with prefix 121 and prefix 122, that is flooded to spines 112, 3975 112, 121 and 122. 3977 6.4. Northbound Partitioned Router and Optional East-West Links 3978 . + + + 3979 . X N1 | N2 | N3 3980 . X | | 3981 .+--+----+ +--+----+ +--+-----+ 3982 .| |0/0> <0/0| |0/0> <0/0| | 3983 .| A01 +----------+ A02 +----------+ A03 | Level 1 3984 .++-+-+--+ ++--+--++ +---+-+-++ 3985 . | | | | | | | | | 3986 . | | +----------------------------------+ | | | 3987 . | | | | | | | | | 3988 . | +-------------+ | | | +--------------+ | 3989 . | | | | | | | | | 3990 . | +----------------+ | +-----------------+ | 3991 . | | | | | | | | | 3992 . | | +------------------------------------+ | | 3993 . | | | | | | | | | 3994 .++-+-+--+ | +---+---+ | +-+---+-++ 3995 .| | +-+ +-+ | | 3996 .| L01 | | L02 | | L03 | Level 0 3997 .+-------+ +-------+ +--------+ 3999 Figure 34: North Partitioned Router 4001 Figure 34 shows a part of a fabric where level 1 is horizontally 4002 connected and A01 lost its only northbound adjacency. Based on N-SPF 4003 rules in Section 5.2.4.1 A01 will compute northbound reachability by 4004 using the link A01 to A02 (whereas A02 will NOT use this link during 4005 N-SPF). Hence A01 will still advertise the default towards level 0 4006 and route unidirectionally using the horizontal link. 4008 As further consideration, the moment A02 looses link N2 the situation 4009 evolves again. A01 will have no more northbound reachability while 4010 still seeing A03 advertising northbound adjacencies in its south node 4011 tie. With that it will stop advertising a default route due to 4012 Section 5.2.3.8. 4014 6.5. Multi-Plane Fabric and Negative Disaggregation 4016 TODO 4018 7. Implementation and Operation: Further Details 4020 7.1. Considerations for Leaf-Only Implementation 4022 RIFT can and is intended to be stretched to the lowest level in the 4023 IP fabric to integrate ToRs or even servers. Since those entities 4024 would run as leafs only, it is worth to observe that a leaf only 4025 version is significantly simpler to implement and requires much less 4026 resources: 4028 1. Under normal conditions, the leaf needs to support a multipath 4029 default route only. In most catastrophic partitioning case it 4030 has to be capable of accommodating all the leaf routes in its own 4031 PoD to prevent black-holing. 4033 2. Leaf nodes hold only their own N-TIEs and S-TIEs of Level 1 nodes 4034 they are connected to; so overall few in numbers. 4036 3. Leaf node does not have to support any type of de-aggregation 4037 computation or propagation. 4039 4. Leaf nodes do not have to support overload bit normally. 4041 5. Unless optional leaf-2-leaf procedures are desired default route 4042 origination and S-TIE origination is unnecessary. 4044 7.2. Considerations for Spine Implementation 4046 In case of spines, i.e. nodes that will never act as Top of Fabric a 4047 full implementation is not required, specifically the node does not 4048 need to perform any computation of negative disaggregation except 4049 respecting northbound disaggregation advertised from the north. 4051 7.3. Adaptations to Other Proposed Data Center Topologies 4052 . +-----+ +-----+ 4053 . | | | | 4054 .+-+ S0 | | S1 | 4055 .| ++---++ ++---++ 4056 .| | | | | 4057 .| | +------------+ | 4058 .| | | +------------+ | 4059 .| | | | | 4060 .| ++-+--+ +--+-++ 4061 .| | | | | 4062 .| | A0 | | A1 | 4063 .| +-+--++ ++---++ 4064 .| | | | | 4065 .| | +------------+ | 4066 .| | +-----------+ | | 4067 .| | | | | 4068 .| +-+-+-+ +--+-++ 4069 .+-+ | | | 4070 . | L0 | | L1 | 4071 . +-----+ +-----+ 4073 Figure 35: Level Shortcut 4075 Strictly speaking, RIFT is not limited to Clos variations only. The 4076 protocol preconditions only a sense of 'compass rose direction' 4077 achieved by configuration (or derivation) of levels and other 4078 topologies are possible within this framework. So, conceptually, one 4079 could include leaf to leaf links and even shortcut between levels but 4080 certain requirements in Section 4 will not be met anymore. As an 4081 example, shortcutting levels illustrated in Figure 35 will lead 4082 either to suboptimal routing when L0 sends traffic to L1 (since using 4083 S0's default route will lead to the traffic being sent back to A0 or 4084 A1) or the leafs need each other's routes installed to understand 4085 that only A0 and A1 should be used to talk to each other. 4087 Whether such modifications of topology constraints make sense is 4088 dependent on many technology variables and the exhausting treatment 4089 of the topic is definitely outside the scope of this document. 4091 7.4. Originating Non-Default Route Southbound 4093 Obviously, an implementation may choose to originate southbound 4094 instead of a strict default route (as described in Section 5.2.3.8) a 4095 shorter prefix P' but in such a scenario all addresses carried within 4096 the RIFT domain must be contained within P'. 4098 8. Security Considerations 4100 8.1. General 4102 One can consider attack vectors where a router may reboot many times 4103 while changing its system ID and pollute the network with many stale 4104 TIEs or TIEs are sent with very long lifetimes and not cleaned up 4105 when the routes vanishes. Those attack vectors are not unique to 4106 RIFT. Given large memory footprints available today those attacks 4107 should be relatively benign. Otherwise a node SHOULD implement a 4108 strategy of discarding contents of all TIEs that were not present in 4109 the SPF tree over a certain, configurable period of time. Since the 4110 protocol, like all modern link-state protocols, is self-stabilizing 4111 and will advertise the presence of such TIEs to its neighbors, they 4112 can be re-requested again if a computation finds that it sees an 4113 adjacency formed towards the system ID of the discarded TIEs. 4115 8.2. ZTP 4117 Section 5.2.7 presents many attack vectors in untrusted environments, 4118 starting with nodes that oscillate their level offers to the 4119 possiblity of a node offering a three way adjacency with the highest 4120 possible level value with a very long holdtime trying to put itself 4121 "on top of the lattice" and with that gaining access to the whole 4122 southbound topology. Session authentication mechanisms are necessary 4123 in environments where this is possible and RIFT provides the 4124 according security envelope to ensure this if desired. 4126 8.3. Lifetime 4128 Traditional IGP protocols are vulnerable to lifetime modification and 4129 replay attacks that can be somewhat mitigated by using techniques 4130 like [RFC7987]. RIFT removes this attack vector by protecting the 4131 lifetime behind a signature computed over it and additional nonce 4132 combination which makes even the replay attack window very small and 4133 for practical purposes irrelevant since lifetime cannot be 4134 artificially shortened by the attacker. 4136 8.4. Packet Number 4138 Optional packet number is carried in the security envelope without 4139 any encryption protection and is hence vulnerable to replay and 4140 modification attacks. Contrary to nonces this number must change on 4141 every packet and would present a very high cryptographic load if 4142 signed. The attack vector packet number present is relatively 4143 benign. Changing the packet number by a man-in-the-middle attack 4144 will only affect operational validation tools and possibly some 4145 performance optimizations on flooding. It is expected that an 4146 implementation detecting too many "fake losses" or "misorderings" due 4147 to the attack on the packet number would simply suppress its further 4148 processing. 4150 8.5. Outer Fingerprint Attacks 4152 A node can try to inject LIE packets observing a conversation on the 4153 wire by using the outer key ID albeit it cannot generate valid hashes 4154 in case it changes the integrity of the message so the only possible 4155 attack is DoS due to excessive LIE validation. 4157 A node can try to replay previous LIEs with changed state that it 4158 recorded but the attack is hard to replicate since the nonce 4159 combination must match the ongoing exchange and is then limited to a 4160 single flap only since both nodes will advance their nonces in case 4161 the adjacency state changed. Even in the most unlikely case the 4162 attack length is limited due to both sides periodically increasing 4163 their nonces. 4165 8.6. TIE Origin Fingerprint DoS Attacks 4167 A compromised node can attempt to generate "fake TIEs" using other 4168 nodes' TIE origin key identifiers. Albeit the ultimate validation of 4169 the origin fingerprint will fail in such scenarios and not progress 4170 further than immediately peering nodes, the resulting denial of 4171 service attack seems unavoidable since the TIE origin key id is only 4172 protected by the, here assumed to be compromised, node. 4174 9. IANA Considerations 4176 This specification will request at an opportune time multiple 4177 registry points to exchange protocol packets in a standardized way, 4178 amongst them multicast address assignments and standard port numbers. 4179 The schema itself defines many values and codepoints which can be 4180 considered registries themselves. 4182 10. Acknowledgments 4184 A new routing protocol in its complexity is not a product of a parent 4185 but of a village as the author list shows already. However, many 4186 more people provided input, fine-combed the specification based on 4187 their experience in design or implementation. This section will make 4188 an inadequate attempt in recording their contribution. 4190 Many thanks to Naiming Shen for some of the early discussions around 4191 the topic of using IGPs for routing in topologies related to Clos. 4192 Russ White to be especially acknowledged for the key conversation on 4193 epistomology that allowed to tie current asynchronous distributed 4194 systems theory results to a modern protocol design presented here. 4195 Adrian Farrel, Joel Halpern, Jeffrey Zhang, Krzysztof Szarkowicz, 4196 Nagendra Kumar provided thoughtful comments that improved the 4197 readability of the document and found good amount of corners where 4198 the light failed to shine. Kris Price was first to mention single 4199 router, single arm default considerations. Jeff Tantsura helped out 4200 with some initial thoughts on BFD interactions while Jeff Haas 4201 corrected several misconceptions about BFD's finer points. Artur 4202 Makutunowicz pointed out many possible improvements and acted as 4203 sounding board in regard to modern protocol implementation techniques 4204 RIFT is exploring. Barak Gafni formalized first time clearly the 4205 problem of partitioned spine and fallen leafs on a (clean) napkin in 4206 Singapore that led to the very important part of the specification 4207 centered around multiple Top-of-Fabric planes and negative 4208 disaggregation. Igor Gashinsky and others shared many thoughts on 4209 problems encountered in design and operation of large-scale data 4210 center fabrics. Xu Benchong found a delicate error in the flooding 4211 procedures while implementing. 