idnits 2.17.1 draft-ietf-spring-segment-routing-msdc-04.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 : ---------------------------------------------------------------------------- No issues found here. Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year -- The document date (March 9, 2017) is 2602 days in the past. Is this intentional? Checking references for intended status: Informational ---------------------------------------------------------------------------- -- Looks like a reference, but probably isn't: '16000' on line 276 -- Looks like a reference, but probably isn't: '23999' on line 276 -- Looks like a reference, but probably isn't: '1000' on line 893 -- Looks like a reference, but probably isn't: '1999' on line 893 -- Looks like a reference, but probably isn't: '2000' on line 894 -- Looks like a reference, but probably isn't: '2999' on line 894 == Outdated reference: A later version (-27) exists of draft-ietf-idr-bgp-prefix-sid-04 == Outdated reference: A later version (-15) exists of draft-ietf-spring-segment-routing-11 ** Obsolete normative reference: RFC 3107 (Obsoleted by RFC 8277) == Outdated reference: A later version (-26) exists of draft-ietf-6man-segment-routing-header-05 == Outdated reference: A later version (-10) exists of draft-ietf-spring-segment-routing-central-epe-04 Summary: 1 error (**), 0 flaws (~~), 5 warnings (==), 7 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 Network Working Group C. Filsfils, Ed. 3 Internet-Draft S. Previdi, Ed. 4 Intended status: Informational Cisco Systems, Inc. 5 Expires: September 10, 2017 J. Mitchell 6 Unaffiliated 7 E. Aries 8 Juniper Networks 9 P. Lapukhov 10 Facebook 11 March 9, 2017 13 BGP-Prefix Segment in large-scale data centers 14 draft-ietf-spring-segment-routing-msdc-04 16 Abstract 18 This document describes the motivation and benefits for applying 19 segment routing in BGP-based large-scale data-centers. It describes 20 the design to deploy segment routing in those data-centers, for both 21 the MPLS and IPv6 dataplanes. 23 Requirements Language 25 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 26 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 27 document are to be interpreted as described in RFC 2119 [RFC2119]. 29 Status of This Memo 31 This Internet-Draft is submitted in full conformance with the 32 provisions of BCP 78 and BCP 79. 34 Internet-Drafts are working documents of the Internet Engineering 35 Task Force (IETF). Note that other groups may also distribute 36 working documents as Internet-Drafts. The list of current Internet- 37 Drafts is at http://datatracker.ietf.org/drafts/current/. 39 Internet-Drafts are draft documents valid for a maximum of six months 40 and may be updated, replaced, or obsoleted by other documents at any 41 time. It is inappropriate to use Internet-Drafts as reference 42 material or to cite them other than as "work in progress." 44 This Internet-Draft will expire on September 10, 2017. 46 Copyright Notice 48 Copyright (c) 2017 IETF Trust and the persons identified as the 49 document authors. All rights reserved. 51 This document is subject to BCP 78 and the IETF Trust's Legal 52 Provisions Relating to IETF Documents 53 (http://trustee.ietf.org/license-info) in effect on the date of 54 publication of this document. Please review these documents 55 carefully, as they describe your rights and restrictions with respect 56 to this document. Code Components extracted from this document must 57 include Simplified BSD License text as described in Section 4.e of 58 the Trust Legal Provisions and are provided without warranty as 59 described in the Simplified BSD License. 61 Table of Contents 63 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 3 64 2. Large Scale Data Center Network Design Summary . . . . . . . 3 65 2.1. Reference design . . . . . . . . . . . . . . . . . . . . 4 66 3. Some open problems in large data-center networks . . . . . . 5 67 4. Applying Segment Routing in the DC with MPLS dataplane . . . 6 68 4.1. BGP Prefix Segment (BGP-Prefix-SID) . . . . . . . . . . . 6 69 4.2. eBGP Labeled Unicast (RFC3107) . . . . . . . . . . . . . 7 70 4.2.1. Control Plane . . . . . . . . . . . . . . . . . . . . 7 71 4.2.2. Data Plane . . . . . . . . . . . . . . . . . . . . . 9 72 4.2.3. Network Design Variation . . . . . . . . . . . . . . 10 73 4.2.4. Global BGP Prefix Segment through the fabric . . . . 10 74 4.2.5. Incremental Deployments . . . . . . . . . . . . . . . 11 75 4.3. iBGP Labeled Unicast (RFC3107) . . . . . . . . . . . . . 12 76 5. Applying Segment Routing in the DC with IPv6 dataplane . . . 14 77 6. Communicating path information to the host . . . . . . . . . 14 78 7. Addressing the open problems . . . . . . . . . . . . . . . . 15 79 7.1. Per-packet and flowlet switching . . . . . . . . . . . . 15 80 7.2. Performance-aware routing . . . . . . . . . . . . . . . . 16 81 7.3. Deterministic network probing . . . . . . . . . . . . . . 17 82 8. Additional Benefits . . . . . . . . . . . . . . . . . . . . . 18 83 8.1. MPLS Dataplane with operational simplicity . . . . . . . 18 84 8.2. Minimizing the FIB table . . . . . . . . . . . . . . . . 18 85 8.3. Egress Peer Engineering . . . . . . . . . . . . . . . . . 18 86 8.4. Anycast . . . . . . . . . . . . . . . . . . . . . . . . . 19 87 9. Preferred SRGB Allocation . . . . . . . . . . . . . . . . . . 19 88 10. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 20 89 11. Manageability Considerations . . . . . . . . . . . . . . . . 20 90 12. Security Considerations . . . . . . . . . . . . . . . . . . . 21 91 13. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 21 92 14. Contributors . . . . . . . . . . . . . . . . . . . . . . . . 21 93 15. References . . . . . . . . . . . . . . . . . . . . . . . . . 22 94 15.1. Normative References . . . . . . . . . . . . . . . . . . 22 95 15.2. Informative References . . . . . . . . . . . . . . . . . 