idnits 2.17.1 draft-przygienda-lsr-flood-reflection-01.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 : ---------------------------------------------------------------------------- ** The document seems to lack an Introduction section. Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year -- The document date (January 7, 2020) is 1564 days in the past. Is this intentional? Checking references for intended status: Proposed Standard ---------------------------------------------------------------------------- (See RFCs 3967 and 4897 for information about using normative references to lower-maturity documents in RFCs) No issues found here. Summary: 1 error (**), 0 flaws (~~), 1 warning (==), 1 comment (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 Network Working Group A. Przygienda 3 Internet-Draft C. Bowers 4 Intended status: Standards Track Juniper 5 Expires: July 10, 2020 Y. Lee 6 A. Sharma 7 Comcast 8 R. White 9 Juniper 10 January 7, 2020 12 IS-IS Flood Reflection 13 draft-przygienda-lsr-flood-reflection-01 15 Abstract 17 This document describes an optional ISIS extension that allows the 18 creation of IS-IS flood reflection topologies. Flood reflection 19 allows the creation of topologies where L1 areas provide transit 20 forwarding for L2 destinations within an L2 topology. It 21 accomplishes this by creating L2 flood reflection adjacencies within 22 each L1 area. The L2 flood reflection adjacencies are used to flood 23 L2 LSPDUs, and they are used in the L2 SPF computation. However, 24 they are not used for forwarding. This arrangement gives the L2 25 topology better scaling properties. In addition, only those routers 26 directly participating in flood reflection have to support the 27 feature. This allows for the incremental deployment of scalable L1 28 transit areas in an existing network, without the necessity of 29 upgrading other routers in the network. 31 Requirements Language 33 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 34 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 35 document are to be interpreted as described in RFC 2119 [RFC2119]. 37 Status of This Memo 39 This Internet-Draft is submitted in full conformance with the 40 provisions of BCP 78 and BCP 79. 42 Internet-Drafts are working documents of the Internet Engineering 43 Task Force (IETF). Note that other groups may also distribute 44 working documents as Internet-Drafts. The list of current Internet- 45 Drafts is at https://datatracker.ietf.org/drafts/current/. 47 Internet-Drafts are draft documents valid for a maximum of six months 48 and may be updated, replaced, or obsoleted by other documents at any 49 time. It is inappropriate to use Internet-Drafts as reference 50 material or to cite them other than as "work in progress." 52 This Internet-Draft will expire on July 10, 2020. 54 Copyright Notice 56 Copyright (c) 2020 IETF Trust and the persons identified as the 57 document authors. All rights reserved. 59 This document is subject to BCP 78 and the IETF Trust's Legal 60 Provisions Relating to IETF Documents 61 (https://trustee.ietf.org/license-info) in effect on the date of 62 publication of this document. Please review these documents 63 carefully, as they describe your rights and restrictions with respect 64 to this document. Code Components extracted from this document must 65 include Simplified BSD License text as described in Section 4.e of 66 the Trust Legal Provisions and are provided without warranty as 67 described in the Simplified BSD License. 69 Table of Contents 71 1. Description . . . . . . . . . . . . . . . . . . . . . . . . . 2 72 2. Further Details . . . . . . . . . . . . . . . . . . . . . . . 8 73 3. Flood Reflection TLV . . . . . . . . . . . . . . . . . . . . 8 74 4. Flood Reflection Discovery Sub-TLV . . . . . . . . . . . . . 10 75 5. Flood Reflection Adjacency Sub-TLV . . . . . . . . . . . . . 10 76 6. Flood Reflection Discovery . . . . . . . . . . . . . . . . . 11 77 7. Flood Reflection Adjacency Formation . . . . . . . . . . . . 12 78 8. Redistribution of Prefixes . . . . . . . . . . . . . . . . . 12 79 9. Route Computation . . . . . . . . . . . . . . . . . . . . . . 