4213 11. References 4215 11.1. Normative References 4217 [ISO10589] 4218 ISO "International Organization for Standardization", 4219 "Intermediate system to Intermediate system intra-domain 4220 routeing information exchange protocol for use in 4221 conjunction with the protocol for providing the 4222 connectionless-mode Network Service (ISO 8473), ISO/IEC 4223 10589:2002, Second Edition.", Nov 2002. 4225 [RFC1982] Elz, R. and R. Bush, "Serial Number Arithmetic", RFC 1982, 4226 DOI 10.17487/RFC1982, August 1996, 4227 . 4229 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 4230 Requirement Levels", BCP 14, RFC 2119, 4231 DOI 10.17487/RFC2119, March 1997, 4232 . 4234 [RFC2328] Moy, J., "OSPF Version 2", STD 54, RFC 2328, 4235 DOI 10.17487/RFC2328, April 1998, 4236 . 4238 [RFC2365] Meyer, D., "Administratively Scoped IP Multicast", BCP 23, 4239 RFC 2365, DOI 10.17487/RFC2365, July 1998, 4240 . 4242 [RFC4271] Rekhter, Y., Ed., Li, T., Ed., and S. Hares, Ed., "A 4243 Border Gateway Protocol 4 (BGP-4)", RFC 4271, 4244 DOI 10.17487/RFC4271, January 2006, 4245 . 4247 [RFC4291] Hinden, R. and S. Deering, "IP Version 6 Addressing 4248 Architecture", RFC 4291, DOI 10.17487/RFC4291, February 4249 2006, . 4251 [RFC5082] Gill, V., Heasley, J., Meyer, D., Savola, P., Ed., and C. 4252 Pignataro, "The Generalized TTL Security Mechanism 4253 (GTSM)", RFC 5082, DOI 10.17487/RFC5082, October 2007, 4254 . 4256 [RFC5120] Przygienda, T., Shen, N., and N. Sheth, "M-ISIS: Multi 4257 Topology (MT) Routing in Intermediate System to 4258 Intermediate Systems (IS-ISs)", RFC 5120, 4259 DOI 10.17487/RFC5120, February 2008, 4260 . 4262 [RFC5303] Katz, D., Saluja, R., and D. Eastlake 3rd, "Three-Way 4263 Handshake for IS-IS Point-to-Point Adjacencies", RFC 5303, 4264 DOI 10.17487/RFC5303, October 2008, 4265 . 4267 [RFC5709] Bhatia, M., Manral, V., Fanto, M., White, R., Barnes, M., 4268 Li, T., and R. Atkinson, "OSPFv2 HMAC-SHA Cryptographic 4269 Authentication", RFC 5709, DOI 10.17487/RFC5709, October 4270 2009, . 4272 [RFC5881] Katz, D. and D. Ward, "Bidirectional Forwarding Detection 4273 (BFD) for IPv4 and IPv6 (Single Hop)", RFC 5881, 4274 DOI 10.17487/RFC5881, June 2010, 4275 . 4277 [RFC5905] Mills, D., Martin, J., Ed., Burbank, J., and W. Kasch, 4278 "Network Time Protocol Version 4: Protocol and Algorithms 4279 Specification", RFC 5905, DOI 10.17487/RFC5905, June 2010, 4280 . 4282 [RFC7752] Gredler, H., Ed., Medved, J., Previdi, S., Farrel, A., and 4283 S. Ray, "North-Bound Distribution of Link-State and 4284 Traffic Engineering (TE) Information Using BGP", RFC 7752, 4285 DOI 10.17487/RFC7752, March 2016, 4286 . 4288 [RFC7987] Ginsberg, L., Wells, P., Decraene, B., Przygienda, T., and 4289 H. Gredler, "IS-IS Minimum Remaining Lifetime", RFC 7987, 4290 DOI 10.17487/RFC7987, October 2016, 4291 . 4293 [RFC8200] Deering, S. and R. Hinden, "Internet Protocol, Version 6 4294 (IPv6) Specification", STD 86, RFC 8200, 4295 DOI 10.17487/RFC8200, July 2017, 4296 . 4298 [RFC8202] Ginsberg, L., Previdi, S., and W. Henderickx, "IS-IS 4299 Multi-Instance", RFC 8202, DOI 10.17487/RFC8202, June 4300 2017, . 4302 [RFC8402] Filsfils, C., Ed., Previdi, S., Ed., Ginsberg, L., 4303 Decraene, B., Litkowski, S., and R. Shakir, "Segment 4304 Routing Architecture", RFC 8402, DOI 10.17487/RFC8402, 4305 July 2018, . 4307 [RFC8505] Thubert, P., Ed., Nordmark, E., Chakrabarti, S., and C. 4308 Perkins, "Registration Extensions for IPv6 over Low-Power 4309 Wireless Personal Area Network (6LoWPAN) Neighbor 4310 Discovery", RFC 8505, DOI 10.17487/RFC8505, November 2018, 4311 . 4313 11.2. Informative References 4315 [CLOS] Yuan, X., "On Nonblocking Folded-Clos Networks in Computer 4316 Communication Environments", IEEE International Parallel & 4317 Distributed Processing Symposium, 2011. 4319 [DIJKSTRA] 4320 Dijkstra, E., "A Note on Two Problems in Connexion with 4321 Graphs", Journal Numer. Math. , 1959. 4323 [DOT] Ellson, J. and L. Koutsofios, "Graphviz: open source graph 4324 drawing tools", Springer-Verlag , 2001. 4326 [DYNAMO] De Candia et al., G., "Dynamo: amazon's highly available 4327 key-value store", ACM SIGOPS symposium on Operating 4328 systems principles (SOSP '07), 2007. 4330 [EPPSTEIN] 4331 Eppstein, D., "Finding the k-Shortest Paths", 1997. 4333 [EUI64] IEEE, "Guidelines for Use of Extended Unique Identifier 4334 (EUI), Organizationally Unique Identifier (OUI), and 4335 Company ID (CID)", IEEE EUI, 4336 . 4338 [FATTREE] Leiserson, C., "Fat-Trees: Universal Networks for 4339 Hardware-Efficient Supercomputing", 1985. 4341 [IEEEstd1588] 4342 IEEE, "IEEE Standard for a Precision Clock Synchronization 4343 Protocol for Networked Measurement and Control Systems", 4344 IEEE Standard 1588, 4345 . 4347 [IEEEstd8021AS] 4348 IEEE, "IEEE Standard for Local and Metropolitan Area 4349 Networks - Timing and Synchronization for Time-Sensitive 4350 Applications in Bridged Local Area Networks", 4351 IEEE Standard 802.1AS, 4352 . 4354 [ISO10589-Second-Edition] 4355 International Organization for Standardization, 4356 "Intermediate system to Intermediate system intra-domain 4357 routeing information exchange protocol for use in 4358 conjunction with the protocol for providing the 4359 connectionless-mode Network Service (ISO 8473)", Nov 2002. 4361 [MAKSIC2013] 4362 Maksic et al., N., "Improving Utilization of Data Center 4363 Networks", IEEE Communications Magazine, Nov 2013. 4365 [RFC0826] Plummer, D., "An Ethernet Address Resolution Protocol: Or 4366 Converting Network Protocol Addresses to 48.bit Ethernet 4367 Address for Transmission on Ethernet Hardware", STD 37, 4368 RFC 826, DOI 10.17487/RFC0826, November 1982, 4369 . 4371 [RFC2131] Droms, R., "Dynamic Host Configuration Protocol", 4372 RFC 2131, DOI 10.17487/RFC2131, March 1997, 4373 . 4375 [RFC3626] Clausen, T., Ed. and P. Jacquet, Ed., "Optimized Link 4376 State Routing Protocol (OLSR)", RFC 3626, 4377 DOI 10.17487/RFC3626, October 2003, 4378 . 4380 [RFC4655] Farrel, A., Vasseur, J., and J. Ash, "A Path Computation 4381 Element (PCE)-Based Architecture", RFC 4655, 4382 DOI 10.17487/RFC4655, August 2006, 4383 . 4385 [RFC4861] Narten, T., Nordmark, E., Simpson, W., and H. Soliman, 4386 "Neighbor Discovery for IP version 6 (IPv6)", RFC 4861, 4387 DOI 10.17487/RFC4861, September 2007, 4388 . 4390 [RFC4862] Thomson, S., Narten, T., and T. Jinmei, "IPv6 Stateless 4391 Address Autoconfiguration", RFC 4862, 4392 DOI 10.17487/RFC4862, September 2007, 4393 . 4395 [RFC6518] Lebovitz, G. and M. Bhatia, "Keying and Authentication for 4396 Routing Protocols (KARP) Design Guidelines", RFC 6518, 4397 DOI 10.17487/RFC6518, February 2012, 4398 . 4400 [RFC7855] Previdi, S., Ed., Filsfils, C., Ed., Decraene, B., 4401 Litkowski, S., Horneffer, M., and R. Shakir, "Source 4402 Packet Routing in Networking (SPRING) Problem Statement 4403 and Requirements", RFC 7855, DOI 10.17487/RFC7855, May 4404 2016, . 4406 [RFC7938] Lapukhov, P., Premji, A., and J. Mitchell, Ed., "Use of 4407 BGP for Routing in Large-Scale Data Centers", RFC 7938, 4408 DOI 10.17487/RFC7938, August 2016, 4409 . 4411 [RFC8415] Mrugalski, T., Siodelski, M., Volz, B., Yourtchenko, A., 4412 Richardson, M., Jiang, S., Lemon, T., and T. Winters, 4413 "Dynamic Host Configuration Protocol for IPv6 (DHCPv6)", 4414 RFC 8415, DOI 10.17487/RFC8415, November 2018, 4415 . 4417 [VAHDAT08] 4418 Al-Fares, M., Loukissas, A., and A. Vahdat, "A Scalable, 4419 Commodity Data Center Network Architecture", SIGCOMM , 4420 2008. 4422 [Wikipedia] 4423 Wikipedia, 4424 "https://en.wikipedia.org/wiki/Serial_number_arithmetic", 4425 2016. 4427 Appendix A. Sequence Number Binary Arithmetic 4429 The only reasonably reference to a cleaner than [RFC1982] sequence 4430 number solution is given in [Wikipedia]. It basically converts the 4431 problem into two complement's arithmetic. Assuming a straight two 4432 complement's substractions on the bit-width of the sequence number 4433 the according >: and =: relations are defined as: 4435 U_1, U_2 are 12-bits aligned unsigned version number 4437 D_f is ( U_1 - U_2 ) interpreted as two complement signed 12-bits 4438 D_b is ( U_2 - U_1 ) interpreted as two complement signed 12-bits 4440 U_1 >: U_2 IIF D_f > 0 AND D_b < 0 4441 U_1 =: U_2 IIF D_f = 0 4443 The >: relationsship is symmetric but not transitive. Observe that 4444 this leaves the case of the numbers having maximum two complement 4445 distance, e.g. ( 0 and 0x800 ) undefined in our 12-bits case since 4446 D_f and D_b are both -0x7ff. 4448 A simple example of the relationship in case of 3-bit arithmetic 4449 follows as table indicating D_f/D_b values and then the relationship 4450 of U_1 to U_2: 4452 U2 / U1 0 1 2 3 4 5 6 7 4453 0 +/+ +/- +/- +/- -/- -/+ -/+ -/+ 4454 1 -/+ +/+ +/- +/- +/- -/- -/+ -/+ 4455 2 -/+ -/+ +/+ +/- +/- +/- -/- -/+ 4456 3 -/+ -/+ -/+ +/+ +/- +/- +/- -/- 4457 4 -/- -/+ -/+ -/+ +/+ +/- +/- +/- 4458 5 +/- -/- -/+ -/+ -/+ +/+ +/- +/- 4459 6 +/- +/- -/- -/+ -/+ -/+ +/+ +/- 4460 7 +/- +/- +/- -/- -/+ -/+ -/+ +/+ 4462 U2 / U1 0 1 2 3 4 5 6 7 4463 0 = > > > ? < < < 4464 1 < = > > > ? < < 4465 2 < < = > > > ? < 4466 3 < < < = > > > ? 4467 4 ? < < < = > > > 4468 5 > ? < < < = > > 4469 6 > > ? < < < = > 4470 7 > > > ? < < < = 4472 Appendix B. Information Elements Schema 4474 This section introduces the schema for information elements. 4476 On schema changes that 4478 1. change field numbers or 4480 2. add new *required* fields or 4482 3. remove any fields or 4484 4. change lists into sets, unions into structures or 4486 5. change multiplicity of fields or 4488 6. changes name of any field or type or 4490 7. change datatypes of any field or 4492 8. adds, changes or removes a default value of any *existing* field 4493 or 4495 9. removes or changes any defined constant or constant value or 4497 10. changes any enumeration type except extending `common.TIEType` 4498 (use of enumeration types is generally discouraged) 4500 major version of the schema MUST increase. All other changes MUST 4501 increase minor version within the same major. 