23 96 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 24 98 1. Introduction 100 Segment Routing (SR), as described in 101 [I-D.ietf-spring-segment-routing] leverages the source routing 102 paradigm. A node steers a packet through an ordered list of 103 instructions, called segments. A segment can represent any 104 instruction, topological or service-based. A segment can have a 105 local semantic to an SR node or global within an SR domain. SR 106 allows to enforce a flow through any topological path and service 107 chain while maintaining per-flow state only at the ingress node to 108 the SR domain. Segment Routing can be applied to the MPLS and IPv6 109 data-planes. 111 The use-cases described in this document should be considered in the 112 context of the BGP-based large-scale data-center (DC) design 113 described in [RFC7938]. We extend it by applying SR both with IPv6 114 and MPLS dataplane. 116 2. Large Scale Data Center Network Design Summary 118 This section provides a brief summary of the informational document 119 [RFC7938] that outlines a practical network design suitable for data- 120 centers of various scales: 122 o Data-center networks have highly symmetric topologies with 123 multiple parallel paths between two server attachment points. The 124 well-known Clos topology is most popular among the operators (as 125 described in [RFC7938]). In a Clos topology, the minimum number 126 of parallel paths between two elements is determined by the 127 "width" of the "Tier-1" stage. See Figure 1 below for an 128 illustration of the concept. 130 o Large-scale data-centers commonly use a routing protocol, such as 131 BGP4 [RFC4271] in order to provide endpoint connectivity. 132 Recovery after a network failure is therefore driven either by 133 local knowledge of directly available backup paths or by 134 distributed signaling between the network devices. 136 o Within data-center networks, traffic is load-shared using the 137 Equal Cost Multipath (ECMP) mechanism. With ECMP, every network 138 device implements a pseudo-random decision, mapping packets to one 139 of the parallel paths by means of a hash function calculated over 140 certain parts of the packet, typically a combination of various 141 packet header fields. 143 The following is a schematic of a five-stage Clos topology, with four 144 devices in the "Tier-1" stage. Notice that number of paths between 145 Node1 and Node12 equals to four: the paths have to cross all of 146 Tier-1 devices. At the same time, the number of paths between Node1 147 and Node2 equals two, and the paths only cross Tier-2 devices. Other 148 topologies are possible, but for simplicity we'll only look into the 149 topologies that have a single path from Tier-1 to Tier-3. The rest 150 could be treated similarly, with a few modifications to the logic. 152 2.1. Reference design 154 Tier-1 155 +-----+ 156 |NODE | 157 +->| 5 |--+ 158 | +-----+ | 159 Tier-2 | | Tier-2 160 +-----+ | +-----+ | +-----+ 161 +------------>|NODE |--+->|NODE |--+--|NODE |-------------+ 162 | +-----| 3 |--+ | 6 | +--| 9 |-----+ | 163 | | +-----+ +-----+ +-----+ | | 164 | | | | 165 | | +-----+ +-----+ +-----+ | | 166 | +-----+---->|NODE |--+ |NODE | +--|NODE |-----+-----+ | 167 | | | +---| 4 |--+->| 7 |--+--| 10 |---+ | | | 168 | | | | +-----+ | +-----+ | +-----+ | | | | 169 | | | | | | | | | | 170 +-----+ +-----+ | +-----+ | +-----+ +-----+ 171 |NODE | |NODE | Tier-3 +->|NODE |--+ Tier-3 |NODE | |NODE | 172 | 1 | | 2 | | 8 | | 11 | | 12 | 173 +-----+ +-----+ +-----+ +-----+ +-----+ 174 | | | | | | | | 175 A O B O <- Servers -> Z O O O 177 Figure 1: 5-stage Clos topology 179 In the reference topology illustrated in Figure 1, we assume: 181 o Each node is its own AS (Node X has AS X). 4-byte AS numbers are 182 recommended ([RFC6793]). 184 * For simple and efficient route propagation filtering, Node5, 185 Node6, Node7 and Node8 use the same AS, Node3 and Node4 use the 186 same AS, Node9 and Node10 use the same AS. 188 * In case of 2-byte autonomous system numbers are used and for 189 efficient usage of the scarce 2-byte Private Use AS pool, 190 different Tier-3 nodes might use the same AS. 192 * Without loss of generality, we will simplify these details in 193 this document and assume that each node has its own AS. 195 o Each node peers with its neighbors with a BGP session. If not 196 specified, eBGP is assumed. In a specific use-case, iBGP will be 197 used but this will be called out explicitly in that case. 199 o Each node originates the IPv4 address of its loopback interface 200 into BGP and announces it to its neighbors. 202 * The loopback of Node X is 192.0.2.x/32. 204 In this document, we also refer to the Tier-1, Tier-2 and Tier-3 205 nodes respectively as Spine, Leaf and ToR (top of rack) nodes. When 206 a ToR node acts as a gateway to the "outside world", we call it a 207 border node. 209 3. Some open problems in large data-center networks 211 The data-center network design summarized above provides means for 212 moving traffic between hosts with reasonable efficiency. There are 213 few open performance and reliability problems that arise in such 214 design: 216 o ECMP routing is most commonly realized per-flow. This means that 217 large, long-lived "elephant" flows may affect performance of 218 smaller, short-lived "mouse" flows and reduce efficiency of per- 219 flow load-sharing. In other words, per-flow ECMP does not perform 220 efficiently when flow lifetime distribution is heavy-tailed. 221 Furthermore, due to hash-function inefficiencies it is possible to 222 have frequent flow collisions, where more flows get placed on one 223 path over the others. 225 o Shortest-path routing with ECMP implements an oblivious routing 226 model, which is not aware of the network imbalances. If the 227 network symmetry is broken, for example due to link failures, 228 utilization hotspots may appear. For example, if a link fails 229 between Tier-1 and Tier-2 devices (e.g. Node5 and Node9), Tier-3 230 devices Node1 and Node2 will not be aware of that, since there are 231 other paths available from perspective of Node3. They will 232 continue sending roughly equal traffic to Node3 and Node4 as if 233 the failure didn't exist which may cause a traffic hotspot. 