13 80 10. Special Considerations . . . . . . . . . . . . . . . . . . . 13 81 11. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 14 82 11.1. New IS-IS TLV Codepoint . . . . . . . . . . . . . . . . 14 83 11.2. Sub TLVs for TLV 242 . . . . . . . . . . . . . . . . . . 14 84 11.3. Sub TLVs for TLV 22, 23, 25, 141, 222, and 223 . . . . . 15 85 12. Security Considerations . . . . . . . . . . . . . . . . . . . 15 86 13. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 15 87 14. References . . . . . . . . . . . . . . . . . . . . . . . . . 15 88 14.1. Informative References . . . . . . . . . . . . . . . . . 15 89 14.2. Normative References . . . . . . . . . . . . . . . . . . 15 90 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 16 92 1. Description 94 Due to the inherent properties of link-state protocols the number of 95 IS-IS routers within a flooding domain is limited by processing and 96 flooding overhead on each node. While that number can be maximized 97 by well written implementations and techniques such as exponential 98 back-offs, IS-IS will still reach a saturation point where no further 99 routers can be added to a single flooding domain. In some L2 100 backbone deployment scenarios, this limit presents a significant 101 challenge. 103 The traditional approach to increasing the scale of an IS-IS 104 deployement is to break it up into multiple L1 flooding domains and a 105 single L2 backbone. This works well for designs where an L2 backbone 106 connects L1 access topologies, but it is limiting where a large L2 is 107 supposed to span large number of routers. In such scenarios, an 108 alternative approach is to consider multiple L2 flooding domains 109 connected together via L1 flooding domains. In other words, L2 110 flooding domains are connected by "L1/L2 lanes" through the L1 areas 111 to form a single L2 backbone again. Unfortunately, in its simplest 112 implementation, this requires the inclusion of most, or all, of the 113 transit L1 routers as L1/L2 to allow traffic to flow along optimal 114 paths through such transit areas. Consequently, this approach fails 115 to reduce the number of L2 routers involved, so it fails to increase 116 the scalability of the L2 backbone. 118 +----+ +-------+ +-------+ +-------+ +----+ 119 | R1 | | R10 +------------+ R20 +---------------+ R30 | | R4 | 120 | L2 +--+ L1/L2 | | L1 | | L1/L2 +--+ L2 | 121 | | | +--------+ +-+ | +------------+ | | | 122 +----+ ++-+--+-+ | | +---+---+----------+ +-+--+-++ +----+ 123 | | | | | | | | | | | | | 124 | | | | | | | | | +-----------+ | | 125 | | +-------+ | | | | | | | | | | 126 | | | | | | | | | | | +------+ | 127 | +------+ +--------+ | +-------+ | | | 128 | | | | | | | | | | | | | 129 +----+ ++------+---+ | +---+---+---+--+ | +-------+------++ +----+ 130 | R2 | | R11 | | | | | R21 | | | | | R31 | | R5 | 131 | L2 +--+ L1/L2 +------------+ L1 +---------------+ L1/L2 +--+ L2 | 132 | | | | | | | | | | | | | | | | 133 +----+ ++------+---+ | | +---+--++ | +-------+------++ +----+ 134 | | | | | | | | | | | | | 135 | +---------------+ | | | | | | | | 136 | | | | | | | | | | | | | 137 | | +--------------+ | +-----------------+ | 138 | | | | | | | | | | | | | 139 +----+ ++-+--+-+ | | +------+---+---+-----+ | | | ++-----++ +----+ 140 | R3 | | R12 | +----------| R22 | | +----+ R32 | | R6 | 141 | L2 +--+ L1/L2 | +--------| L1 +-------+ | | L1/L2 +--+ L2 | 142 | | | +------------+ |---------------+ | | | 143 +----+ +-------+ +-------+-------------+ +-------+ +----+ 145 Figure 1: Example topology 147 Figure 1 is an example of a network where a topologically rich L1 148 area is used to provide transit between six different L2-only routers 149 (R1-R6). Note that the six L2-only routers do not have connectivity 150 to one another over L2 links. To take advantage of the abundance of 151 paths in the L1 transit area, all the intermediate systems could be 152 placed into both L1 and L2, but this essentially combines the 153 separate L2 flooding domains into a single one, triggering again 154 maximum L2 scale limitation we try to address in first place. 156 A more effective solution would allow to reduce the number of links 157 and routers exposed in L2, while still utilizing the full L1 topology 158 when forwarding through the network. 