4503 Observe however that introducing an optional field does not cause a 4504 major version increase even if the fields inside the structure are 4505 optional with defaults. 4507 All signed integer as forced by Thrift support must be cast for 4508 internal purposes to equivalent unsigned values without discarding 4509 the signedness bit. An implementation SHOULD try to avoid using the 4510 signedness bit when generating values. 4512 The schema is normative. 4514 B.1. common.thrift 4516 /** 4517 Thrift file with common definitions for RIFT 4518 */ 4519 /** @note MUST be interpreted in implementation as unsigned 64 bits. 4520 * The implementation SHOULD NOT use the MSB. 4521 */ 4522 typedef i64 SystemIDType 4523 typedef i32 IPv4Address 4524 /** this has to be of length long enough to accomodate prefix */ 4525 typedef binary IPv6Address 4526 /** @note MUST be interpreted in implementation as unsigned */ 4527 typedef i16 UDPPortType 4528 /** @note MUST be interpreted in implementation as unsigned */ 4529 typedef i32 TIENrType 4530 /** @note MUST be interpreted in implementation as unsigned */ 4531 typedef i32 MTUSizeType 4532 /** @note MUST be interpreted in implementation as unsigned rollling over number */ 4533 typedef i16 SeqNrType 4534 /** @note MUST be interpreted in implementation as unsigned */ 4535 typedef i32 LifeTimeInSecType 4536 /** @note MUST be interpreted in implementation as unsigned */ 4537 typedef i8 LevelType 4538 /** optional, recommended monotonically increasing number _per packet type per adjacency_ 4539 that can be used to detect losses/misordering/restarts. 4540 This will be moved into envelope in the future. 4541 @note MUST be interpreted in implementation as unsigned rollling over number */ 4542 typedef i16 PacketNumberType 4543 /** @note MUST be interpreted in implementation as unsigned */ 4544 typedef i32 PodType 4545 /** @note MUST be interpreted in implementation as unsigned. This is carried in the 4546 security envelope and MUST fit into 8 bits. */ 4547 typedef i8 VersionType 4548 /** @note MUST be interpreted in implementation as unsigned */ 4549 typedef i16 MinorVersionType 4550 /** @note MUST be interpreted in implementation as unsigned */ 4551 typedef i32 MetricType 4552 /** @note MUST be interpreted in implementation as unsigned and unstructured */ 4553 typedef i64 RouteTagType 4554 /** @note MUST be interpreted in implementation as unstructured label value */ 4555 typedef i32 LabelType 4556 /** @note MUST be interpreted in implementation as unsigned */ 4557 typedef i32 BandwithInMegaBitsType 4558 /** @note Key Value key ID type */ 4559 typedef string KeyIDType 4560 /** node local, unique identification for a link (interface/tunnel 4561 * etc. Basically anything RIFT runs on). This is kept 4562 * at 32 bits so it aligns with BFD [RFC5880] discriminator size. 4563 */ 4564 typedef i32 LinkIDType 4565 typedef string KeyNameType 4566 typedef i8 PrefixLenType 4567 /** timestamp in seconds since the epoch */ 4568 typedef i64 TimestampInSecsType 4569 /** security nonce. 4570 * @note MUST be interpreted in implementation as rolling over unsigned value */ 4571 typedef i16 NonceType 4572 /** LIE FSM holdtime type */ 4573 typedef i16 TimeIntervalInSecType 4574 /** Transaction ID type for prefix mobility as specified by RFC6550, value 4575 MUST be interpreted in implementation as unsigned */ 4576 typedef i8 PrefixTransactionIDType 4577 /** timestamp per IEEE 802.1AS, values MUST be interpreted in implementation as unsigned */ 4578 struct IEEE802_1ASTimeStampType { 4579 1: required i64 AS_sec; 4580 2: optional i32 AS_nsec; 4581 } 4582 /** generic counter type */ 4583 typedef i64 CounterType 4584 /** Platform Interface Index type, i.e. index of interface on hardware, can be used e.g. with 4585 RFC5837 */ 4586 typedef i32 PlatformInterfaceIndex 4588 /** Flags indicating nodes behavior in case of ZTP and support 4589 for special optimization procedures. It will force level to `leaf_level` or 4590 `top-of-fabric` level accordingly and enable according procedures 4591 */ 4592 enum HierarchyIndications { 4593 leaf_only = 0, 4594 leaf_only_and_leaf_2_leaf_procedures = 1, 4595 top_of_fabric = 2, 4596 } 4598 const PacketNumberType undefined_packet_number = 0 4599 /** This MUST be used when node is configured as top of fabric in ZTP. 4600 This is kept reasonably low to alow for fast ZTP convergence on 4601 failures. */ 4602 const LevelType top_of_fabric_level = 24 4603 /** default bandwidth on a link */ 4604 const BandwithInMegaBitsType default_bandwidth = 100 4605 /** fixed leaf level when ZTP is not used */ 4606 const LevelType leaf_level = 0 4607 const LevelType default_level = leaf_level 4608 const PodType default_pod = 0 4609 const LinkIDType undefined_linkid = 0 4611 /** default distance used */ 4612 const MetricType default_distance = 1 4613 /** any distance larger than this will be considered infinity */ 4614 const MetricType infinite_distance = 0x7FFFFFFF 4615 /** represents invalid distance */ 4616 const MetricType invalid_distance = 0 4617 const bool overload_default = false 4618 const bool flood_reduction_default = true 4619 /** default LIE FSM holddown time */ 4620 const TimeIntervalInSecType default_lie_holdtime = 3 4621 /** default ZTP FSM holddown time */ 4622 const TimeIntervalInSecType default_ztp_holdtime = 1 4623 /** by default LIE levels are ZTP offers */ 4624 const bool default_not_a_ztp_offer = false 4625 /** by default e'one is repeating flooding */ 4626 const bool default_you_are_flood_repeater = true 4627 /** 0 is illegal for SystemID */ 4628 const SystemIDType IllegalSystemID = 0 4629 /** empty set of nodes */ 4630 const set empty_set_of_nodeids = {} 4631 /** default lifetime of TIE is one week */ 4632 const LifeTimeInSecType default_lifetime = 604800 4633 /** default lifetime when TIEs are purged is 5 minutes */ 4634 const LifeTimeInSecType purge_lifetime = 300 4635 /** round down interval when TIEs are sent with security hashes 4636 to prevent excessive computation. **/ 4637 const LifeTimeInSecType rounddown_lifetime_interval = 60 4638 /** any `TieHeader` that has a smaller lifetime difference 4639 than this constant is equal (if other fields equal). This 4640 constant MUST be larger than `purge_lifetime` to avoid 4641 retransmissions */ 4642 const LifeTimeInSecType lifetime_diff2ignore = 400 4644 /** default UDP port to run LIEs on */ 4645 const UDPPortType default_lie_udp_port = 911 4646 /** default UDP port to receive TIEs on, that can be peer specific */ 4647 const UDPPortType default_tie_udp_flood_port = 912 4649 /** default MTU link size to use */ 4650 const MTUSizeType default_mtu_size = 1400 4651 /** default link being BFD capable */ 4652 const bool bfd_default = true 4654 /** undefined nonce, equivalent to missing nonce */ 4655 const NonceType undefined_nonce = 0; 4656 /** outer security key id */ 4657 typedef i8 OuterSecurityKeyID 4658 /** outer security key id */ 4659 typedef i32 InnerSecurityKeyID 4660 /** security key id */ 4661 typedef i32 TIESecurityKeyID 4662 /** undefined key */ 4663 const TIESecurityKeyID undefined_securitykey_id = 0; 4664 /** Maximum delta (negative or positive) that a mirrored nonce can 4665 deviate from local value to be considered valid. If nonces are 4666 changed every minute on both sides this opens statistically 4667 a `maximum_valid_nonce_delta` minutes window of identical LIEs, 4668 TIE, TI(x)E replays. 4669 The interval cannot be too small since LIE FSM may change 4670 states fairly quickly during ZTP without sending LIEs*/ 4671 const i16 maximum_valid_nonce_delta = 5; 4673 /** indicates whether the direction is northbound/east-west 4674 * or southbound */ 4675 enum TieDirectionType { 4676 Illegal = 0, 4677 South = 1, 4678 North = 2, 4679 DirectionMaxValue = 3, 4680 } 4682 enum AddressFamilyType { 4683 Illegal = 0, 4684 AddressFamilyMinValue = 1, 4685 IPv4 = 2, 4686 IPv6 = 3, 4687 AddressFamilyMaxValue = 4, 4688 } 4690 struct IPv4PrefixType { 4691 1: required IPv4Address address; 4692 2: required PrefixLenType prefixlen; 4693 } 4695 struct IPv6PrefixType { 4696 1: required IPv6Address address; 4697 2: required PrefixLenType prefixlen; 4698 } 4700 union IPAddressType { 4701 1: optional IPv4Address ipv4address; 4702 2: optional IPv6Address ipv6address; 4703 } 4705 /** Prefix representing reachablity. Observe that for interface 4706 addresses the protocol can propagate the address part beyond 4707 the subnet mask and on reachability computation that has to 4708 be normalized. The non-significant bits can be used for operational 4709 purposes. 4710 */ 4711 union IPPrefixType { 4712 1: optional IPv4PrefixType ipv4prefix; 4713 2: optional IPv6PrefixType ipv6prefix; 4714 } 4716 /** @note: Sequence of a prefix. Comparison function: 4717 if diff(timestamps) < 200msecs better transactionid wins 4718 else better time wins 4719 */ 4720 struct PrefixSequenceType { 4721 1: required IEEE802_1ASTimeStampType timestamp; 4722 2: optional PrefixTransactionIDType transactionid; 4723 } 4725 /** Type of TIE. 4727 This enum indicates what TIE type the TIE is carrying. 4728 In case the value is not known to the receiver, 4729 re-flooded the same way as prefix TIEs. This allows for 4730 future extensions of the protocol within the same schema major 4731 with types opaque to some nodes unless the flooding scope is not 4732 the same as prefix TIE, then a major version revision MUST 4733 be performed. 4734 */ 4735 enum TIETypeType { 4736 Illegal = 0, 4737 TIETypeMinValue = 1, 4738 /** first legal value */ 4739 NodeTIEType = 2, 4740 PrefixTIEType = 3, 4741 PositiveDisaggregationPrefixTIEType = 4, 4742 NegativeDisaggregationPrefixTIEType = 5, 4743 PGPrefixTIEType = 6, 4744 KeyValueTIEType = 7, 4745 ExternalPrefixTIEType = 8, 4746 TIETypeMaxValue = 9, 4747 } 4749 /** @note: route types which MUST be ordered on their preference 4750 PGP prefixes are most preferred attracting 4751 traffic north (towards spine) and then south 4752 normal prefixes are attracting traffic south (towards leafs), 4753 i.e. prefix in NORTH PREFIX TIE is preferred over SOUTH PREFIX TIE. 4755 @note: The only purpose of those values is to introduce an 4756 ordering whereas an implementation can choose internally 4757 any other values as long the ordering is preserved 4758 */ 4760 enum RouteType { 4761 Illegal = 0, 4762 RouteTypeMinValue = 1, 4763 /** First legal value. */ 4764 /** Discard routes are most prefered */ 4765 Discard = 2, 4767 /** Local prefixes are directly attached prefixes on the 4768 * system such as e.g. interface routes. 