235 o The absence of path visibility leaves transport protocols, such as 236 TCP, with a "blackbox" view of the network. Some TCP metrics, 237 such as SRTT, MSS, CWND and few others could be inferred and 238 cached based on past history, but those apply to destinations, 239 regardless of the path that has been chosen to get there. Thus, 240 for instance, TCP is not capable of remembering "bad" paths, such 241 as those that exhibited poor performance in the past. This means 242 that every new connection will be established obliviously (memory- 243 less) with regards to the paths chosen before, or chosen by other 244 nodes. 246 o Isolating faults in the network with multiple parallel paths and 247 ECMP-based routing is non-trivial due to lack of determinism. 248 Specifically, the connections from HostA to HostB may take a 249 different path every time a new connection is formed, thus making 250 consistent reproduction of a failure much more difficult. This 251 complexity scales linearly with the number of parallel paths in 252 the network, and stems from the random nature of path selection by 253 the network devices. 255 Further in this document, we are going to demonstrate how these 256 problems could be addressed within the framework of Segment Routing. 258 First, we will explain how to apply SR in the DC, for MPLS and IPv6 259 data-planes. 261 4. Applying Segment Routing in the DC with MPLS dataplane 263 4.1. BGP Prefix Segment (BGP-Prefix-SID) 265 A BGP Prefix Segment is a segment associated with a BGP prefix. A 266 BGP Prefix Segment is a network-wide instruction to forward the 267 packet along the ECMP-aware best path to the related prefix. 269 The BGP Prefix Segment is defined as the BGP-Prefix-SID Attribute in 270 [I-D.ietf-idr-bgp-prefix-sid] which contains an index. Throughout 271 this document the BGP Prefix Segment Attribute is referred as the 272 BGP-Prefix-SID and the encoded index as the label-index. 274 In this document, we make the network design decision to assume that 275 all the nodes are allocated the same SRGB (Segment Routing Global 276 Block), e.g. [16000, 23999]. This provides operational 277 simplification as explained in Section 9, but this is not a 278 requirement. 280 For illustration purpose, when considering an MPLS data-plane, we 281 assume that the label-index allocated to prefix 192.0.2.x/32 is X. 282 As a result, a local label (16000+x) is allocated for prefix 283 192.0.2.x/32 by each node throughout the DC fabric. 285 When IPv6 data-plane is considered, we assume that Node X is 286 allocated IPv6 address (segment) 2001:DB8::X. 288 4.2. eBGP Labeled Unicast (RFC3107) 290 Referring to Figure 1 and [RFC7938], the following design 291 modifications are introduced: 293 o Each node peers with its neighbors via a eBGP session with 294 extensions defined in [RFC3107] (named "eBGP3107" throughout this 295 document) and with the BGP-Prefix-SID attribute extension defined 296 in this document. 298 o The forwarding plane at Tier-2 and Tier-1 is MPLS. 300 o The forwarding plane at Tier-3 is either IP2MPLS (if the host 301 sends IP traffic) or MPLS2MPLS (if the host sends MPLS- 302 encapsulated traffic). 304 Figure 2 zooms into a path from server A to server Z within the 305 topology of Figure 1. 307 +-----+ +-----+ +-----+ 308 +---------->|NODE | |NODE | |NODE | 309 | | 4 |--+->| 7 |--+--| 10 |---+ 310 | +-----+ +-----+ +-----+ | 311 | | 312 +-----+ +-----+ 313 |NODE | |NODE | 314 | 1 | | 11 | 315 +-----+ +-----+ 316 | | 317 A <- Servers -> Z 319 Figure 2: Path from A to Z via nodes 1, 4, 7, 10 and 11 321 Referring to Figure 1 and Figure 2 and assuming the IP address with 322 the AS and label-index allocation previously described, the following 323 sections detail the control plane operation and the data plane states 324 for the prefix 192.0.2.11/32 (loopback of Node11) 326 4.2.1. Control Plane 328 Node11 originates 192.0.2.11/32 in BGP and allocates to it a BGP- 329 Prefix-SID with label-index: index11) [I-D.ietf-idr-bgp-prefix-sid]. 331 Node11 sends the following eBGP3107 update to Node10: 333 . IP Prefix: 192.0.2.11/32 334 . Label: Implicit-Null 335 . Next-hop: Node11's interface address on the link to Node10 336 . AS Path: {11} 337 . BGP-Prefix-SID: Label-Index 11 339 Node10 receives the above update. As it is SR capable, Node10 is 340 able to interpret the BGP-Prefix-SID and hence understands that it 341 should allocate the label from its own SRGB block, offset by the 342 Label-Index received in the BGP-Prefix-SID (16000+11 hence 16011) to 343 the NLRI instead of allocating a non-deterministic label out of a 344 dynamically allocated portion of the local label space. The 345 implicit-null label in the NLRI tells Node10 that it is the 346 penultimate hop and MUST pop the top label on the stack before 347 forwarding traffic for this prefix to Node11. 349 Then, Node10 sends the following eBGP3107 update to Node7: 351 . IP Prefix: 192.0.2.11/32 352 . Label: 16011 353 . Next-hop: Node10's interface address on the link to Node7 354 . AS Path: {10, 11} 355 . BGP-Prefix-SID: Label-Index 11 357 Node7 receives the above update. As it is SR capable, Node7 is able 358 to interpret the BGP-Prefix-SID and hence allocates the local 359 (incoming) label 16011 (16000 + 11) to the NLRI (instead of 360 allocating a "dynamic" local label from its label manager). Node7 361 uses the label in the received eBGP3107 NLRI as the outgoing label 362 (the index is only used to derive the local/incoming label). 364 Node7 sends the following eBGP3107 update to Node4: 366 . IP Prefix: 192.0.2.11/32 367 . Label: 16011 368 . Next-hop: Node7's interface address on the link to Node4 369 . AS Path: {7, 10, 11} 370 . BGP-Prefix-SID: Label-Index 11 372 Node4 receives the above update. As it is SR capable, Node4 is able 373 to interpret the BGP-Prefix-SID and hence allocates the local 374 (incoming) label 16011 to the NLRI (instead of allocating a "dynamic" 375 local label from its label manager). Node4 uses the label in the 376 received eBGP3107 NLRI as outgoing label (the index is only used to 377 derive the local/incoming label). 