idnits 2.17.1 draft-ietf-idr-bgp-optimal-route-reflection-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 : ---------------------------------------------------------------------------- ** The abstract seems to contain references ([RFC4456]), which it shouldn't. Please replace those with straight textual mentions of the documents in question. Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year == Line 84 has weird spacing: '...routing from...' == Line 95 has weird spacing: '... potato routi...' == The document seems to lack the recommended RFC 2119 boilerplate, even if it appears to use RFC 2119 keywords. (The document does seem to have the reference to RFC 2119 which the ID-Checklist requires). -- The document date (January 2, 2014) is 3766 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) == Unused Reference: 'RFC2119' is defined on line 929, but no explicit reference was found in the text == Unused Reference: 'RFC4271' is defined on line 932, but no explicit reference was found in the text == Unused Reference: 'RFC4360' is defined on line 935, but no explicit reference was found in the text == Unused Reference: 'RFC5492' is defined on line 938, but no explicit reference was found in the text == Unused Reference: 'RFC1997' is defined on line 948, but no explicit reference was found in the text == Unused Reference: 'RFC1998' is defined on line 951, but no explicit reference was found in the text == Unused Reference: 'RFC4384' is defined on line 955, but no explicit reference was found in the text == Unused Reference: 'RFC4893' is defined on line 962, but no explicit reference was found in the text == Unused Reference: 'RFC5668' is defined on line 969, but no explicit reference was found in the text == Outdated reference: A later version (-15) exists of draft-ietf-idr-add-paths-09 -- Obsolete informational reference (is this intentional?): RFC 4893 (Obsoleted by RFC 6793) Summary: 1 error (**), 0 flaws (~~), 14 warnings (==), 2 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 IDR Working Group R. Raszuk 3 Internet-Draft NTT I3 4 Intended status: Standards Track C. Cassar 5 Expires: July 6, 2014 Cisco Systems 6 E. Aman 7 TeliaSonera 8 B. Decraene 9 S. Litkowski 10 Orange 11 January 2, 2014 13 BGP Optimal Route Reflection (BGP-ORR) 14 draft-ietf-idr-bgp-optimal-route-reflection-06 16 Abstract 18 [RFC4456] asserts that, because the Interior Gateway Protocol (IGP) 19 cost to a given point in the network will vary across routers, "the 20 route reflection approach may not yield the same route selection 21 result as that of the full IBGP mesh approach." One practical 22 implication of this assertion is that the deployment of route 23 reflection may thwart the ability to achieve hot potato routing. Hot 24 potato routing attempts to direct traffic to the closest AS egress 25 point in cases where no higher priority policy dictates otherwise. 26 As a consequence of the route reflection method, the choice of exit 27 point for a route reflector and its clients will be the egress point 28 closest to the route reflector - and not necessarily closest to the 29 RR clients. 31 Section 11 of [RFC4456] describes a deployment approach and a set of 32 constraints which, if satsified, would result in the deployment of 33 route reflection yielding the same results as the iBGP full mesh 34 approach. Such a deployment approach would make route reflection 35 compatible with the application of hot potato routing policy. 37 As networks evolved to accommodate architectural requirements of new 38 services, tunneled (LSP/IP tunneling) networks with centralized route 39 reflectors became commonplace. This is one type of common deployment 40 where it would be impractical to satisfy the constraints described in 41 Section 11 of [RFC4456]. Yet, in such an environment, hot potato 42 routing policy remains desirable. 44 This document proposes two new solutions which can be deployed to 45 facilitate the application of closest exit point policy centralized 46 route reflection deployments. 48 Status of This Memo 50 This Internet-Draft is submitted in full conformance with the 51 provisions of BCP 78 and BCP 79. 53 Internet-Drafts are working documents of the Internet Engineering 54 Task Force (IETF). Note that other groups may also distribute 55 working documents as Internet-Drafts. The list of current Internet- 56 Drafts is at http://datatracker.ietf.org/drafts/current/. 58 Internet-Drafts are draft documents valid for a maximum of six months 59 and may be updated, replaced, or obsoleted by other documents at any 60 time. It is inappropriate to use Internet-Drafts as reference 61 material or to cite them other than as "work in progress." 63 This Internet-Draft will expire on July 6, 2014. 65 Copyright Notice 67 Copyright (c) 2014 IETF Trust and the persons identified as the 68 document authors. All rights reserved. 70 This document is subject to BCP 78 and the IETF Trust's Legal 71 Provisions Relating to IETF Documents 72 (http://trustee.ietf.org/license-info) in effect on the date of 73 publication of this document. Please review these documents 74 carefully, as they describe your rights and restrictions with respect 75 to this document. Code Components extracted from this document must 76 include Simplified BSD License text as described in Section 4.e of 77 the Trust Legal Provisions and are provided without warranty as 78 described in the Simplified BSD License. 