4769 */ 4770 LocalPrefix = 3, 4771 /** advertised in S-TIEs */ 4772 SouthPGPPrefix = 4, 4773 /** advertised in N-TIEs */ 4774 NorthPGPPrefix = 5, 4775 /** advertised in N-TIEs */ 4776 NorthPrefix = 6, 4777 /** externally imported north */ 4778 NorthExternalPrefix = 7, 4779 /** advertised in S-TIEs, either normal prefix or positive disaggregation */ 4780 SouthPrefix = 8, 4781 /** externally imported south */ 4782 SouthExternalPrefix = 9, 4783 /** negative, transitive prefixes are least preferred of 4784 local variety */ 4785 NegativeSouthPrefix = 10, 4786 RouteTypeMaxValue = 11, 4787 } 4789 B.2. encoding.thrift 4791 /** 4792 Thrift file for packet encodings for RIFT 4793 */ 4795 include "common.thrift" 4797 /** 4798 Thrift file for packet encodings for RIFT 4800 Copyright (c) Juniper Networks, Inc., 2016- 4801 All rights reserved. 4802 */ 4804 include "common.thrift" 4805 namespace rs models 4806 namespace py encoding 4808 /** represents protocol encoding schema major version */ 4809 const common.VersionType protocol_major_version = 2 4810 /** represents protocol encoding schema minor version */ 4811 const common.MinorVersionType protocol_minor_version = 0 4813 /** common RIFT packet header */ 4814 struct PacketHeader { 4815 1: required common.VersionType major_version = protocol_major_version; 4816 2: required common.VersionType minor_version = protocol_minor_version; 4817 /** this is the node sending the packet, in case of LIE/TIRE/TIDE 4818 also the originator of it */ 4819 3: required common.SystemIDType sender; 4820 /** level of the node sending the packet, required on everything except 4821 * LIEs. Lack of presence on LIEs indicates UNDEFINED_LEVEL and is used 4822 * in ZTP procedures. 4823 */ 4824 4: optional common.LevelType level; 4825 } 4827 /** Community serves as community for PGP purposes */ 4828 struct Community { 4829 1: required i32 top; 4830 2: required i32 bottom; 4831 } 4833 /** Neighbor structure */ 4834 struct Neighbor { 4835 1: required common.SystemIDType originator; 4836 2: required common.LinkIDType remote_id; 4837 } 4839 /** Capabilities the node supports. The schema may add to this 4840 field future capabilities to indicate whether it will support 4841 interpretation of future schema extensions on the same major 4842 revision. Such fields MUST be optional and have an implicit or 4843 explicit false default value. If a future capability changes route 4844 selection or generates blackholes if some nodes are not supporting 4845 it then a major version increment is unavoidable. 4846 */ 4847 struct NodeCapabilities { 4848 /** can this node participate in flood reduction */ 4849 1: optional bool flood_reduction = 4850 common.flood_reduction_default; 4851 /** does this node restrict itself to be top-of-fabric or 4852 leaf only (in ZTP) and does it support leaf-2-leaf procedures */ 4854 2: optional common.HierarchyIndications hierarchy_indications; 4855 } 4857 /* Link capabilities */ 4858 struct LinkCapabilities { 4859 /* indicates that the link's `local ID` can be used as its BFD 4860 discriminator and the link is supporting BFD */ 4861 1: optional bool bfd = 4862 common.bfd_default; 4863 } 4865 /** RIFT LIE packet 4867 @note this node's level is already included on the packet header */ 4868 struct LIEPacket { 4869 /** optional node or adjacency name */ 4870 1: optional string name; 4871 /** local link ID */ 4872 2: required common.LinkIDType local_id; 4873 /** UDP port to which we can receive flooded TIEs */ 4874 3: required common.UDPPortType flood_port = 4875 common.default_tie_udp_flood_port; 4876 /** layer 3 MTU, used to discover to mismatch. */ 4877 4: optional common.MTUSizeType link_mtu_size = 4878 common.default_mtu_size; 4879 /** local link bandwidth on the interface */ 4880 5: optional common.BandwithInMegaBitsType link_bandwidth = 4881 common.default_bandwidth; 4882 /** this will reflect the neighbor once received to provide 4883 3-way connectivity */ 4884 6: optional Neighbor neighbor; 4885 7: optional common.PodType pod = 4886 common.default_pod; 4887 /** optional node capabilities shown in the LIE. The capabilies 4888 MUST match the capabilities shown in the Node TIEs, otherwise 4889 the behavior is unspecified. A node detecting the mismatch 4890 SHOULD generate according error */ 4891 10: optional NodeCapabilities node_capabilities; 4892 11: optional LinkCapabilities link_capabilities; 4893 /** required holdtime of the adjacency, i.e. how much time 4894 MUST expire without LIE for the adjacency to drop */ 4895 12: required common.TimeIntervalInSecType holdtime = 4896 common.default_lie_holdtime; 4897 /** optional downstream assigned locally significant label 4898 value for the adjacency */ 4899 13: optional common.LabelType label; 4900 /** indicates that the level on the LIE MUST NOT be used 4901 to derive a ZTP level by the receiving node */ 4903 21: optional bool not_a_ztp_offer = 4904 common.default_not_a_ztp_offer; 4905 /** indicates to northbound neighbor that it should 4906 be reflooding this node's N-TIEs to achieve flood reduction and 4907 balancing for northbound flooding. To be ignored if received from a 4908 northbound adjacency */ 4909 22: optional bool you_are_flood_repeater = 4910 common.default_you_are_flood_repeater; 4911 /** can be optionally set to indicate to neighbor that packet losses are seen on 4912 reception based on packet numbers or the rate is too high. The receiver SHOULD 4913 temporarily slow down flooding rates. 4914 */ 4915 23: optional bool you_are_sending_too_quickly = 4916 false; 4918 } 4920 /** LinkID pair describes one of parallel links between two nodes */ 4921 struct LinkIDPair { 4922 /** node-wide unique value for the local link */ 4923 1: required common.LinkIDType local_id; 4924 /** received remote link ID for this link */ 4925 2: required common.LinkIDType remote_id; 4927 /** optionally describes the local interface index of the link */ 4928 10: optional common.PlatformInterfaceIndex platform_interface_index; 4929 /** optionally describes the local interface name */ 4930 11: optional string platform_interface_name; 4931 /** optional indication whether the link is secured, i.e. protected by outer key, absence 4932 of this element means no indication, undefined outer key means not secured */ 4933 12: optional common.OuterSecurityKeyID trusted_outer_security_key; 4934 /** more properties of the link can go in here */ 4935 } 4937 /** ID of a TIE 4939 @note: TIEID space is a total order achieved by comparing the elements 4940 in sequence defined and comparing each value as an 4941 unsigned integer of according length. 4942 */ 4943 struct TIEID { 4944 /** indicates direction of the TIE */ 4945 1: required common.TieDirectionType direction; 4946 /** indicates originator of the TIE */ 4947 2: required common.SystemIDType originator; 4948 3: required common.TIETypeType tietype; 4949 4: required common.TIENrType tie_nr; 4950 } 4951 /** Header of a TIE. 4953 @note: TIEID space is a total order achieved by comparing the elements 4954 in sequence defined and comparing each value as an 4955 unsigned integer of according length. 4957 After sequence number the lifetime received on the envelope 4958 must be used for comparison before further fields. 4960 `origination_time` and `origination_lifetime` are disregarded 4961 for comparison purposes and carried purely for debugging/security 4962 purposes if present. 4963 */ 4964 struct TIEHeader { 4965 2: required TIEID tieid; 4966 3: required common.SeqNrType seq_nr; 4968 /** optional absolute timestamp when the TIE 4969 was generated. This can be used on fabrics with 4970 synchronized clock to prevent lifetime modification attacks. */ 4971 10: optional common.IEEE802_1ASTimeStampType origination_time; 4972 /** optional original lifetime when the TIE 4973 was generated. This can be used on fabrics with 4974 synchronized clock to prevent lifetime modification attacks. */ 4975 12: optional common.LifeTimeInSecType origination_lifetime; 4976 } 4978 /** Header of a TIE as described in TIRE/TIDE. 4979 */ 4980 struct TIEHeaderWithLifeTime { 4981 1: required TIEHeader header; 4982 /** remaining lifetime that expires down to 0 just like in ISIS. 4983 TIEs with lifetimes differing by less than `lifetime_diff2ignore` MUST 4984 be considered EQUAL. */ 4985 2: required common.LifeTimeInSecType remaining_lifetime; 4986 } 4988 /** A TIDE with sorted TIE headers, if headers unsorted, behavior is undefined */ 4989 struct TIDEPacket { 4990 /** all 00s marks starts */ 4991 1: required TIEID start_range; 4992 /** all FFs mark end */ 4993 2: required TIEID end_range; 4994 /** _sorted_ list of headers */ 4995 3: required list headers; 4996 } 4998 /** A TIRE packet */ 4999 struct TIREPacket { 5000 1: required set headers; 5001 } 5003 /** Neighbor of a node */ 5004 struct NodeNeighborsTIEElement { 5005 /** Level of neighbor */ 5006 1: required common.LevelType level; 5007 /** Cost to neighbor. 5009 @note: All parallel links to same node 5010 incur same cost, in case the neighbor has multiple 5011 parallel links at different cost, the largest distance 5012 (highest numerical value) MUST be advertised 5013 @note: any neighbor with cost <= 0 MUST be ignored in computations */ 5014 3: optional common.MetricType cost = common.default_distance; 5015 /** can carry description of multiple parallel links in a TIE */ 5016 4: optional set link_ids; 5018 /** total bandwith to neighbor, this will be normally sum of the 5019 bandwidths of all the parallel links. */ 5020 5: optional common.BandwithInMegaBitsType bandwidth = 5021 common.default_bandwidth; 5022 } 5024 /** Flags the node sets */ 5025 struct NodeFlags { 5026 /** node is in overload, do not transit traffic through it */ 5027 1: optional bool overload = common.overload_default; 5028 } 5030 /** Description of a node. 5032 It may occur multiple times in different TIEs but if either 5033 * capabilities values do not match or 5034 * flags values do not match or 5035 * neighbors repeat with different values 5037 the behavior is undefined and a warning SHOULD be generated. 5038 Neighbors can be distributed across multiple TIEs however if 5039 the sets are disjoint. Miscablings SHOULD be repeated in every 5040 node TIE, otherwise the behavior is undefined. 