160 [RFC8099] describes Topology Transparent Zones (TTZ) for OSPF. The 161 TTZ mechanism represents a group of OSPF routers as a full mesh of 162 adjacencies between the routers at the edge of the group. A similar 163 mechanism could be applied to ISIS as well. However, a full mesh of 164 adjacencies between edge routers (or L1/L2 nodes) significantly 165 limits the scale of the topology. The topology in Figure 1 has 6 L1/ 166 L2 nodes. Figure 2 illustrates a full mesh of L2 adjacencies between 167 the 6 L1/L2 nodes, resulting in (5 * 6)/2 = 15 L2 adjacencies. In a 168 somewhat larger topology containing 20 L1/L2 nodes, the number of L2 169 adjacencies in a full mesh rises to 190. 171 +----+ +-------+ +-------------------------------+-------+ +----+ 172 | R1 | | R10 | | | R30 | | R4 | 173 | L2 +--+ L1/L2 +------------------------------------+ L1/L2 +--+ L2 | 174 | | | | | | | | | 175 +----+ ++-+-+--+-+ | +-+--+---++ +----+ 176 | | | | | | | | 177 | +----------------------------------------------+ | 178 | | | | | | | | 179 | +-----------------------------------+ | | | | 180 | | | | | | | | 181 | +----------------------------------------+ | | 182 | | | | | | | | 183 +----+ ++-----+- | | | | -----+-++ +----+ 184 | R2 | | R11 | | | | | | R31 | | R5 | 185 | L2 +--+ L1/L2 +------------------------------------+ L1/L2 +--+ L2 | 186 | | | | | | | | | | | | 187 +----+ ++------+------------------------------+ | | +----+-++ +----+ 188 | | | | | | | | 189 | | | | | | | | 190 | +-------------------------------------------+ | 191 | | | | | | | | 192 | | | | +----------+ | 193 | | | | | | | | 194 | | | | +-----+ | | 195 | | | | | | | | 196 +----+ ++----+-+-+ | +-+-+--+-++ +----+ 197 | R3 | | R12 | | L2 adjacency | R32 | | R6 | 198 | L2 +--+ L1/L2 +------------------------------------+ L1/L2 +--+ L2 | 199 | | | | | | | | | 200 +----+ +-------+----+ +-------+ +----+ 202 Figure 2: Example topology represented in L2 with a full mesh of L2 203 adjacencies between L1/L2 nodes 205 BGP, as specified in [RFC4271], faced a similar scaling problem, 206 which has been solved in many networks by deploying BGP route 207 reflectors [RFC4456]. We note that BGP route reflectors do not 208 necessarily have to be in the forwarding path of the traffic. This 209 incongruity of forwarding and control path for BGP route reflectors 210 allows the control plane to scale independently of the forwarding 211 plane. 213 We propose here a similar solution for IS-IS. A simple example of 214 what a flood reflector control plane approach would look like is 215 shown in Figure 3, where router R21 plays the role of a flood 216 reflector. Each L1/L2 ingress/egress router builds a tunnel to the 217 flood reflector, and an L2 adjacency is built over each tunnel. In 218 this solution, we need only 6 L2 adjacencies, instead of the 15 219 needed for a full mesh. In a somewhat larger topology containing 20 220 L1/L2 nodes, this solution requires only 20 L2 adjacencies, instead 221 of the 190 need for a full mesh. Multiple flood reflectors can be 222 used, allowing the network operator to balance between resilience, 223 path utilization, and state in the control plane. The resulting L2 224 adjacency scale is R*n, where R is the number of flood reflectors 225 used and n is the number of L1/L2 nodes. This compares quite 226 favorably with n*(n-1)/2 L2 adjacencies required in a fully meshed L2 227 solution. 229 +----+ +-------+ +-------+ +----+ 230 | R1 | | R10 | | R30 | | R4 | 231 | L2 +--+ L1/L2 +--------------+ +-----------------+ L1/L2 +--+ L2 | 232 | | | | L2 adj | | L2 adj | | | | 233 +----+ +-------+ over | | over +-------+ +----+ 234 tunnel | | tunnel 235 +----+ +-------+ +--+---+--+ +-------+ +----+ 236 | R2 | | R11 | | R21 | | R31 | | R5 | 237 | L2 +--+ L1/L2 +-----------+ L1/L2 +--------------+ L1/L2 +--+ L2 | 238 | | | | L2 adj | flood | L2 adj | | | | 239 +----+ +-------+ over |reflector| over +-------+ +----+ 240 tunnel +--+---+--+ tunnel 241 +----+ +-------+ | | +-------+ +----+ 242 | R3 | | R12 +--------------+ +-----------------+ R32 | | R6 | 243 | L2 +--+ L1/L2 | L2 adj L2 