80 Table of Contents 82 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 3 83 2. Proposed solutions . . . . . . . . . . . . . . . . . . . . . 5 84 3. Best path selection for BGP hot potato routing from 85 customized IGP network position . . . . . . . . . . . . . . . 6 86 3.1. Client's perspective best path selection algorithm . . . 7 87 3.1.1. Flat IGP network . . . . . . . . . . . . . . . . . . 7 88 3.1.2. Hierarchical IGP network . . . . . . . . . . . . . . 8 89 3.2. Aside: Configuration-based flexible route reflector 90 placement . . . . . . . . . . . . . . . . . . . . . . . . 9 91 3.3. Route reflector client grouping . . . . . . . . . . . . . 10 92 3.3.1. Route Reflector Client Group ID . . . . . . . . . . . 10 93 3.4. Discussion . . . . . . . . . . . . . . . . . . . . . . . 11 94 3.5. Advantages . . . . . . . . . . . . . . . . . . . . . . . 12 95 4. Angular distance approximation for BGP warm potato routing . 13 96 4.1. Problem statement . . . . . . . . . . . . . . . . . . . . 13 97 4.2. Proposed solution . . . . . . . . . . . . . . . . . . . . 14 98 4.3. Centralized vs distributed route reflectors . . . . . . . 15 99 5. Client's perspective policy based best path selection . . . . 16 100 5.1. Proposal . . . . . . . . . . . . . . . . . . . . . . . . 17 101 5.2. Example . . . . . . . . . . . . . . . . . . . . . . . . . 17 102 5.3. Avoiding routing loops . . . . . . . . . . . . . . . . . 18 103 6. Deployment considerations . . . . . . . . . . . . . . . . . . 19 104 7. Security considerations . . . . . . . . . . . . . . . . . . . 20 105 8. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 20 106 9. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . 20 107 10. References . . . . . . . . . . . . . . . . . . . . . . . . . 20 108 10.1. Normative References . . . . . . . . . . . . . . . . . . 20 109 10.2. Informative References . . . . . . . . . . . . . . . . . 21 110 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 21 112 1. Introduction 114 There are three types of BGP deployments within Autonomous Systems 115 today: full mesh, confederations and route reflection. 117 BGP route reflection is the most popular way to distribute BGP routes 118 between BGP speakers belonging to the same administrative domain. 119 Traditionally route reflectors have been deployed in the forwarding 120 path and carefully placed on the POP to core boundaries. That model 121 of BGP route reflector placement has started to evolve. The 122 placement of route reflectors outside the forwarding path was 123 triggered by applications which required traffic to be tunneled from 124 AS ingress PE to egress PE: for example L3VPN. 126 This evolving model of intra-domain network design has enabled 127 deployments of centralized route reflectors. Initially this model 128 was only employed for new address families e.g. L3VPNs, L2VPNs etc 130 With edge to edge MPLS or IP encapsulation also being used to carry 131 internet traffic, this model has been gradually extended to other BGP 132 address families including IPv4 and IPv6 Internet routing. This is 133 also applicable to new services achieved with BGP as control plane 134 for example 6PE. 136 Such centralized route reflectors can be placed on the POP to core 137 boundaries, but they are often placed in arbitrary locations in the 138 core of large networks. 140 Such deployments suffer from a critical drawback in the context of 141 best path selection. A route reflector with knowledge of multiple 142 paths for a given prefix will pick the best path and only advertise 143 that best path to the the route reflector clients. If the best path 144 for a prefix is selected on the basis of an IGP tie break, the best 145 path advertised from the route reflector to its clients will be the 146 exit point closest to the route reflector. But route reflector 147 clients will be in a place in the network toplogy which is different 148 from the route reflector. In networks with centralized route 149 reflectors, this difference will be even more acute. It follows that 150 the best path chosen by the route reflector is not necessarily the 151 same as the path which would have been chosen by the client if the 152 client considered the same set of candidate paths as the route 153 reflector. Furthermore, the path chosen by the client might have 154 been a better path from that chosen by the route reflector for 155 traffic entering the network at the client. The path chosen by the 156 client would have guaranteed the lowest cost and delay trajectory 157 through the network. 159 Route reflector clients switch packets using routing information 160 learnt from route reflectors which are not on the forwarding path of 161 the packet through the network even in the absence of end-to-end 162 encapsulation. In those cases the path chosen as best and propagated 163 to the clients will often not be the optimal path chosen by the 164 client given all available paths. 