379 Node4 sends the following eBGP3107 update to Node1: 381 . IP Prefix: 192.0.2.11/32 382 . Label: 16011 383 . Next-hop: Node4's interface address on the link to Node1 384 . AS Path: {4, 7, 10, 11} 385 . BGP-Prefix-SID: Label-Index 11 387 Node1 receives the above update. As it is SR capable, Node1 is able 388 to interpret the BGP-Prefix-SID and hence allocates the local 389 (incoming) label 16011 to the NLRI (instead of allocating a "dynamic" 390 local label from its label manager). Node1 uses the label in the 391 received eBGP3107 NLRI as outgoing label (the index is only used to 392 derive the local/incoming label). 394 4.2.2. Data Plane 396 Referring to Figure 1, and assuming all nodes apply the same 397 advertisement rules described above and all nodes have the same SRGB 398 (16000-23999), here are the IP/MPLS forwarding tables for prefix 399 192.0.2.11/32 at Node1, Node4, Node7 and Node10. 401 ----------------------------------------------- 402 Incoming label | outgoing label | Outgoing 403 or IP destination | | Interface 404 ------------------+----------------+----------- 405 16011 | 16011 | ECMP{3, 4} 406 192.0.2.11/32 | 16011 | ECMP{3, 4} 407 ------------------+----------------+----------- 409 Figure 3: Node1 Forwarding Table 411 ----------------------------------------------- 412 Incoming label | outgoing label | Outgoing 413 or IP destination | | Interface 414 ------------------+----------------+----------- 415 16011 | 16011 | ECMP{7, 8} 416 192.0.2.11/32 | 16011 | ECMP{7, 8} 417 ------------------+----------------+----------- 419 Figure 4: Node4 Forwarding Table 421 ----------------------------------------------- 422 Incoming label | outgoing label | Outgoing 423 or IP destination | | Interface 424 ------------------+----------------+----------- 425 16011 | 16011 | 10 426 192.0.2.11/32 | 16011 | 10 427 ------------------+----------------+----------- 429 Figure 5: Node7 Forwarding Table 431 ----------------------------------------------- 432 Incoming label | outgoing label | Outgoing 433 or IP destination | | Interface 434 ------------------+----------------+----------- 435 16011 | POP | 11 436 192.0.2.11/32 | N/A | 11 437 ------------------+----------------+----------- 439 Node10 Forwarding Table 441 4.2.3. Network Design Variation 443 A network design choice could consist of switching all the traffic 444 through Tier-1 and Tier-2 as MPLS traffic. In this case, one could 445 filter away the IP entries at Node4, Node7 and Node10. This might be 446 beneficial in order to optimize the forwarding table size. 448 A network design choice could consist in allowing the hosts to send 449 MPLS-encapsulated traffic based on the Egress Peer Engineering (EPE) 450 use-case as defined in [I-D.ietf-spring-segment-routing-central-epe]. 451 For example, applications at HostA would send their Z-destined 452 traffic to Node1 with an MPLS label stack where the top label is 453 16011 and the next label is an EPE peer segment 454 ([I-D.ietf-spring-segment-routing-central-epe]) at Node11 directing 455 the traffic to Z. 457 4.2.4. Global BGP Prefix Segment through the fabric 459 When the previous design is deployed, the operator enjoys global BGP- 460 Prefix-SID and label allocation throughout the DC fabric. 462 A few examples follow: 464 o Normal forwarding to Node11: a packet with top label 16011 465 received by any node in the fabric will be forwarded along the 466 ECMP-aware BGP best-path towards Node11 and the label 16011 is 467 penultimate-popped at Node10. 469 o Traffic-engineered path to Node11: an application on a host behind 470 Node1 might want to restrict its traffic to paths via the Spine 471 node Node5. The application achieves this by sending its packets 472 with a label stack of {16005, 16011}. BGP Prefix SID 16005 directs 473 the packet up to Node5 along the path (Node1, Node3, Node5). BGP- 474 Prefix-SID 16011 then directs the packet down to Node11 along the 475 path (Node5, Node9, Node11). 477 4.2.5. Incremental Deployments 479 The design previously described can be deployed incrementally. Let 480 us assume that Node7 does not support the BGP-Prefix-SID and let us 481 show how the fabric connectivity is preserved. 483 From a signaling viewpoint, nothing would change: even though Node7 484 does not support the BGP-Prefix-SID, it does propagate the attribute 485 unmodified to its neighbors. 487 From a label allocation viewpoint, the only difference is that Node7 488 would allocate a dynamic (random) label to the prefix 192.0.2.11/32 489 (e.g. 123456) instead of the "hinted" label as instructed by the BGP- 490 Prefix-SID. The neighbors of Node7 adapt automatically as they 491 always use the label in the BGP3107 NLRI as outgoing label. 493 Node4 does understand the BGP-Prefix-SID and hence allocates the 494 indexed label in the SRGB (16011) for 192.0.2.11/32. 496 As a result, all the data-plane entries across the network would be 497 unchanged except the entries at Node7 and its neighbor Node4 as shown 498 in the figures below. 500 The key point is that the end-to-end Label Switched Path (LSP) is 501 preserved because the outgoing label is always derived from the 502 received label within the BGP3107 NLRI. The index in the BGP-Prefix- 503 SID is only used as a hint on how to allocate the local label (the 504 incoming label) but never for the outgoing label. 506 ------------------------------------------ 507 Incoming label | outgoing | Outgoing 508 or IP destination | label | Interface 509 -------------------+---------------------- 510 12345 | 16011 | 10 512 Figure 7: Node7 Forwarding Table 514 ------------------------------------------ 515 Incoming label | outgoing | Outgoing 516 or IP destination | label | Interface 517 -------------------+---------------------- 518 16011 | 12345 | 7 520 Figure 8: Node4 Forwarding Table 522 The BGP-Prefix-SID can thus be deployed incrementally one node at a 523 time. 525 When deployed together with a homogeneous SRGB (same SRGB across the 526 fabric), the operator incrementally enjoys the global prefix segment 527 benefits as the deployment progresses through the fabric. 529 4.3. iBGP Labeled Unicast (RFC3107) 531 The same exact design as eBGP3107 is used with the following 532 modifications: 534 All nodes use the same AS number. 536 Each node peers with its neighbors via an internal BGP session 537 (iBGP) with extensions defined in [RFC3107] (named "iBGP3107" 538 throughout this document) and with the BGP-Prefix-SID attribute 539 extension defined in this document. 