adj | L1/L2 +--+ L2 | 244 | | | | over over | | | | 245 +----+ +-------+ tunnel tunnel +-------+ +----+ 247 Figure 3: Example topology represented in L2 with L2 adjacencies from 248 each L1/L2 node to a single flood reflector 250 As illustrated in Figure 3, when R21 plays the role of flood 251 reflector, it provides L2 connectivity among all of the previously 252 disconnected L2 islands by reflooding all L2 LSPDUs. At the same 253 time, R20 and R22 remain L1-only routers. L1-only routers and 254 L1-only links are not visible in L2. In this manner, the flood 255 reflector allows us provide L2 control plane connectivity in a 256 scalable manner. 258 As described so far, the solution illustrated in Figure 3 relies only 259 on currently standardized ISIS functionality. Without new 260 functionality, however, the data traffic will traverse only R21. 261 This will unnecessarily create a bottleneck at R21 since there is 262 still available capacity in the paths crossing the L1-only routers 263 R20 and R22. 265 Hence, some new functionality is necessary to allow the L1/L2 edge 266 nodes (R10-12 and R30-32 in Figure 3) to recognize that the L2 267 adjacency to R21 should not be used for forwarding. The L1/L2 edge 268 nodes should forward traffic that would normally be forwarded over 269 the L2 adjacency to R21 over L1 links instead. This would allow the 270 forwarding within the L1 area to use the L1-only nodes and links 271 shown in Figure 1 as well. It allows networks to be built that use 272 the entire forwarding capacity of the L1 areas, while at the same 273 time introducing control plane scaling benefits provided by L2 flood 274 reflectors. 276 This document defines all extensions necessary to support flood 277 reflector deployment: 279 o A 'flood reflector adjacency' for all the adjacencies built for 280 the purpose of reflecting flooding information. This allows these 281 'flood reflectors' to participate in the IS-IS control plane 282 without being used in the forwarding plane. This is a purely 283 local operation on the L1/L2 ingress; it does not require 284 replacing or modifying any routers not involved in the reflection 285 process. Deployment-wise, it is far less tricky to just upgrade 286 the routers involved in flood reflection rather than have a flag 287 day on the whole ISIS domain. 289 o A full mesh of L1 tunnels between the L1/L2 routers, ideally load- 290 balancing across all available L1 links. This harnesses all 291 forwarding paths between the L1/L2 edge nodes without injecting 292 unneeded state into the L2 flooding domain or creating 'choke 293 points' at the 'flood reflectors' themselves. A solution without 294 tunnels is also possible by judicious scoping of reachability 295 information between the levels. 297 o Some way to support reflector redundancy, and potentially some way 298 to auto-discover and advertise such adjacencies as flood reflector 299 adjacencies. Such advertisements may allow L2 nodes outside the 300 L1 to perform optimizations in the future based on this 301 information. 303 2. Further Details 305 Several considerations should be noted in relation to such a flood 306 reflection mechanism. 308 First, this allows multi-area IS-IS deployments to scale without any 309 major modifications in the IS-IS implementation on most of the nodes 310 deployed in the network. Unmodified (traditional) L2 routers will 311 compute reachability across the transit L1 area using the flood 312 reflector adjacencies. 314 Second, the flood reflectors are not required to participate in 315 forwarding traffic through the L1 transit area. These flood 316 reflectors can be hosted on virtual devices outside the forwarding 317 topology. 319 Third, astute readers will realize that flooding reflection may cause 320 the use of suboptimal paths. This is similar to the BGP route 321 reflection suboptimal routing problem described in 322 [ID.draft-ietf-idr-bgp-optimal-route-reflection-19]. The L2 323 computation determines the egress L1/L2 and with that can create 324 illusions of ECMP where there is none. And in certain scenarios lead 325 to an L1/L2 egress which is not globally optimal. This represents a 326 straightforward instance of the trade-off between the amount of 327 control plane state and the optimal use of paths through the network 328 often encountered when aggregating routing information. 