5042 @note: observe that absence of fields implies defined defaults 5043 */ 5044 struct NodeTIEElement { 5045 1: required common.LevelType level; 5046 /** If neighbor systemID repeats in other node TIEs of same node 5047 the behavior is undefined. */ 5048 2: required map neighbors; 5050 3: optional NodeCapabilities capabilities; 5051 4: optional NodeFlags flags; 5052 /** optional node name for easier operations */ 5053 5: optional string name; 5054 /** PoD to which the node belongs */ 5055 6: optional common.PodType pod; 5057 /** if any local links are miscabled, the indication is flooded. */ 5058 10: optional set miscabled_links; 5060 } 5062 struct PrefixAttributes { 5063 2: required common.MetricType metric = common.default_distance; 5064 /** generic unordered set of route tags, can be redistributed to other protocols or use 5065 within the context of real time analytics */ 5066 3: optional set tags; 5067 /** optional monotonic clock for mobile addresses */ 5068 4: optional common.PrefixSequenceType monotonic_clock; 5069 /** optionally indicates the interface is a node loopback */ 5070 6: optional bool loopback = false; 5071 /** indicates that the prefix is directly attached, i.e. should be routed to even if 5072 the node is in overload. **/ 5073 7: optional bool directly_attached = true; 5075 /** in case of locally originated prefixes, i.e. interface addresses this can 5076 describe which link the address belongs to. */ 5077 10: optional common.LinkIDType from_link; 5078 } 5080 /** multiple prefixes */ 5081 struct PrefixTIEElement { 5082 /** prefixes with the associated attributes. 5083 if the same prefix repeats in multiple TIEs of same node 5084 behavior is unspecified */ 5085 1: required map prefixes; 5086 } 5088 /** keys with their values */ 5089 struct KeyValueTIEElement { 5090 /** if the same key repeats in multiple TIEs of same node 5091 or with different values, behavior is unspecified */ 5092 1: required map keyvalues; 5093 } 5094 /** single element in a TIE. enum `common.TIETypeType` 5095 in TIEID indicates which elements MUST be present 5096 in the TIEElement. In case of mismatch the unexpected 5097 elements MUST be ignored. In case of lack of expected 5098 element the TIE an error MUST be reported and the TIE 5099 MUST be ignored. 5101 This type can be extended with new optional elements 5102 for new `common.TIETypeType` values without breaking 5103 the major but if it is necessary to understand whether 5104 all nodes support the new type a node capability must 5105 be added as well. 5106 */ 5107 union TIEElement { 5108 /** in case of enum common.TIETypeType.NodeTIEType */ 5109 1: optional NodeTIEElement node; 5110 /** in case of enum common.TIETypeType.PrefixTIEType */ 5111 2: optional PrefixTIEElement prefixes; 5112 /** positive prefixes (always southbound) 5113 It MUST NOT be advertised within a North TIE. 5114 */ 5115 3: optional PrefixTIEElement positive_disaggregation_prefixes; 5116 /** transitive, negative prefixes (always southbound) which 5117 MUST be aggregated and propagated 5118 according to the specification 5119 southwards towards lower levels to heal 5120 pathological upper level partitioning, otherwise 5121 blackholes may occur in multiplane fabrics. 5122 It MUST NOT be advertised within a North TIE. 5123 */ 5124 4: optional PrefixTIEElement negative_disaggregation_prefixes; 5125 /** externally reimported prefixes */ 5126 5: optional PrefixTIEElement external_prefixes; 5127 /** Key-Value store elements */ 5128 6: optional KeyValueTIEElement keyvalues; 5129 /** @todo: policy guided prefixes */ 5130 } 5132 struct TIEPacket { 5133 1: required TIEHeader header; 5134 2: required TIEElement element; 5135 } 5137 union PacketContent { 5138 1: optional LIEPacket lie; 5139 2: optional TIDEPacket tide; 5140 3: optional TIREPacket tire; 5141 4: optional TIEPacket tie; 5143 } 5145 /** protocol packet structure */ 5146 struct ProtocolPacket { 5147 1: required PacketHeader header; 5148 2: required PacketContent content; 5149 } 5151 Appendix C. Finite State Machines and Precise Operational 5152 Specifications 5154 Some FSM figures are provided as [DOT] description due to limitations 5155 of ASCII art. 5157 On Entry action is performed every time and right before the 5158 according state is entered, i.e. after any transitions from previous 5159 state. 5161 On Exit action is performed every time and immediately when a state 5162 is exited, i.e. before any transitions towards target state are 5163 performed. 5165 Any attempt to transition from a state towards another on reception 5166 of an event where no action is specified MUST be considered an 5167 unrecoverable error. 5169 The FSMs and procedures are NOT normative in the sense that an 5170 implementation MUST implement them literally (which would be 5171 overspecification) but an implementation MUST exhibit externally 5172 observable behavior that is identical to the execution of the 5173 specified FSMs. 5175 Where a FSM representation is inconvenient, i.e. the amount of 5176 procedures and kept state exceeds the amount of transitions, we defer 5177 to a more procedural description on data structures. 5179 C.1. LIE FSM 5181 Initial state is `OneWay`. 5183 Event `MultipleNeighbors` occurs normally when more than two nodes 5184 see each other on the same link or a remote node is quickly 5185 reconfigured or rebooted without regressing to `OneWay` first. Each 5186 occurence of the event SHOULD generate a clear, according 5187 notification to help operational deployments. 5189 The machine sends LIEs on several transitions to accelerate adjacency 5190 bring-up without waiting for the timer tic. 5192 digraph Ga556dde74c30450aae125eaebc33bd57 { 5193 Nd16ab5092c6b421c88da482eb4ae36b6[label="ThreeWay"][shape="oval"]; 5194 N54edd2b9de7641688608f44fca346303[label="OneWay"][shape="oval"]; 5195 Nfeef2e6859ae4567bd7613a32cc28c0e[label="TwoWay"][shape="oval"]; 5196 N7f2bb2e04270458cb5c9bb56c4b96e23[label="Enter"][style="invis"][shape="plain"]; 5197 N292744a4097f492f8605c926b924616b[label="Enter"][style="dashed"][shape="plain"]; 5198 Nc48847ba98e348efb45f5b78f4a5c987[label="Exit"][style="invis"][shape="plain"]; 5199 Nd16ab5092c6b421c88da482eb4ae36b6 -> N54edd2b9de7641688608f44fca346303 5200 [label="|NeighborChangedLevel|\n|NeighborChangedAddress|\n|UnacceptableHeader|\n|MTUMismatch|\n|PODMismatch|\n|HoldtimeExpired|\n|MultipleNeighbors|"] 5201 [color="black"][arrowhead="normal" dir="both" arrowtail="none"]; 5202 Nd16ab5092c6b421c88da482eb4ae36b6 -> Nd16ab5092c6b421c88da482eb4ae36b6 5203 [label="|TimerTick|\n|LieRcvd|\n|SendLie|"][color="black"] 5204 [arrowhead="normal" dir="both" arrowtail="none"]; 5205 Nfeef2e6859ae4567bd7613a32cc28c0e -> Nfeef2e6859ae4567bd7613a32cc28c0e 5206 [label="|TimerTick|\n|LieRcvd|\n|SendLie|"][color="black"] 5207 [arrowhead="normal" dir="both" arrowtail="none"]; 5208 N54edd2b9de7641688608f44fca346303 -> Nd16ab5092c6b421c88da482eb4ae36b6 5209 [label="|ValidReflection|"][color="red"][arrowhead="normal" dir="both" arrowtail="none"]; 5210 Nd16ab5092c6b421c88da482eb4ae36b6 -> Nd16ab5092c6b421c88da482eb4ae36b6 5211 [label="|HALChanged|\n|HATChanged|\n|HALSChanged|\n|UpdateZTPOffer|"][color="blue"] 5212 [arrowhead="normal" dir="both" arrowtail="none"]; 5213 Nd16ab5092c6b421c88da482eb4ae36b6 -> Nd16ab5092c6b421c88da482eb4ae36b6 5214 [label="|ValidReflection|"][color="red"][arrowhead="normal" dir="both" arrowtail="none"]; 5215 Nfeef2e6859ae4567bd7613a32cc28c0e -> N54edd2b9de7641688608f44fca346303 5216 [label="|LevelChanged|"][color="blue"][arrowhead="normal" dir="both" arrowtail="none"]; 5217 Nfeef2e6859ae4567bd7613a32cc28c0e -> N54edd2b9de7641688608f44fca346303 5218 [label="|NeighborChangedLevel|\n|NeighborChangedAddress|\n|UnacceptableHeader|\n|MTUMismatch|\n|PODMismatch|\n|HoldtimeExpired|\n|MultipleNeighbors|"] 5219 [color="black"][arrowhead="normal" dir="both" arrowtail="none"]; 5220 Nfeef2e6859ae4567bd7613a32cc28c0e -> Nd16ab5092c6b421c88da482eb4ae36b6 5221 [label="|ValidReflection|"][color="red"][arrowhead="normal" dir="both" arrowtail="none"]; 5222 N54edd2b9de7641688608f44fca346303 -> N54edd2b9de7641688608f44fca346303 5223 [label="|TimerTick|\n|LieRcvd|\n|NeighborChangedLevel|\n|NeighborChangedAddress|\n|UnacceptableHeader|\n|MTUMismatch|\n|PODMismatch|\n|HoldtimeExpired|\n|SendLie|"] 5224 [color="black"][arrowhead="normal" dir="both" arrowtail="none"]; 5225 N292744a4097f492f8605c926b924616b -> N54edd2b9de7641688608f44fca346303 5226 [label=""][color="black"][arrowhead="normal" dir="both" arrowtail="none"]; 5227 Nd16ab5092c6b421c88da482eb4ae36b6 -> N54edd2b9de7641688608f44fca346303 5228 [label="|LevelChanged|"][color="blue"][arrowhead="normal" dir="both" arrowtail="none"]; 5229 N54edd2b9de7641688608f44fca346303 -> Nfeef2e6859ae4567bd7613a32cc28c0e 5230 [label="|NewNeighbor|"][color="black"][arrowhead="normal" dir="both" arrowtail="none"]; 5231 N54edd2b9de7641688608f44fca346303 -> N54edd2b9de7641688608f44fca346303 5232 [label="|LevelChanged|\n|HALChanged|\n|HATChanged|\n|HALSChanged|\n|UpdateZTPOffer|"] 5233 [color="blue"][arrowhead="normal" dir="both" arrowtail="none"]; 5234 Nfeef2e6859ae4567bd7613a32cc28c0e -> Nfeef2e6859ae4567bd7613a32cc28c0e 5235 [label="|HALChanged|\n|HATChanged|\n|HALSChanged|\n|UpdateZTPOffer|"] 5236 [color="blue"][arrowhead="normal" dir="both" arrowtail="none"]; 5237 Nd16ab5092c6b421c88da482eb4ae36b6 -> Nfeef2e6859ae4567bd7613a32cc28c0e 5238 [label="|NeighborDroppedReflection|"] 5239 [color="red"][arrowhead="normal" dir="both" arrowtail="none"]; 5240 N54edd2b9de7641688608f44fca346303 -> N54edd2b9de7641688608f44fca346303 5241 [label="|NeighborDroppedReflection|"][color="red"] 5242 [arrowhead="normal" dir="both" arrowtail="none"]; 5243 } 5245 LIE FSM DOT 5247 .. To be updated .. 