541 Each node acts as a route-reflector for each of its neighbors and 542 with the next-hop-self option. Next-hop-self is a well known 543 operational feature which consists of rewriting the next-hop of a 544 BGP update prior to send it to the neighbor. Usually, it's a 545 common practice to apply next-hop-self behavior towards iBGP peers 546 for eBGP learned routes. In the case outlined in this section we 547 propose to use the next-hop-self mechanism also to iBGP learned 548 routes. 550 Cluster-1 551 +-----------+ 552 | Tier-1 | 553 | +-----+ | 554 | |NODE | | 555 | | 5 | | 556 Cluster-2 | +-----+ | Cluster-3 557 +---------+ | | +---------+ 558 | Tier-2 | | | | Tier-2 | 559 | +-----+ | | +-----+ | | +-----+ | 560 | |NODE | | | |NODE | | | |NODE | | 561 | | 3 | | | | 6 | | | | 9 | | 562 | +-----+ | | +-----+ | | +-----+ | 563 | | | | | | 564 | | | | | | 565 | +-----+ | | +-----+ | | +-----+ | 566 | |NODE | | | |NODE | | | |NODE | | 567 | | 4 | | | | 7 | | | | 10 | | 568 | +-----+ | | +-----+ | | +-----+ | 569 +---------+ | | +---------+ 570 | | 571 | +-----+ | 572 | |NODE | | 573 Tier-3 | | 8 | | Tier-3 574 +-----+ +-----+ | +-----+ | +-----+ +-----+ 575 |NODE | |NODE | +-----------+ |NODE | |NODE | 576 | 1 | | 2 | | 11 | | 12 | 577 +-----+ +-----+ +-----+ +-----+ 579 Figure 9: iBGP Sessions with Reflection and Next-Hop-Self 581 For simple and efficient route propagation filtering and as 582 illustrated in Figure 9: 584 Node5, Node6, Node7 and Node8 use the same Cluster ID (Cluster- 585 1) 587 Node3 and Node4 use the same Cluster ID (Cluster-2) 589 Node9 and Node10 use the same Cluster ID (Cluster-3) 591 AIGP metric ([RFC7311]) is likely applied to the BGP-Prefix-SID as 592 part of a large-scale multi-domain design such as Seamless MPLS 593 [I-D.ietf-mpls-seamless-mpls]. 595 The control-plane behavior is mostly the same as described in the 596 previous section: the only difference is that the eBGP3107 path 597 propagation is simply replaced by an iBGP3107 path reflection with 598 next-hop changed to self. 600 The data-plane tables are exactly the same. 602 5. Applying Segment Routing in the DC with IPv6 dataplane 604 The design described in [RFC7938] is reused with one single 605 modification. We highlight it using the example of the reachability 606 to Node11 via spine node Node5. 608 Node5 originates 2001:DB8::5/128 with the attached BGP-Prefix-SID 609 advertising the support of the Segment Routing extension header (SRH, 610 [I-D.ietf-6man-segment-routing-header]) for IPv6 packets destined to 611 segment 2001:DB8::5 ([I-D.ietf-idr-bgp-prefix-sid]). 613 Tor11 originates 2001:DB8::11/128 with the attached BGP-Prefix-SID 614 advertising the support of the SRH for IPv6 packets destined to 615 segment 2001:DB8::11. 617 The control-plane and data-plane processing of all the other nodes in 618 the fabric is unchanged. Specifically, the routes to 2001:DB8::5 and 619 2001:DB8::11 are installed in the FIB along the eBGP best-path to 620 Node5 (spine node) and Node11 (ToR node) respectively. 622 An application on HostA which needs to send traffic to HostZ via only 623 Node5 (spine node) can do so by sending IPv6 packets with a SRH 624 extension header. The destination address and active segment is set 625 to 2001:DB8::5. The next and last segment is set to 2001:DB8::11. 627 The application must only use IPv6 addresses that have been 628 advertised as capable for SRv6 segment processing (e.g. for which the 629 BGP prefix segment capability has been advertised). How applications 630 learn this (e.g.: centralized controller and orchestration) is 631 outside the scope of this document. 633 6. Communicating path information to the host 635 There are two general methods for communicating path information to 636 the end-hosts: "proactive" and "reactive", aka "push" and "pull" 637 models. There are multiple ways to implement either of these 638 methods. Here, we note that one way could be using a centralized 639 controller: the controller either tells the hosts of the prefix-to- 640 path mappings beforehand and updates them as needed (network event 641 driven push), or responds to the hosts making request for a path to 642 specific destination (host event driven pull). It is also possible 643 to use a hybrid model, i.e., pushing some state from the controller 644 in response to particular network events, while the host pulls other 645 state on demand. 647 We note, that when disseminating network-related data to the end- 648 hosts a trade-off is made to balance the amount of information Vs. 649 the level of visibility in the network state. This applies both to 650 push and pull models. In the extreme case, the host would request 651 path information on every flow, and keep no local state at all. On 652 the other end of the spectrum, information for every prefix in the 653 network along with available paths could be pushed and continuously 654 updated on all hosts. 656 7. Addressing the open problems 658 This section demonstrates how the problems describe above could be 659 solved using the segment routing concept. It is worth noting that 660 segment routing signaling and data-plane are only parts of the 661 solution. Additional enhancements, e.g., such as the centralized 662 controller mentioned previously, and host networking stack support 663 are required to implement the proposed solutions. 665 7.1. Per-packet and flowlet switching 667 A flowlet is defined as a burst of packets from the same flow 668 followed by an idle interval. [KANDULA04] developed a scheme that 669 uses flowlets to split traffic across multiple parallel paths in 670 order to optimize traffic load sharing. 672 With the ability to choose paths on the host, one may go from per- 673 flow load-sharing in the network to per-packet or per-flowlet. The 674 host may select different segment routing instructions either per 675 packet, or per flowlet, and route them over different paths. This 676 allows for solving the "elephant flow" problem in the data-center and 677 avoiding link imbalances. 