330 One possible solution to this problem is to expose additional 331 topology information into the L2 flooding domains. In the example 332 network given, links from router 01 to router 02 can be exposed into 333 L2 even when 01 and 02 are participating in flood reflection. This 334 information would allow the L2 nodes to build 'shortcuts' when the L2 335 flood reflected part of the topology looks more expensive to cross 336 distance wise. 338 Another possible variation is for an implementation to approximate 339 with the L1 tunnel cost the cost of the underlying topology. 341 Redundancy can be achieved by building multiple flood reflectors in 342 the L1 area. Multiple flood reflectors do not need any 343 synchronization mechanisms amongst themselves, except standard ISIS 344 flooding and database maintenance procedures. 346 3. Flood Reflection TLV 348 The Flood Reflection TLV is a new top-level TLV that MAY appear in 349 IIHs. The Flood Reflection TLV indicates the flood reflector cluster 350 (based on Flood Reflection Cluster ID) that a given router is 351 configured to participate in. It also indicates whether the router 352 is configured to play the role of either flood reflector or flood 353 reflector client. The Flood Reflection Cluster ID and flood 354 reflector roles advertised in the IIHs are used to ensure that flood 355 reflector adjacencies are only formed between a flood reflector and 356 flood reflector client, and that the Flood Reflection Cluster IDs 357 match. The Flood Reflection TLV has the following format: 359 0 1 2 3 360 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 1 361 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 362 | Type | Length |C| RESERVED | 363 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 364 | Flood Reflection Cluster ID | 365 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 366 | Sub-TLVs ... 367 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 369 Type: TBD 371 Length: The length, in octets, of the following fields. 373 C (Client): This bit is set to indicate that the router acts as a 374 flood reflector client. When this bit is NOT set, the router acts 375 as a flood reflector. On a given router, the same value of the 376 C-bit MUST be advertised across all interfaces advertising the 377 Flood Reflection TLV in IIHs. 379 RESERVED: This field is reserved for future use. It MUST be set to 380 0 when sent and MUST be ignored when received. 382 Flood Reflection Cluster ID: Flood Reflection Cluster Identifier. 383 These same 32-bit value MUST be assigned to all of the flood 384 reflectors and flood reflector clients in the L1 area. The value 385 MUST be unique across different L1 areas within the IGP domain. 386 On a given router, the same value of the Flood Reflection Cluster 387 ID MUST be advertised across all interfaces advertising the Flood 388 Reflection TLV in IIHs. 390 Sub-TLVs: Optional sub-TLVs. For future extensibility, the format 391 of the Flood Reflection TLV allows for the possibility of 392 including optional sub-TLVs. No sub-TLVs of the Flood Reflection 393 TLV are defined in this document. 395 The Flood Reflection TLV MUST NOT appear more than once in an IIH. A 396 router receiving multiple Flood Reflection TLVs in the same IIH 397 SHOULD use the values in the first TLV. 399 4. Flood Reflection Discovery Sub-TLV 401 Flood Reflection Discovery sub-TLV is advertised as a sub-TLV of the 402 IS-IS Router Capability TLV-242, defined in [RFC7981]. The Flood 403 Reflection Discovery sub-TLV is advertised in L1 LSPs with area 404 flooding scope in order to enable the auto-discovery of flood 405 reflection capabilities and the automatic creation of L2 tunnels to 406 be used as flood reflector adjacencies. The Flood Reflection 407 Discovery sub-TLV has the following format: 409 0 1 2 3 410 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 1 411 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 412 | Type | Length |C| Reserved | 413 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 414 | Flood Reflection Cluster ID | 415 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 417 Type: TBD 419 Length: The length, in octets, of the following fields. 