5249 LIE FSM Figure 5251 Events 5253 o TimerTick: one second timer tic 5255 o LevelChanged: node's level has been changed by ZTP or 5256 configuration 5258 o HALChanged: best HAL computed by ZTP has changed 5260 o HATChanged: HAT computed by ZTP has changed 5262 o HALSChanged: set of HAL offering systems computed by ZTP has 5263 changed 5265 o LieRcvd: received LIE 5267 o NewNeighbor: new neighbor parsed 5269 o ValidReflection: received own reflection from neighbor 5271 o NeighborDroppedReflection: lost previous own reflection from 5272 neighbor 5274 o NeighborChangedLevel: neighbor changed advertised level 5276 o NeighborChangedAddress: neighbor changed IP address 5278 o UnacceptableHeader: unacceptable header seen 5280 o MTUMismatch: MTU mismatched 5282 o PODMismatch: Unacceptable PoD seen 5283 o HoldtimeExpired: adjacency hold down expired 5285 o MultipleNeighbors: more than one neighbor seen on interface 5287 o SendLie: send a LIE out 5289 o UpdateZTPOffer: update this node's ZTP offer 5291 Actions 5293 on TimerTick in TwoWay finishes in TwoWay: PUSH SendLie event, if 5294 holdtime expired PUSH HoldtimeExpired event 5296 on HALChanged in TwoWay finishes in TwoWay: store new HAL 5298 on MTUMismatch in ThreeWay finishes in OneWay: no action 5300 on HALChanged in ThreeWay finishes in ThreeWay: store new HAL 5302 on ValidReflection in TwoWay finishes in ThreeWay: no action 5304 on ValidReflection in OneWay finishes in ThreeWay: no action 5306 on NeighborDroppedReflection in ThreeWay finishes in TwoWay: no 5307 action 5309 on LieRcvd in ThreeWay finishes in ThreeWay: PROCESS_LIE 5311 on MultipleNeighbors in TwoWay finishes in OneWay: no action 5313 on UnacceptableHeader in ThreeWay finishes in OneWay: no action 5315 on MTUMismatch in TwoWay finishes in OneWay: no action 5317 on LevelChanged in OneWay finishes in OneWay: update level with 5318 event value, PUSH SendLie event 5320 on UnacceptableHeader in TwoWay finishes in OneWay: no action 5322 on HALSChanged in TwoWay finishes in TwoWay: store HALS 5324 on UpdateZTPOffer in TwoWay finishes in TwoWay: send offer to ZTP 5325 FSM 5327 on NeighborChangedLevel in TwoWay finishes in OneWay: no action 5329 on NewNeighbor in OneWay finishes in TwoWay: PUSH SendLie event 5330 on NeighborChangedAddress in ThreeWay finishes in OneWay: no 5331 action 5333 on HALChanged in OneWay finishes in OneWay: store new HAL 5335 on NeighborChangedLevel in OneWay finishes in OneWay: no action 5337 on HoldtimeExpired in TwoWay finishes in OneWay: no action 5339 on SendLie in TwoWay finishes in TwoWay: SEND_LIE 5341 on LevelChanged in TwoWay finishes in OneWay: update level with 5342 event value 5344 on NeighborChangedAddress in OneWay finishes in OneWay: no action 5346 on HATChanged in TwoWay finishes in TwoWay: store HAT 5348 on LieRcvd in TwoWay finishes in TwoWay: PROCESS_LIE 5350 on MultipleNeighbors in ThreeWay finishes in OneWay: no action 5352 on MTUMismatch in OneWay finishes in OneWay: no action 5354 on SendLie in OneWay finishes in OneWay: SEND_LIE 5356 on LieRcvd in OneWay finishes in OneWay: PROCESS_LIE 5358 on TimerTick in ThreeWay finishes in ThreeWay: PUSH SendLie event, 5359 if holdtime expired PUSH HoldtimeExpired event 5361 on TimerTick in OneWay finishes in OneWay: PUSH SendLie event 5363 on PODMismatch in ThreeWay finishes in OneWay: no action 5365 on LevelChanged in ThreeWay finishes in OneWay: update level with 5366 event value 5368 on NeighborChangedLevel in ThreeWay finishes in OneWay: no action 5370 on UpdateZTPOffer in OneWay finishes in OneWay: send offer to ZTP 5371 FSM 5373 on UpdateZTPOffer in ThreeWay finishes in ThreeWay: send offer to 5374 ZTP FSM 5376 on HATChanged in OneWay finishes in OneWay: store HAT 5377 on HATChanged in ThreeWay finishes in ThreeWay: store HAT 5379 on HoldtimeExpired in OneWay finishes in OneWay: no action 5381 on UnacceptableHeader in OneWay finishes in OneWay: no action 5383 on PODMismatch in OneWay finishes in OneWay: no action 5385 on SendLie in ThreeWay finishes in ThreeWay: SEND_LIE 5387 on NeighborChangedAddress in TwoWay finishes in OneWay: no action 5389 on ValidReflection in ThreeWay finishes in ThreeWay: no action 5391 on HALSChanged in OneWay finishes in OneWay: store HALS 5393 on HoldtimeExpired in ThreeWay finishes in OneWay: no action 5395 on HALSChanged in ThreeWay finishes in ThreeWay: store HALS 5397 on NeighborDroppedReflection in OneWay finishes in OneWay: no 5398 action 5400 on PODMismatch in TwoWay finishes in OneWay: no action 5402 on Entry into OneWay: CLEANUP 5404 Following words are used for well known procedures: 5406 1. PUSH Event: pushes an event to be executed by the FSM upon exit 5407 of this action 5409 2. CLEANUP: neighbor MUST be reset to unknown 5411 3. SEND_LIE: create a new LIE packet 5413 1. reflecting the neighbor if known and valid and 5415 2. setting the necessary `not_a_ztp_offer` variable if level was 5416 derived from last known neighbor on this interface and 5418 3. setting `you_are_not_flood_repeater` to computed value 5420 4. PROCESS_LIE: 5422 1. if lie has wrong major version OR our own system ID or 5423 invalid system ID then CLEANUP else 5425 2. if lie has non matching MTUs then CLEANUP, PUSH 5426 UpdateZTPOffer, PUSH MTUMismatch else 5428 3. if PoD rules do not allow adjacency forming then CLEANUP, 5429 PUSH PODMismatch, PUSH MTUMismatch else 5431 4. if lie has undefined level OR my level is undefined OR this 5432 node is leaf and remote level lower than HAT OR (lie's level 5433 is not leaf AND its difference is more than one from my 5434 level) then CLEANUP, PUSH UpdateZTPOffer, PUSH 5435 UnacceptableHeader else 5437 5. PUSH UpdateZTPOffer, construct temporary new neighbor 5438 structure with values from lie, if no current neighbor exists 5439 then set neighbor to new neighbor, PUSH NewNeighbor event, 5440 CHECK_THREE_WAY else 5442 1. if current neighbor system ID differs from lie's system 5443 ID then PUSH MultipleNeighbors else 5445 2. if current neighbor stored level differs from lie's level 5446 then PUSH NeighborChangedLevel else 5448 3. if current neighbor stored IPv4/v6 address differs from 5449 lie's address then PUSH NeighborChangedAddress else 5451 4. if any of neighbor's flood address port, name, local 5452 linkid changed then PUSH NeighborChangedMinorFields and 5454 5. CHECK_THREE_WAY 5456 5. CHECK_THREE_WAY: if current state is one-way do nothing else 5458 1. if lie packet does not contain neighbor then if current state 5459 is three-way then PUSH NeighborDroppedReflection else 5461 2. if packet reflects this system's ID and local port and state 5462 is three-way then PUSH event ValidReflection else PUSH event 5463 MultipleNeighbors 5465 C.2. ZTP FSM 5467 Initial state is ComputeBestOffer. 5469 digraph Gd436cc3ced8c471eb30bd4f3ac946261 { 5470 N06108ba9ac894d988b3e4e8ea5ace007 5471 [label="Enter"] 5472 [style="invis"] 5473 [shape="plain"]; 5474 Na47ff5eac9aa4b2eaf12839af68aab1f 5475 [label="MultipleNeighborsWait"] 5476 [shape="oval"]; 5477 N57a829be68e2489d8dc6b84e10597d0b 5478 [label="OneWay"] 5479 [shape="oval"]; 5480 Na641d400819a468d987e31182cdb013e 5481 [label="ThreeWay"] 5482 [shape="oval"]; 5483 Necfbfc2d8e5b482682ee66e604450c7b 5484 [label="Enter"] 5485 [style="dashed"] 5486 [shape="plain"]; 5487 N16db54bf2c5d48f093ad6c18e70081ee 5488 [label="TwoWay"] 5489 [shape="oval"]; 5490 N1b89016876b44cc1b9c1e4a735769560 5491 [label="Exit"] 5492 [style="invis"] 5493 [shape="plain"]; 5494 N16db54bf2c5d48f093ad6c18e70081ee -> N57a829be68e2489d8dc6b84e10597d0b 5495 [label="|NeighborChangedLevel|\n|NeighborChangedAddress|\n|UnacceptableHeader|\n|MTUMismatch|\n|PODMismatch|\n|HoldtimeExpired|"] 5496 [color="black"] 5497 [arrowhead="normal" dir="both" arrowtail="none"]; 5498 N57a829be68e2489d8dc6b84e10597d0b -> N57a829be68e2489d8dc6b84e10597d0b 5499 [label="|NeighborDroppedReflection|"] 5500 [color="red"] 5501 [arrowhead="normal" dir="both" arrowtail="none"]; 5502 N57a829be68e2489d8dc6b84e10597d0b -> Na47ff5eac9aa4b2eaf12839af68aab1f 5503 [label="|MultipleNeighbors|"] 5504 [color="black"] 5505 [arrowhead="normal" dir="both" arrowtail="none"]; 5506 Necfbfc2d8e5b482682ee66e604450c7b -> N57a829be68e2489d8dc6b84e10597d0b 5507 [label=""] 5508 [color="black"] 5509 [arrowhead="normal" dir="both" arrowtail="none"]; 5510 N57a829be68e2489d8dc6b84e10597d0b -> N16db54bf2c5d48f093ad6c18e70081ee 5511 [label="|NewNeighbor|"] 5512 [color="black"] 5513 [arrowhead="normal" dir="both" arrowtail="none"]; 5514 Na641d400819a468d987e31182cdb013e -> Na47ff5eac9aa4b2eaf12839af68aab1f 5515 [label="|MultipleNeighbors|"] 5516 [color="black"] 5517 [arrowhead="normal" dir="both" arrowtail="none"]; 5518 N16db54bf2c5d48f093ad6c18e70081ee -> N16db54bf2c5d48f093ad6c18e70081ee 5519 [label="|HALChanged|\n|HATChanged|\n|HALSChanged|\n|UpdateZTPOffer|"] 5520 [color="blue"] 5521 [arrowhead="normal" dir="both" arrowtail="none"]; 5522 Na641d400819a468d987e31182cdb013e -> N16db54bf2c5d48f093ad6c18e70081ee 5523 [label="|NeighborDroppedReflection|"] 5524 [color="red"] 5525 [arrowhead="normal" dir="both" arrowtail="none"]; 5526 Na47ff5eac9aa4b2eaf12839af68aab1f -> Na47ff5eac9aa4b2eaf12839af68aab1f 5527 [label="|TimerTick|\n|MultipleNeighbors|"] 5528 [color="black"] 5529 [arrowhead="normal" dir="both" arrowtail="none"]; 5530 N57a829be68e2489d8dc6b84e10597d0b -> N57a829be68e2489d8dc6b84e10597d0b 5531 [label="|LevelChanged|\n|HALChanged|\n|HATChanged|\n|HALSChanged|\n|UpdateZTPOffer|"] 5532 [color="blue"] 5533 [arrowhead="normal" dir="both" arrowtail="none"]; 5534 Na641d400819a468d987e31182cdb013e -> Na641d400819a468d987e31182cdb013e 5535 [label="|HALChanged|\n|HATChanged|\n|HALSChanged|\n|UpdateZTPOffer|"] 5536 [color="blue"] 5537 [arrowhead="normal" dir="both" arrowtail="none"]; 5538 Na641d400819a468d987e31182cdb013e -> N57a829be68e2489d8dc6b84e10597d0b 5539 [label="|NeighborChangedLevel|\n|NeighborChangedAddress|\n|UnacceptableHeader|\n|MTUMismatch|\n|PODMismatch|\n|HoldtimeExpired|"] 5540 [color="black"] 5541 [arrowhead="normal" dir="both" arrowtail="none"]; 5542 Na47ff5eac9aa4b2eaf12839af68aab1f -> Na47ff5eac9aa4b2eaf12839af68aab1f 5543 [label="|HALChanged|\n|HATChanged|\n|HALSChanged|\n|UpdateZTPOffer|"] 5544 [color="blue"] 5545 [arrowhead="normal" dir="both" arrowtail="none"]; 5546 N16db54bf2c5d48f093ad6c18e70081ee -> N57a829be68e2489d8dc6b84e10597d0b 5547 [label="|LevelChanged|"] 5548 [color="blue"] 5549 [arrowhead="normal" dir="both" arrowtail="none"]; 5550 Na641d400819a468d987e31182cdb013e -> N57a829be68e2489d8dc6b84e10597d0b 5551 [label="|LevelChanged|"] 5552 [color="blue"] 5553 [arrowhead="normal" dir="both" arrowtail="none"]; 5554 N16db54bf2c5d48f093ad6c18e70081ee -> Na47ff5eac9aa4b2eaf12839af68aab1f 5555 [label="|MultipleNeighbors|"] 5556 [color="black"] 5557 [arrowhead="normal" dir="both" arrowtail="none"]; 5558 Na47ff5eac9aa4b2eaf12839af68aab1f -> N57a829be68e2489d8dc6b84e10597d0b 5559 [label="|MultipleNeighborsDone|"] 5560 [color="black"] 5561 [arrowhead="normal" dir="both" arrowtail="none"]; 5562 N16db54bf2c5d48f093ad6c18e70081ee -> Na641d400819a468d987e31182cdb013e 5563 [label="|ValidReflection|"] 5564 [color="red"] 5565 [arrowhead="normal" dir="both" arrowtail="none"]; 5566 Na47ff5eac9aa4b2eaf12839af68aab1f -> N57a829be68e2489d8dc6b84e10597d0b 5567 [label="|LevelChanged|"] 5568 [color="blue"] 5569 [arrowhead="normal" dir="both" arrowtail="none"]; 5570 Na641d400819a468d987e31182cdb013e -> Na641d400819a468d987e31182cdb013e 5571 [label="|TimerTick|\n|LieRcvd|\n|SendLie|"] 5572 [color="black"] 5573 [arrowhead="normal" dir="both" arrowtail="none"]; 5574 N57a829be68e2489d8dc6b84e10597d0b -> N57a829be68e2489d8dc6b84e10597d0b 5575 [label="|TimerTick|\n|LieRcvd|\n|NeighborChangedLevel|\n|NeighborChangedAddress|\n|NeighborAddressAdded|\n|UnacceptableHeader|\n|MTUMismatch|\n|PODMismatch|\n|HoldtimeExpired|\n|SendLie|"] 5576 [color="black"] 5577 [arrowhead="normal" dir="both" arrowtail="none"]; 5578 N57a829be68e2489d8dc6b84e10597d0b -> Na641d400819a468d987e31182cdb013e 5579 [label="|ValidReflection|"] 5580 [color="red"] 5581 [arrowhead="normal" dir="both" arrowtail="none"]; 5582 N16db54bf2c5d48f093ad6c18e70081ee -> N16db54bf2c5d48f093ad6c18e70081ee 5583 [label="|TimerTick|\n|LieRcvd|\n|SendLie|"] 5584 [color="black"] 5585 [arrowhead="normal" dir="both" arrowtail="none"]; 5586 Na641d400819a468d987e31182cdb013e -> Na641d400819a468d987e31182cdb013e 5587 [label="|ValidReflection|"] 5588 [color="red"] 5589 [arrowhead="normal" dir="both" arrowtail="none"]; 5590 } 5592 ZTP FSM DOT 5594 Events 5596 o TimerTick: one second timer tic 5598 o LevelChanged: node's level has been changed by ZTP or 5599 configuration 5601 o HALChanged: best HAL computed by ZTP has changed 5603 o HATChanged: HAT computed by ZTP has changed 5605 o HALSChanged: set of HAL offering systems computed by ZTP has 5606 changed 5608 o LieRcvd: received LIE 5610 o NewNeighbor: new neighbor parsed 5612 o ValidReflection: received own reflection from neighbor 5613 o NeighborDroppedReflection: lost previous own reflection from 5614 neighbor 5616 o NeighborChangedLevel: neighbor changed advertised level 5618 o NeighborChangedAddress: neighbor changed IP address 5620 o UnacceptableHeader: unacceptable header seen 5622 o MTUMismatch: MTU mismatched 5624 o PODMismatch: Unacceptable PoD seen 5626 o HoldtimeExpired: adjacency hold down expired 5628 o MultipleNeighbors: more than one neighbor seen on interface 5630 o MultipleNeighborsDone: cooldown for multiple neighbors expired 5632 o SendLie: send a LIE out 5634 o UpdateZTPOffer: update this node's ZTP offer 5636 Actions 5638 on MTUMismatch in OneWay finishes in OneWay: no action 5640 on HoldtimeExpired in OneWay finishes in OneWay: no action 5642 on LevelChanged in ThreeWay finishes in OneWay: update level with 5643 event value 5645 on MultipleNeighbors in MultipleNeighborsWait finishes in 5646 MultipleNeighborsWait: start multiple neighbors timer as 4 * 5647 DEFAULT_LIE_HOLDTIME 5649 on HALChanged in MultipleNeighborsWait finishes in 5650 MultipleNeighborsWait: store new HAL 5652 on NeighborChangedAddress in ThreeWay finishes in OneWay: no 5653 action 5655 on ValidReflection in OneWay finishes in ThreeWay: no action 5657 on MTUMismatch in TwoWay finishes in OneWay: no action 5658 on TimerTick in MultipleNeighborsWait finishes in 5659 MultipleNeighborsWait: decrement MultipleNeighbors timer, if 5660 expired PUSH MultipleNeighborsDone 5662 on MultipleNeighborsDone in MultipleNeighborsWait finishes in 5663 OneWay: decrement MultipleNeighbors timer, if expired PUSH 5664 MultipleNeighborsDone 5666 on HATChanged in ThreeWay finishes in ThreeWay: store HAT 5668 on UpdateZTPOffer in TwoWay finishes in TwoWay: send offer to ZTP 5669 FSM 5671 on HALSChanged in TwoWay finishes in TwoWay: store HALS 5673 on PODMismatch in TwoWay finishes in OneWay: no action 5675 on LieRcvd in TwoWay finishes in TwoWay: PROCESS_LIE 5677 on PODMismatch in ThreeWay finishes in OneWay: no action 5679 on TimerTick in TwoWay finishes in TwoWay: PUSH SendLie event, if 5680 holdtime expired PUSH HoldtimeExpired event 5682 on SendLie in TwoWay finishes in TwoWay: SEND_LIE 5684 on SendLie in OneWay finishes in OneWay: SEND_LIE 5686 on TimerTick in OneWay finishes in OneWay: PUSH SendLie event 5688 on HALChanged in OneWay finishes in OneWay: store new HAL 5690 on HALSChanged in ThreeWay finishes in ThreeWay: store HALS 5692 on NeighborChangedLevel in TwoWay finishes in OneWay: no action 5694 on PODMismatch in OneWay finishes in OneWay: no action 5696 on HoldtimeExpired in TwoWay finishes in OneWay: no action 5698 on TimerTick in ThreeWay finishes in ThreeWay: PUSH SendLie event, 5699 if holdtime expired PUSH HoldtimeExpired event 5701 on MultipleNeighbors in TwoWay finishes in MultipleNeighborsWait: 5702 start multiple neighbors timer as 4 * DEFAULT_LIE_HOLDTIME 5704 on UpdateZTPOffer in MultipleNeighborsWait finishes in 5705 MultipleNeighborsWait: send offer to ZTP FSM 5706 on LieRcvd in OneWay finishes in OneWay: PROCESS_LIE 5708 on LevelChanged in MultipleNeighborsWait finishes in OneWay: 5709 update level with event value 5711 on UpdateZTPOffer in ThreeWay finishes in ThreeWay: send offer to 5712 ZTP FSM 5714 on HALChanged in TwoWay finishes in TwoWay: store new HAL 5716 on UnacceptableHeader in OneWay finishes in OneWay: no action 5718 on HALSChanged in OneWay finishes in OneWay: store HALS 5720 on HALSChanged in MultipleNeighborsWait finishes in 5721 MultipleNeighborsWait: store HALS 5723 on SendLie in ThreeWay finishes in ThreeWay: SEND_LIE 5725 on MTUMismatch in ThreeWay finishes in OneWay: no action 5727 on HATChanged in MultipleNeighborsWait finishes in 5728 MultipleNeighborsWait: store HAT 5730 on NeighborChangedAddress in OneWay finishes in OneWay: no action 5732 on ValidReflection in TwoWay finishes in ThreeWay: no action 5734 on MultipleNeighbors in OneWay finishes in MultipleNeighborsWait: 5735 start multiple neighbors timer as 4 * DEFAULT_LIE_HOLDTIME 5737 on NeighborChangedLevel in OneWay finishes in OneWay: no action 5739 on HATChanged in OneWay finishes in OneWay: store HAT 5741 on NeighborDroppedReflection in OneWay finishes in OneWay: no 5742 action 5744 on HALChanged in ThreeWay finishes in ThreeWay: store new HAL 5746 on NeighborAddressAdded in OneWay finishes in OneWay: no action 5748 on NeighborChangedAddress in TwoWay finishes in OneWay: no action 5750 on LieRcvd in ThreeWay finishes in ThreeWay: PROCESS_LIE 5752 on UnacceptableHeader in TwoWay finishes in OneWay: no action 5753 on LevelChanged in TwoWay finishes in OneWay: update level with 5754 event value 5756 on HATChanged in TwoWay finishes in TwoWay: store HAT 5758 on UpdateZTPOffer in OneWay finishes in OneWay: send offer to ZTP 5759 FSM 5761 on ValidReflection in ThreeWay finishes in ThreeWay: no action 5763 on UnacceptableHeader in ThreeWay finishes in OneWay: no action 5765 on HoldtimeExpired in ThreeWay finishes in OneWay: no action 5767 on NeighborChangedLevel in ThreeWay finishes in OneWay: no action 5769 on LevelChanged in OneWay finishes in OneWay: update level with 5770 event value, PUSH SendLie event 5772 on NewNeighbor in OneWay finishes in TwoWay: PUSH SendLie event 5774 on NeighborDroppedReflection in ThreeWay finishes in TwoWay: no 5775 action 5777 on MultipleNeighbors in ThreeWay finishes in 5778 MultipleNeighborsWait: start multiple neighbors timer as 4 * 5779 DEFAULT_LIE_HOLDTIME 5781 on Entry into OneWay: CLEANUP 5783 Following words are used for well known procedures: 5785 1. PUSH Event: pushes an event to be executed by the FSM upon exit 5786 of this action 5788 2. CLEANUP: neighbor MUST be reset to unknown 5790 3. SEND_LIE: create a new LIE packet 5792 1. reflecting the neighbor if known and valid and 5794 2. setting the necessary `not_a_ztp_offer` variable if level was 5795 derived from last known neighbor on this interface and 5797 3. setting `you_are_not_flood_repeater` to computed value 5799 4. PROCESS_LIE: 5801 1. if lie has wrong major version OR our own system ID or 5802 invalid system ID then CLEANUP else 5804 2. if lie has non matching MTUs then CLEANUP, PUSH 5805 UpdateZTPOffer, PUSH MTUMismatch else 5807 3. if PoD rules do not allow adjacency forming then CLEANUP, 5808 PUSH PODMismatch, PUSH MTUMismatch else 5810 4. if lie has undefined level OR my level is undefined OR this 5811 node is leaf and remote level lower than HAT OR (lie's level 5812 is not leaf AND its difference is more than one from my 5813 level) then CLEANUP, PUSH UpdateZTPOffer, PUSH 5814 UnacceptableHeader else 5816 5. PUSH UpdateZTPOffer, construct temporary new neighbor 5817 structure with values from lie, if no current neighbor exists 5818 then set neighbor to new neighbor, PUSH NewNeighbor event, 5819 CHECK_THREE_WAY else 5821 1. if current neighbor system ID differs from lie's system 5822 ID then PUSH MultipleNeighbors else 5824 2. if current neighbor stored level differs from lie's level 5825 then PUSH NeighborChangedLevel else 5827 3. if current neighbor stored IPv4/v6 address differs from 5828 lie's address then PUSH NeighborChangedAddress else 5830 4. if any of neighbor's flood address port, name, local 5831 linkid changed then PUSH NeighborChangedMinorFields and 5833 5. CHECK_THREE_WAY 5835 5. CHECK_THREE_WAY: if current state is one-way do nothing else 5837 1. if lie packet does not contain neighbor then if current state 5838 is three-way then PUSH NeighborDroppedReflection else 5840 2. if packet reflects this system's ID and local port and state 5841 is three-way then PUSH event ValidReflection else PUSH event 5842 MultipleNeighbors 5844 C.3. Flooding Procedures 5846 Flooding Procedures are described in terms of a flooding state of an 5847 adjacency and resulting operations on it driven by packet arrivals. 5849 The FSM has basically a single state and is not well suited to 5850 represent the behavior. 5852 RIFT does not specify any kind of flood rate limiting since such 5853 specifications always assume particular points in available 5854 technology speeds and feeds and those points are shifting at faster 5855 and faster rate (speed of light holding for the moment). The encoded 5856 packets provide hints to react accordingly to losses or overruns. 5858 Flooding of all according topology exchange elements SHOULD be 5859 performed at highest feasible rate whereas the rate of transmission 5860 MUST be throttled by reacting to adequate features of the system such 5861 as e.g. queue lengths or congestion indications in the protocol 5862 packets. 5864 C.3.1. FloodState Structure per Adjacency 5866 The structure contains conceptually the following elements. The word 5867 collection or queue indicates a set of elements that can be iterated: 5869 TIES_TX: Collection containing all the TIEs to transmit on the 5870 adjacency. 5872 TIES_ACK: Collection containing all the TIEs that have to be 5873 acknowledged on the adjacency. 5875 TIES_REQ: Collection containing all the TIE headers that have to be 5876 requested on the adjacency. 5878 TIES_RTX: Collection containing all TIEs that need retransmission 5879 with the according time to retransmit. 5881 Following words are used for well known procedures operating on this 5882 structure: 5884 TIE Describes either a full RIFT TIE or accordingly just the 5885 `TIEHeader` or `TIEID`. The according meaning is unambiguously 5886 contained in the context of the algorithm. 