679 Note that traditional ECMP routing could be easily simulated with on- 680 host path selection, using method proposed in [GREENBERG09]. The 681 hosts would randomly pick a Tier-2 or Tier-1 device to "bounce" the 682 packet off of, depending on whether the destination is under the same 683 Tier-2 nodes, or has to be reached across Tier-1. The host would use 684 a hash function that operates on per-flow invariants, to simulate 685 per-flow load-sharing in the network. 687 Using Figure 1 as reference, let us illustrate this concept assuming 688 that HostA has an elephant flow to HostZ called Flow-f. 690 Normally, a flow is hashed on to a single path. Let's assume HostA 691 sends its packets associated with Flow-f with top label 16011 (the 692 label for the remote ToR, Node11, where HostZ is connected) and Node1 693 would hash all the packets of Flow-F via the same next-hop (e.g. 694 Node3). Similarly, let's assume that leaf Node3 would hash all the 695 packets of Flow-F via the same next-hop (e.g.: spine node Node5). 696 This normal operation would restrict the elephant flow on a small 697 subset of the ECMP paths to HostZ and potentially create imbalance 698 and congestion in the fabric. 700 Leveraging the flowlet proposal, assuming HostA is made aware of 4 701 disjoint paths via intermediate segment 16005, 16006, 16007 and 16008 702 (the BGP prefix SID's of the 4 spine nodes) and also made aware of 703 the prefix segment of the remote ToR connected to the destination 704 (16011), then the application can break the elephant flow F into 705 flowlets F1, F2, F3, F4 and associate each flowlet with one of the 706 following 4 label stacks: {16005, 16011}, {16006, 16011}, {16007, 707 16011} and {16008, 16011}. This would spread the load of the elephant 708 flow through all the ECMP paths available in the fabric and re- 709 balance the load. 711 7.2. Performance-aware routing 713 Knowing the path associated with flows/packets, the end host may 714 deduce certain characteristics of the path on its own, and 715 additionally use the information supplied with path information 716 pushed from the controller or received via pull request. The host 717 may further share its path observations with the centralized agent, 718 so that the latter may keep up-to-date network health map to assist 719 other hosts with this information. 721 For example, an application A.1 at HostA may pin a TCP flow destined 722 to HostZ via Spine node Node5 using label stack {16005, 16011}. The 723 application A.1 may collect information on packet loss, deduced from 724 TCP retransmissions and other signals (e.g. RTT increases). A.1 may 725 additionally publish this information to a centralized agent, e.g. 726 after a flow completes, or periodically for longer lived flows. 727 Next, using both local and/or global performance data, application 728 A.1 as well as other applications sharing the same resources in the 729 DC fabric may pick up the best path for the new flow, or update an 730 existing path (e.g.: when informed of congestion on an existing 731 path). 733 One particularly interesting instance of performance-aware routing is 734 dynamic fault-avoidance. If some links or devices in the network 735 start discarding packets due to a fault, the end-hosts could probe 736 and detect the path(s) that are affected and hence steer the affected 737 flows away from the problem spot. Similar logic applies to failure 738 cases where packets get completely black-holed, e.g., when a link 739 goes down and the failure is detected by the host while probing the 740 path. 742 For example, an application A.1 informed about 5 paths to Z {16005, 743 16011}, {16006, 16011}, {16007, 16011}, {16008, 16011} and {16011} 744 might use the last one by default (for simplicity). When performance 745 is degrading, A.1 might then start to pin TCP flows to each of the 4 746 other paths (each via a distinct spine) and monitor the performance. 747 It would then detect the faulty path and assign a negative preference 748 to the faulty path to avoid further flows using it. Gradually, over 749 time, it may re-assign flows on the faulty path to eventually detect 750 the resolution of the trouble and start reusing the path. 752 By leveraging Segment Routing, one avoids issues associated with 753 oblivious ECMP hashing. For example, if in the topology depicted on 754 Figure 1 a link between spine node Node5 and leaf node Node9 fails, 755 HostA may exclude the segment corresponding to Node5 from the prefix 756 matching the servers under Tier-2 devices Node9. In the push path 757 discovery model, the affected path mappings may be explicitly pushed 758 to all the servers for the duration of the failure. The new mapping 759 would instruct them to avoid the particular Tier-1 node until the 760 link has recovered. Alternatively, in pull path, the centralized 761 controller may start steering new flows immediately after it 762 discovers the issue. Until then, the existing flows may recover 763 using local detection of the path issues. 765 7.3. Deterministic network probing 767 Active probing is a well-known technique for monitoring network 768 elements' health, constituting of sending continuous packet streams 769 simulating network traffic to the hosts in the data-center. Segment 770 routing makes possible to prescribe the exact paths that each probe 771 or series of probes would be taking toward their destination. This 772 allows for fast correlation and detection of failed paths, by 773 processing information from multiple actively probing agents. This 774 complements the data collected from the hosts routing stacks as 775 described in Section 7.2. 777 For example, imagine a probe agent sending packets to all machines in 778 the data-center. For every host, it may send packets over each of 779 the possible paths, knowing exactly which links and devices these 780 packets will be crossing. Correlating results for multiple 781 destinations with the topological data, it may automatically isolate 782 possible problem to a link or device in the network. 784 8. Additional Benefits 786 8.1. MPLS Dataplane with operational simplicity 788 As required by [RFC7938], no new signaling protocol is introduced. 