421 C (Client): This bit is set to indicate that the router acts as a 422 flood reflector client. When this bit is NOT set, the router acts 423 as a flood reflector. 425 RESERVED: This field is reserved for future use. It MUST be set to 426 0 when sent and MUST be ignored when received. 428 Flood Reflection Cluster ID: The Flood Reflection Cluster Identifier 429 is the same as that defined in the Flood Reflection TLV. 431 The Flood Reflection Discovery sub-TLV MUST NOT appear more than once 432 in TLV 242. A router receiving multiple Flood Reflection Discovery 433 sub-TLVs in TLV 242 SHOULD use the values in the first sub-TLV. 435 5. Flood Reflection Adjacency Sub-TLV 437 The Flood Reflection Adjacency sub-TLV is advertised as a sub-TLV of 438 TLVs 22, 23, 25, 141, 222, and 223. Its presence indicates that a 439 given adjacency is a flood reflector adjacency. It is included in L2 440 area scope flooded LSPs. Flood Reflection Adjacency sub-TLV has the 441 following format: 443 0 1 2 3 444 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 1 445 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 446 | Type | Length |C| Reserved | 447 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 448 | Flood Reflection Cluster ID | 449 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 451 Type: TBD 453 Length: The length, in octets, of the following fields. 455 C (Client): This bit is set to indicate that the router advertising 456 this adjacency is a flood reflector client. When this bit is NOT 457 set, the router advertising this adjacency is a flood reflector. 459 RESERVED: This field is reserved for future use. It MUST be set to 460 0 when sent and MUST be ignored when received. 462 Flood Reflection Cluster ID: The Flood Reflection Cluster Identifier 463 is the same as that defined in the Flood Reflection TLV. 465 The Flood Reflection Adjacency sub-TLV MUST NOT appear more than once 466 in a given TLV. A router receiving multiple Flood Reflection 467 Adjacency sub-TLVs in a TLV SHOULD use the values in the first sub- 468 TLV. 470 6. Flood Reflection Discovery 472 A router participating in flood reflection MUST be configured as an 473 L1/L2 router. It originates the Flood Reflection Discovery sub-TLV 474 with area flooding scope in L1 only. Normally, all routers on the 475 edge of the L1 area (those having traditional L2 adjacencies) will 476 advertise themselves as route reflector clients. Therefore, a flood 477 reflector client will have both traditional L2 adjacencies and flood 478 reflector L2 adjacencies. 480 A router acting as a flood reflector MUST NOT have any traditional L2 481 adjacencies. It will be an L1/L2 router only by virtue of having 482 flood reflector L2 adjacencies. A router desiring to act as a flood 483 reflector will advertise itself as such using the Flood Reflection 484 Discovery sub-TLV in L1. 486 A given flood reflector or flood reflector client can only 487 participate in a single cluster, as determined by the value of its 488 Flood Reflection Cluster ID. 490 Upon reception of Flood Reflection Discovery sub-TLVs, a router 491 acting as flood reflector client MUST initiate a tunnel towards each 492 flood reflector with which it shares an Flood Reflection Cluster ID. 493 The L2 adjacencies formed over such tunnels MUST be marked as flood 494 reflector adjacencies. If the client has a direct L2 adjacency with 495 the flood reflector it SHOULD use it instead of instantiating a new 496 tunnel. 