5888 is_flood_reduced(TIE): returns whether a TIE can be flood reduced or 5889 not. 5891 is_tide_entry_filtered(TIE): returns whether a header should be 5892 propagated in TIDE according to flooding scopes. 5894 is_request_filtered(TIE): returns whether a TIE request should be 5895 propagated to neighbor or not according to flooding scopes. 5897 is_flood_filtered(TIE): returns whether a TIE requested be flooded 5898 to neighbor or not according to flooding scopes. 5900 try_to_transmit_tie(TIE): 5902 A. if not is_flood_filtered(TIE) then 5904 1. remove TIE from TIES_RTX if present 5906 2. if TIE" with same key on TIES_ACK then 5908 a. if TIE" same or newer than TIE do nothing else 5910 b. remove TIE" from TIES_ACK and add TIE to TIES_TX 5912 3. else insert TIE into TIES_TX 5914 ack_tie(TIE): remove TIE from all collections and then insert TIE 5915 into TIES_ACK. 5917 tie_been_acked(TIE): remove TIE from all collections. 5919 remove_from_all_queues(TIE): same as `tie_been_acked`. 5921 request_tie(TIE): if not is_request_filtered(TIE) then 5922 remove_from_all_queues(TIE) and add to TIES_REQ. 5924 move_to_rtx_list(TIE): remove TIE from TIES_TX and then add to 5925 TIES_RTX using TIE retransmission interval. 5927 clear_requests(TIEs): remove all TIEs from TIES_REQ. 5929 bump_own_tie(TIE): for self-originated TIE originate an empty or re- 5930 generate with version number higher then the one in TIE. 5932 The collection SHOULD be served with following priorities if the 5933 system cannot process all the collections in real time: 5935 Elements on TIES_ACK should be processed with highest priority 5937 TIES_TX 5939 TIES_REQ and TIES_RTX 5941 C.3.2. TIDEs 5943 `TIEID` and `TIEHeader` space forms a strict total order (modulo 5944 uncomparable sequence numbers in the very unlikely event that can 5945 occur if a TIE is "stuck" in a part of a network while the originator 5946 reboots and reissues TIEs many times to the point its sequence# rolls 5947 over and forms incomparable distance to the "stuck" copy) which 5948 implies that a comparison relation is possible between two elements. 5949 With that it is implictly possible to compare TIEs, TIEHeaders and 5950 TIEIDs to each other whereas the shortest viable key is always 5951 implied. 5953 When generating and sending TIDEs an implementation SHOULD ensure 5954 that enough bandwidth is left to send elements of Floodstate 5955 structure. 5957 C.3.2.1. TIDE Generation 5959 As given by timer constant, periodically generate TIDEs by: 5961 NEXT_TIDE_ID: ID of next TIE to be sent in TIDE. 5963 TIDE_START: Begin of TIDE packet range. 5965 a. NEXT_TIDE_ID = MIN_TIEID 5967 b. while NEXT_TIDE_ID not equal to MAX_TIEID do 5969 1. TIDE_START = NEXT_TIDE_ID 5971 2. HEADERS = At most TIRDEs_PER_PKT headers in TIEDB starting at 5972 NEXT_TIDE_ID or higher that SHOULD be filtered by 5973 is_tide_entry_filtered and MUST either have a lifetime left > 5974 0 or have no content 5976 3. if HEADERS is empty then START = MIN_TIEID else START = first 5977 element in HEADERS 5979 4. if HEADERS' size less than TIRDEs_PER_PKT then END = 5980 MAX_TIEID else END = last element in HEADERS 5982 5. send sorted HEADERS as TIDE setting START and END as its 5983 range 5985 6. NEXT_TIDE_ID = END 5987 The constant `TIRDEs_PER_PKT` SHOULD be generated and used by the 5988 implementation to limit the amount of TIE headers per TIDE so the 5989 sent TIDE PDU does not exceed interface MTU. 5991 TIDE PDUs SHOULD be spaced on sending to prevent packet drops. 5993 C.3.2.2. TIDE Processing 5995 On reception of TIDEs the following processing is performed: 5997 TXKEYS: Collection of TIE Headers to be send after processing of 5998 the packet 6000 REQKEYS: Collection of TIEIDs to be requested after processing of 6001 the packet 6003 CLEARKEYS: Collection of TIEIDs to be removed from flood state 6004 queues 6006 LASTPROCESSED: Last processed TIEID in TIDE 6008 DBTIE: TIE in the LSDB if found 6010 a. LASTPROCESSED = TIDE.start_range 6012 b. for every HEADER in TIDE do 6014 1. DBTIE = find HEADER in current LSDB 6016 2. if HEADER < LASTPROCESSED then report error and reset 6017 adjacency and return 6019 3. put all TIEs in LSDB where (TIE.HEADER > LASTPROCESSED and 6020 TIE.HEADER < HEADER) into TXKEYS 6022 4. LASTPROCESSED = HEADER 6024 5. if DBTIE not found then 6026 I) if originator is this node then bump_own_tie 6028 II) else put HEADER into REQKEYS 6030 6. if DBTIE.HEADER < HEADER then 6032 I) if originator is this node then bump_own_tie else 6033 i. if this is a N-TIE header from a northbound 6034 neighbor then override DBTIE in LSDB with HEADER 6036 ii. else put HEADER into REQKEYS 6038 7. if DBTIE.HEADER > HEADER then put DBTIE.HEADER into TXKEYS 6040 8. if DBTIE.HEADER = HEADER then 6042 I) if DBTIE has content already then put DBTIE.HEADER 6043 into CLEARKEYS 6045 II) else put HEADER into REQKEYS 6047 c. put all TIEs in LSDB where (TIE.HEADER > LASTPROCESSED and 6048 TIE.HEADER <= TIDE.end_range) into TXKEYS 6050 d. for all TIEs in TXKEYS try_to_transmit_tie(TIE) 6052 e. for all TIEs in REQKEYS request_tie(TIE) 6054 f. for all TIEs in CLEARKEYS remove_from_all_queues(TIE) 6056 C.3.3. TIREs 6058 C.3.3.1. TIRE Generation 6060 There is not much to say here. Elements from both TIES_REQ and 6061 TIES_ACK MUST be collected and sent out as fast as feasible as TIREs. 6062 When sending TIREs with elements from TIES_REQ the `lifetime` field 6063 MUST be set to 0 to force reflooding from the neighbor even if the 6064 TIEs seem to be same. 6066 C.3.3.2. TIRE Processing 6068 On reception of TIREs the following processing is performed: 6070 TXKEYS: Collection of TIE Headers to be send after processing of 6071 the packet 6073 REQKEYS: Collection of TIEIDs to be requested after processing of 6074 the packet 6076 ACKKEYS: Collection of TIEIDs that have been acked 6078 DBTIE: TIE in the LSDB if found 6080 a. for every HEADER in TIRE do 6081 1. DBTIE = find HEADER in current LSDB 6083 2. if DBTIE not found then do nothing 6085 3. if DBTIE.HEADER < HEADER then put HEADER into REQKEYS 6087 4. if DBTIE.HEADER > HEADER then put DBTIE.HEADER into TXKEYS 6089 5. if DBTIE.HEADER = HEADER then put DBTIE.HEADER into ACKKEYS 6091 b. for all TIEs in TXKEYS try_to_transmit_tie(TIE) 6093 c. for all TIEs in REQKEYS request_tie(TIE) 6095 d. for all TIEs in ACKKEYS tie_been_acked(TIE) 6097 C.3.4. TIEs Processing on Flood State Adjacency 6099 On reception of TIEs the following processing is performed: 6101 ACKTIE: TIE to acknowledge 6103 TXTIE: TIE to transmit 6105 DBTIE: TIE in the LSDB if found 6107 a. DBTIE = find TIE in current LSDB 6109 b. if DBTIE not found then 6111 1. if originator is this node then bump_own_tie with a short 6112 remaining lifetime 6114 2. else insert TIE into LSDB and ACKTIE = TIE 6116 else 6118 1. if DBTIE.HEADER = TIE.HEADER then 6120 i. if DBTIE has content already then ACKTIE = TIE 6122 ii. else process like the "DBTIE.HEADER < TIE.HEADER" case 6124 2. if DBTIE.HEADER < TIE.HEADER then 6126 i. if originator is this node then bump_own_tie 6128 ii. else insert TIE into LSDB and ACKTIE = TIE 6130 3. if DBTIE.HEADER > TIE.HEADER then 6132 i. if DBTIE has content already then TXTIE = DBTIE 6134 ii. else ACKTIE = DBTIE 6136 c. if TXTIE is set then try_to_transmit_tie(TXTIE) 6138 d. if ACKTIE is set then ack_tie(TIE) 6140 C.3.5. TIEs Processing When LSDB Received Newer Version on Other 6141 Adjacencies 6143 The Link State Database can be considered to be a switchboard that 6144 does not need any flooding procedures but can be given new versions 6145 of TIEs by a peer. Consecutively, a peer receives from the LSDB 6146 newer versions of TIEs received by other peeers and processes them 6147 (without any filtering) just like receving TIEs from its remote peer. 6148 This publisher model can be implemented in many ways. 6150 C.3.6. Sending TIEs 6152 On a periodic basis all TIEs with lifetime left > 0 MUST be sent out 6153 on the adjacency, removed from TIES_TX list and requeued onto 6154 TIES_RTX list. 6156 Appendix D. Constants 6158 D.1. Configurable Protocol Constants 6160 This section gather constants that are provided in the schema files 6161 and the document. 6163 +----------------+--------------+-----------------------------------+ 6164 | | Type | Value | 6165 +----------------+--------------+-----------------------------------+ 6166 | LIE IPv4 | Default | 224.0.0.120 or all-rift-routers | 6167 | Multicast | Value, | to be assigned in IPv4 Multicast | 6168 | Address | Configurable | Address Space Registry in Local | 6169 | | | Network Control Block | 6170 +----------------+--------------+-----------------------------------+ 6171 | LIE IPv6 | Default | FF02::A1F7 or all-rift-routers to | 6172 | Multicast | Value, | be assigned in IPv6 Multicast | 6173 | Address | Configurable | Address Assignments | 6174 +----------------+--------------+-----------------------------------+ 6175 | LIE | Default | 911 | 6176 | Destination | Value, | | 6177 | Port | Configurable | | 6178 +----------------+--------------+-----------------------------------+ 6179 | Level value | Constant | 24 | 6180 | for | | | 6181 | TOP_OF_FABRIC | | | 6182 | flag | | | 6183 +----------------+--------------+-----------------------------------+ 6184 | Default LIE | Default | 3 seconds | 6185 | Holdtime | Value, | | 6186 | | Configurable | | 6187 +----------------+--------------+-----------------------------------+ 6188 | TIE | Default | 1 second | 6189 | Retransmission | Value | | 6190 | Interval | | | 6191 +----------------+--------------+-----------------------------------+ 6192 | TIDE | Default | 5 seconds | 6193 | Generation | Value, | | 6194 | Interval | Configurable | | 6195 +----------------+--------------+-----------------------------------+ 6196 | MIN_TIEID | Constant | TIE Key with minimal values: | 6197 | signifies | | TIEID(originator=0, | 6198 | start of TIDEs | | tietype=TIETypeMinValue, | 6199 | | | tie_nr=0, direction=South) | 6200 +----------------+--------------+-----------------------------------+ 6201 | MAX_TIEID | Constant | TIE Key with maximal values: | 6202 | signifies end | | TIEID(originator=MAX_UINT64, | 6203 | of TIDEs | | tietype=TIETypeMaxValue, | 6204 | | | tie_nr=MAX_UINT64, | 6205 | | | direction=North) | 6206 +----------------+--------------+-----------------------------------+ 6208 Table 6: all_constants 6210 Appendix E. TODO 6212 o section on E-W superspine/ToF flooding scope to connect partitions 6214 Author's Address 6216 The RIFT Team