789 The BGP-Prefix-SID is a lightweight extension to BGP Labeled Unicast 790 (RFC3107 [RFC3107]). It applies either to eBGP or iBGP based 791 designs. 793 Specifically, LDP and RSVP-TE are not used. These protocols would 794 drastically impact the operational complexity of the Data Center and 795 would not scale. This is in line with the requirements expressed in 796 [RFC7938]. 798 Provided the same SRGB is configured on all nodes, all nodes use the 799 same MPLS label for a given IP prefix. This is simpler from an 800 operation standpoint, as discussed in Section 9 802 8.2. Minimizing the FIB table 804 The designer may decide to switch all the traffic at Tier-1 and Tier- 805 2's based on MPLS, hence drastically decreasing the IP table size at 806 these nodes. 808 This is easily accomplished by encapsulating the traffic either 809 directly at the host or the source ToR node by pushing the BGP- 810 Prefix-SID of the destination ToR for intra-DC traffic, or the BGP- 811 Prefix-SID for the the border node for inter-DC or DC-to-outside- 812 world traffic. 814 8.3. Egress Peer Engineering 816 It is straightforward to combine the design illustrated in this 817 document with the Egress Peer Engineering (EPE) use-case described in 818 [I-D.ietf-spring-segment-routing-central-epe]. 820 In such case, the operator is able to engineer its outbound traffic 821 on a per host-flow basis, without incurring any additional state at 822 intermediate points in the DC fabric. 824 For example, the controller only needs to inject a per-flow state on 825 the HostA to force it to send its traffic destined to a specific 826 Internet destination D via a selected border node (say Node12 in 827 Figure 1 instead of another border node, Node11) and a specific 828 egress peer of Node12 (say peer AS 9999 of local PeerNode segment 829 9999 at Node12 instead of any other peer which provides a path to the 830 destination D). Any packet matching this state at host A would be 831 encapsulated with SR segment list (label stack) {16012, 9999}. 16012 832 would steer the flow through the DC fabric, leveraging any ECMP, 833 along the best path to border node Node12. Once the flow gets to 834 border node Node12, the active segment is 9999 (because of PHP on the 835 upstream neighbor of Node12). This EPE PeerNode segment forces 836 border node Node12 to forward the packet to peer AS 9999, without any 837 IP lookup at the border node. There is no per-flow state for this 838 engineered flow in the DC fabric. A benefit of segment routing is 839 the per-flow state is only required at the source. 841 As well as allowing full traffic engineering control such a design 842 also offers FIB table minimization benefits as the Internet-scale FIB 843 at border node Node12 is not required if all FIB lookups are avoided 844 there by using EPE. 846 8.4. Anycast 848 The design presented in this document preserves the availability and 849 load-balancing properties of the base design presented in 850 [I-D.ietf-spring-segment-routing]. 852 For example, one could assign an anycast loopback 192.0.2.20/32 and 853 associate segment index 20 to it on the border Node11 and Node12 (in 854 addition to their node-specific loopbacks). Doing so, the EPE 855 controller could express a default "go-to-the-Internet via any border 856 node" policy as segment list {16020}. Indeed, from any host in the DC 857 fabric or from any ToR node, 16020 steers the packet towards the 858 border Node11 or Node12 leveraging ECMP where available along the 859 best paths to these nodes. 861 9. Preferred SRGB Allocation 863 In the MPLS case, we recommend to use same SRGBs at each node. 865 Different SRGBs in each node likely increase the complexity of the 866 solution both from an operational viewpoint and from a controller 867 viewpoint. 869 From an operation viewpoint, it is much simpler to have the same 870 global label at every node for the same destination (the MPLS 871 troubleshooting is then similar to the IPv6 troubleshooting where 872 this global property is a given). 874 From a controller viewpoint, this allows us to construct simple 875 policies applicable across the fabric. 877 Let us consider two applications A and B respectively connected to 878 ToR1 and ToR2. A has two flows FA1 and FA2 destined to Z. B has two 879 flows FB1 and FB2 destined to Z. The controller wants FA1 and FB1 to 880 be load-shared across the fabric while FA2 and FB2 must be 881 respectively steered via Node5 and Node8. 883 Assuming a consistent unique SRGB across the fabric as described in 884 the document, the controller can simply do it by instructing A and B 885 to use {16011} respectively for FA1 and FB1 and by instructing A and 886 B to use {16005 16011} and {16008 16011} respectively for FA2 and 887 FB2. 889 Let us assume a design where the SRGB is different at every node and 890 where the SRGB of each node is advertised using the Originator SRGB 891 TLV of the BGP-Prefix-SID as defined in 892 [I-D.ietf-idr-bgp-prefix-sid]: SRGB of Node K starts at value K*1000 893 and the SRGB length is 1000 (e.g. ToR1's SRGB is [1000, 1999], 894 ToR2's SRGB is [2000, 2999], ...). 896 In this case, not only the controller would need to collect and store 897 all of these different SRGB's (e.g., through the Originator SRGB TLV 898 of the BGP-Prefix-SID), furthermore it would need to adapt the policy 899 for each host. Indeed, the controller would instruct A to use {1011} 900 for FA1 while it would have to instruct B to use {2011} for FB1 901 (while with the same SRGB, both policies are the same {16011}). 903 Even worse, the controller would instruct A to use {1005, 5011} for 904 FA1 while it would instruct B to use {2011, 8011} for FB1 (while with 905 the same SRGB, the second segment is the same across both policies: 906 16011). When combining segments to create a policy, one need to 907 carefully update the label of each segment. This is obviously more 908 error-prone, more complex and more difficult to troubleshoot. 910 10. IANA Considerations 912 This document does not make any IANA request. 