498 Upon reception of Flood Reflection Discover TLVs, a router acting as 499 a flood reflector client MAY initiate tunnels with L1-only 500 adjacencies towards all the other flood reflector clients in its 501 cluster. These tunnels MAY be used for forwarding to improve the 502 load-balancing characteristics of the L1 area. 504 7. Flood Reflection Adjacency Formation 506 In order to simplify both implementations and network deployments, we 507 do not allow the formation of complex hierarchies of flood reflectors 508 and clients. All flood reflectors and flood reflector clients in the 509 same L1 area MUST share the same Flood Reflector Cluster ID. A flood 510 reflector MUST only form flood reflection adjacencies with flood 511 reflector clients. A flood reflector MUST NOT form any traditional 512 L2 adjacencies. Flood reflector clients MUST only form flood 513 reflection adjacencies with flood reflectors. Flood reflector 514 clients may form traditional L2 adjacencies with flood reflector 515 clients or nodes not participating in flood reflection. 517 The Flood Reflector Cluster ID and flood reflector roles advertised 518 in the Flood Reflection TLVs in IIHs are used to ensure that flood 519 reflection adjacencies that are established meet the above criteria. 521 Once a flood reflection adjacency is established, the flood reflector 522 and the flood reflector client MUST advertise the adjacency by 523 including the Flood Reflection Adjacency Sub-TLV in the Extended IS 524 reachability TLV or MT-ISN TLV. 526 8. Redistribution of Prefixes 528 In some scenarios, L2 prefixes need to be redistributed into L1 by 529 the route reflector clients. However, if a L1 area edge router 530 doesn't have any L2 flood reflector adjacencies, then it cannot be 531 the shortest path egress in the L2 topology. Therefore, flood 532 reflector client SHOULD only redistribute L2 prefixes into L1 if it 533 has an L2 flood reflector adjacency. The L2 prefixes advertisements 534 redistributed into L1 SHOULD be normally limited to L2 intra-area 535 routes (as defined in [RFC7775]), if the information exists to 536 distinguish them from other L2 prefix advertisements. 538 On the other hand, in topologies that make use of flood reflection to 539 hide the structure of L1 areas while still providing transit 540 forwarding across them, we generally do not need to redistribute L1 541 prefixes advertisements into L2. 543 In deployment scenarios where L1 tunnels are not used, all L1/L2 edge 544 nodes MUST be flood reflector clients. 546 9. Route Computation 548 To ensure loop-free routing, the route reflection client MUST follow 549 the normal L2 computation to determine L2 routes. This is because 550 nodes outside the L1 area will generally not be aware that flood 551 reflection is being performed. The flood reflection clients need to 552 produce the same result for the L2 route computation as a router not 553 participating in flood reflection. However, a flood reflector client 554 will not necessarily use a given L2 route for forwarding. For an L2 555 route that uses a flood reflection adjacency as a next-hop, the flood 556 reflection client may use the next-hop from an L1 route instead. 558 On the reflection client, after L2 and L1 computation, all flood 559 reflector adjacencies used as next-hops for L2 routes MUST be 560 examined and replaced with the correct L1 tunnel next-hop to the 561 egress. Alternatively, if the ingress has adequate reachability 562 information to ensure forwarding towards destination via L1 routes, 563 L2 routes using flood reflector adjacencies as next-hops can be 564 omitted entirely. Due to the rules in Section 7 the computation in 565 the resulting topology is relatively simple, the L2 SPF from a flood 566 reflector client is guaranteed to reach within a hop the Flood 567 Reflector and in the following hop the L2 egress to which it has a L1 568 forwarding tunnel. However, if the topology has L2 paths which are 569 not route reflected and look "shorter" than the path through the 570 Flood Reflector then the computation will have to track the egress 571 out of the L1 domain by a more advanced algorithm. 573 10. Special Considerations 575 In pathological cases setting the overload bit in L1 (but not in L2) 576 can partition L1 forwarding, while allowing L2 reachability through 577 flood reflector adjacencies to exist. In such a case a node cannot 578 replace a route through a flood reflector adjacency with a L1 579 shortcut and the client can use the L2 tunnel to the flood reflector 580 for forwarding while it MUST initiate an alarm and declare 581 misconfiguration. 583 A flood reflector with directly L2 attached prefixes should advertise 584 those in L1 as well since based on preference of L1 routes the 585 clients will not try to use the L2 flood reflector adjacency to route 586 the packet towards them. A very, very corner case is when the flood 587 reflector is reachable via L2 flood reflector adjacency (due to 588 underlying L1 partition) only in which case the client can use the L2 589 tunnel to the flood reflector for forwarding towards those prefixes 590 while it MUST initiate an alarm and declare misconfiguration. 592 Instead of modifying the computation procedures one could imagine a 593 flood reflector solution where the Flood Reflector would re-advertise 594 the L2 prefixes with a 'third-party' next-hop but that would have 595 less desirable convergence properties than the solution proposed and 596 force a fork-lift of all L2 routers to make sure they disregard such 597 prefixes unless in the same L1 domain as the Flood Reflector. 599 Depending on pseudo-node choice in case of a broadcast domain with 600 multiple flood reflectors attached this can lead to a partitioned LAN 601 and hence a router discovering such a condition MUST initiate an 602 alarm and declare misconfiguration. 604 11. IANA Considerations 606 This document requests allocation for the following IS-IS TLVs and 607 Sub-TLVs. 609 11.1. New IS-IS TLV Codepoint 611 This document requests the following IS-IS TLV: 613 Value Name IIH LSP SNP Purge 614 ----- --------------------------------- --- --- --- ----- 615 TBD1 Flood Reflection y n n n 617 11.2. Sub TLVs for TLV 242 619 This document request the following registration in the "sub-TLVs for 620 TLV 242" registry. 622 Type Description 623 ---- ----------- 624 TBD2 Flood Reflection Discovery 626 11.3. Sub TLVs for TLV 22, 23, 25, 141, 222, and 223 628 This document requests the following registration in the "sub-TLVs 629 for TLV 22, 23, 25, 141, 222, and 223" registry. 631 Type Description 22 23 25 141 222 223 632 ---- -------------------------------- --- --- --- --- --- --- 633 TBD3 Flood Reflector Adjacency y y y(s) y y y 635 12. Security Considerations 637 This document introduces no new security concerns to ISIS or other 638 specifications referenced in this document. 640 13. Acknowledgements 642 The authors thank Shraddha Hegde, Peter Psenak, and Les Ginsberg for 643 their thorough review and detailed discussions. 645 14. References 647 14.1. Informative References 649 [ID.draft-ietf-idr-bgp-optimal-route-reflection-19] 650 Raszuk et al., R., "BGP Optimal Route Reflection", July 651 2019. 653 [RFC4271] Rekhter, Y., Ed., Li, T., Ed., and S. Hares, Ed., "A 654 Border Gateway Protocol 4 (BGP-4)", RFC 4271, 655 DOI 10.17487/RFC4271, January 2006, 656 . 658 [RFC4456] Bates, T., Chen, E., and R. Chandra, "BGP Route 659 Reflection: An Alternative to Full Mesh Internal BGP 660 (IBGP)", RFC 4456, DOI 10.17487/RFC4456, April 2006, 661 . 663 [RFC8099] Chen, H., Li, R., Retana, A., Yang, Y., and Z. Liu, "OSPF 664 Topology-Transparent Zone", RFC 8099, 665 DOI 10.17487/RFC8099, February 2017, 666 . 668 14.2. Normative References 670 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 671 Requirement Levels", BCP 14, RFC 2119, 672 DOI 10.17487/RFC2119, March 1997, 673 . 675 [RFC7775] Ginsberg, L., Litkowski, S., and S. Previdi, "IS-IS Route 676 Preference for Extended IP and IPv6 Reachability", 677 RFC 7775, DOI 10.17487/RFC7775, February 2016, 678 . 680 [RFC7981] Ginsberg, L., Previdi, S., and M. Chen, "IS-IS Extensions 681 for Advertising Router Information", RFC 7981, 682 DOI 10.17487/RFC7981, October 2016, 683 . 685 Authors' Addresses 687 Tony Przygienda 688 Juniper 689 1137 Innovation Way 691 Sunnyvale, CA 693 USA 695 Email: prz@juniper.net 697 Chris Bowers 698 Juniper 699 1137 Innovation Way 701 Sunnyvale, CA 703 USA 705 Email: cbowers@juniper.net 707 Yiu Lee 708 Comcast 709 1800 Bishops Gate Blvd 710 Mount Laurel, NJ 08054 711 US 713 Email: Yiu_Lee@comcast.com 714 Alankar Sharma 715 Comcast 716 1800 Bishops Gate Blvd 717 Mount Laurel, NJ 08054 718 US 720 Email: Alankar_Sharma@comcast.com 722 Russ White 723 Juniper 724 1137 Innovation Way 726 Sunnyvale, CA 728 USA 730 Email: russw@juniper.net