914 11. Manageability Considerations 916 The design and deployment guidelines described in this document are 917 based on the network design described in [RFC7938]. 919 The deployment model assumed in this document is based on a single 920 domain where the interconnected DCs are part of the same 921 administrative domain (which, of course, is split into different 922 autonomous systems). The operator has full control of the whole 923 domain and the usual operational and management mechanisms and 924 procedures are used in order to prevent any information related to 925 internal prefixes and topology to be leaked outside the domain. 927 As recommended in [I-D.ietf-spring-segment-routing], the same SRGB 928 SHOULD be allocated in all nodes in order to facilitate the design, 929 deployment and operations of the domain. 931 When EPE ([I-D.ietf-spring-segment-routing-central-epe]) is used (as 932 explained in Section 8.3, the same operational model is assumed. EPE 933 information is originated and propagated throughout the domain 934 towards an internal server and unless explicitly configured by the 935 operator, no EPE information is leaked outside the domain boundaries. 937 12. Security Considerations 939 This document proposes to apply Segment Routing to a well known 940 scalability requirement expressed in [RFC7938] using the BGP-Prefix- 941 SID as defined in [I-D.ietf-idr-bgp-prefix-sid]. 943 It has to be noted, as described in Section 11 that the design 944 illustrated in [RFC7938] and in this document, refer to a deployment 945 model where all nodes are under the same administration. In this 946 context, we assume that the operator doesn't want to leak outside of 947 the domain any information related to internal prefixes and topology. 948 The internal information includes prefix-sid and EPE information. In 949 order to prevent such leaking, the standard BGP mechanisms (filters) 950 are applied on the boundary of the domain. 952 Therefore, the solution proposed in this document does not introduce 953 any additional security concerns from what expressed in [RFC7938] and 954 [I-D.ietf-idr-bgp-prefix-sid]. It is assumed that the security and 955 confidentiality of the prefix and topology information is preserved 956 by outbound filters at each peering point of the domain as described 957 in Section 11. 959 13. Acknowledgements 961 The authors would like to thank Benjamin Black, Arjun Sreekantiah, 962 Keyur Patel, Acee Lindem and Anoop Ghanwani for their comments and 963 review of this document. 965 14. Contributors 967 Gaya Nagarajan 968 Facebook 969 US 971 Email: gaya@fb.com 972 Dmitry Afanasiev 973 Yandex 974 RU 976 Email: fl0w@yandex-team.ru 978 Tim Laberge 979 Cisco 980 US 982 Email: tlaberge@cisco.com 984 Edet Nkposong 985 Salesforce.com Inc. 986 US 988 Email: enkposong@salesforce.com 990 Mohan Nanduri 991 Microsoft 992 US 994 Email: mnanduri@microsoft.com 996 James Uttaro 997 ATT 998 US 1000 Email: ju1738@att.com 1002 Saikat Ray 1003 Unaffiliated 1004 US 1006 Email: raysaikat@gmail.com 1008 15. References 1010 15.1. Normative References 1012 [I-D.ietf-idr-bgp-prefix-sid] 1013 Previdi, S., Filsfils, C., Lindem, A., Patel, K., 1014 Sreekantiah, A., Ray, S., and H. Gredler, "Segment Routing 1015 Prefix SID extensions for BGP", draft-ietf-idr-bgp-prefix- 1016 sid-04 (work in progress), December 2016. 1018 [I-D.ietf-spring-segment-routing] 1019 Filsfils, C., Previdi, S., Decraene, B., Litkowski, S., 1020 and R. Shakir, "Segment Routing Architecture", draft-ietf- 1021 spring-segment-routing-11 (work in progress), February 1022 2017. 1024 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 1025 Requirement Levels", BCP 14, RFC 2119, 1026 DOI 10.17487/RFC2119, March 1997, 1027 . 1029 [RFC3107] Rekhter, Y. and E. Rosen, "Carrying Label Information in 1030 BGP-4", RFC 3107, DOI 10.17487/RFC3107, May 2001, 1031 . 1033 [RFC4271] Rekhter, Y., Ed., Li, T., Ed., and S. Hares, Ed., "A 1034 Border Gateway Protocol 4 (BGP-4)", RFC 4271, 1035 DOI 10.17487/RFC4271, January 2006, 1036 . 1038 [RFC7311] Mohapatra, P., Fernando, R., Rosen, E., and J. Uttaro, 1039 "The Accumulated IGP Metric Attribute for BGP", RFC 7311, 1040 DOI 10.17487/RFC7311, August 2014, 1041 . 1043 15.2. Informative References 1045 [GREENBERG09] 1046 Greenberg, A., Hamilton, J., Jain, N., Kadula, S., Kim, 1047 C., Lahiri, P., Maltz, D., Patel, P., and S. Sengupta, 1048 "VL2: A Scalable and Flexible Data Center Network", 2009. 1050 [I-D.ietf-6man-segment-routing-header] 1051 Previdi, S., Filsfils, C., Field, B., Leung, I., Linkova, 1052 J., Aries, E., Kosugi, T., Vyncke, E., and D. Lebrun, 1053 "IPv6 Segment Routing Header (SRH)", draft-ietf-6man- 1054 segment-routing-header-05 (work in progress), February 1055 2017. 1057 [I-D.ietf-mpls-seamless-mpls] 1058 Leymann, N., Decraene, B., Filsfils, C., Konstantynowicz, 1059 M., and D. Steinberg, "Seamless MPLS Architecture", draft- 1060 ietf-mpls-seamless-mpls-07 (work in progress), June 2014. 1062 [I-D.ietf-spring-segment-routing-central-epe] 1063 Filsfils, C., Previdi, S., Aries, E., and D. Afanasiev, 1064 "Segment Routing Centralized BGP Egress Peer Engineering", 1065 draft-ietf-spring-segment-routing-central-epe-04 (work in 1066 progress), February 2017. 1068 [KANDULA04] 1069 Sinha, S., Kandula, S., and D. Katabi, "Harnessing TCP's 1070 Burstiness with Flowlet Switching", 2004. 1072 [RFC6793] Vohra, Q. and E. Chen, "BGP Support for Four-Octet 1073 Autonomous System (AS) Number Space", RFC 6793, 1074 DOI 10.17487/RFC6793, December 2012, 1075 . 1077 [RFC7938] Lapukhov, P., Premji, A., and J. Mitchell, Ed., "Use of 1078 BGP for Routing in Large-Scale Data Centers", RFC 7938, 1079 DOI 10.17487/RFC7938, August 2016, 1080 . 1082 Authors' Addresses 1084 Clarence Filsfils (editor) 1085 Cisco Systems, Inc. 1086 Brussels 1087 BE 1089 Email: cfilsfil@cisco.com 1091 Stefano Previdi (editor) 1092 Cisco Systems, Inc. 1093 Via Del Serafico, 200 1094 Rome 00142 1095 Italy 1097 Email: sprevidi@cisco.com 1099 Jon Mitchell 1100 Unaffiliated 1102 Email: jrmitche@puck.nether.net 1103 Ebben Aries 1104 Juniper Networks 1105 1133 Innovation Way 1106 Sunnyvale CA 94089 1107 US 1109 Email: exa@juniper.net 1111 Petr Lapukhov 1112 Facebook 1113 US 1115 Email: petr@fb.com