166 Eliminating the IGP distance to the BGP nexthop as a tie breaker on 167 centralized route reflectors does not address the issue. Ignoring 168 IGP distance to the BGP next hop results in the tie breaking 169 procedure contributing the best path by differentiating between paths 170 using attributes otherwise considered less important than IGP cost to 171 the BGP nexthop. 173 One possible valid solution or workaround to this problem requires 174 sending all domain external paths from the RR to all its clients. 175 This approach suffers the significant drawback of pushing a large 176 amount of BGP state to all the edge routers. In many networks, the 177 number of EBGP peers over which full Internet routing information is 178 received would correlate directly to the number of paths present in 179 each ASBR. This could easily result in tens of paths for each 180 prefix. 182 Notwithstanding this drawback, there are a number of reasons for 183 sending more than just the single best path to the clients. Improved 184 path diversity at the edge is a requirement for fast connectivity 185 restoration, and a requirement for effective BGP level load 186 balancing. Protocol extensions like add-paths 187 [I-D.ietf-idr-add-paths] or [RFC6774] diverse-path allow for such 188 improved path diversity and can be used to address the same problems 189 addressed by the mechanisms proposed in this draft. 191 In practical terms, add/diverse path deployments are expected to 192 result in the distribution of 2, 3 or n (where n is a small number) 193 'good' paths rather than all domain external paths. While the route 194 reflector chooses one set of n paths and distributes those same n 195 paths to all its route reflector clients, those n paths may not be 196 the right n paths for all clients. In the context of the problem 197 described above, those n paths will not necessarily include the 198 closest egress point out of the network for each route reflector 199 client. The mechanisms proposed in this document are likely to be 200 complementary to mechanisms aimed at improving path diversity. 202 2. Proposed solutions 204 This document proposes two simple solutions to the problem described 205 above. Both of these solutions make it possible for route reflector 206 clients to direct traffic to their closest exit point in hot potato 207 routing deployments, without requiring further state to be pushed out 208 to the edge. These solutions are primarily applicable in deployments 209 using centralized route reflectors, which are typically implemented 210 in devices without a capable forwarding plane. 212 The two alternatives are: 214 "Best path selection for BGP hot potato routing from client's IGP 215 network position" 217 "Angular distance approximation for BGP warm potato routing" 219 Both solutions rely upon all route reflectors learning all paths 220 which are eligible for consideration for hot potato routing. In 221 order to satisfy this requirement, path diversity enhancing 222 mechanisms such as add paths/diverse paths may need to be deployed 223 between route reflectors. 225 In both of these solutions the route reflector selects and 226 distributes a route to each client based on what would be optimal 227 from the client's perspective. By optimal we refer in this document 228 to the decision made during best path selection at the IGP metric to 229 BGP next hop comparison step. Clearly the overall path selection 230 preference may be chosen based other policy step and provisions as 231 defined in this document would not apply. 233 In the respective solutions the choice is made either factoring in 234 IGP costs or the configured angular distance to the next hop. The 235 route reflector makes different decisions for different clients only 236 in the case where the tie breaker for path selection would have been 237 the IGP distance to the BGP nexthop (as in hot potato routing). 239 A significant advantage of this approach is that the RR clients do 240 not need to run new software or hardware. 242 Besides these solutions to manage hot potato routing, there are 243 deployment scenarios where service providers want to have more 244 control of traffic exiting the AS by assigning per client preference 245 to gateways. 247 This document proposes to introduce a solution to perform a policy 248 based route-reflection to address those scenarios. This solution has 249 the same requirements (regarding path diversity) and advantages than 250 the two IGP metric based solutions. 252 3. Best path selection for BGP hot potato routing from customized IGP 253 network position 255 This section describes a method for calculating the order of 256 preference of BGP paths from the point of view of each separate route 257 reflector client. More specifically, the route reflector will 258 compute the IGP metric to the BGP nexthop from the position of the 259 client to which the resulting path will be distributed, if the IGP 260 metric is the tie breaker applied to a set of possible paths. In the 261 subsequent model authors will propose virtual reflector placement at 262 operator's selected IGP location. 264 In the case of a hierarchical IGP deployment where the client is in a 265 different level in the hierarchy to the route reflector, the route 266 reflector will compute IGP distance to the BGP nexthop from the Area 267 Border Routers (ABR) leading to the client in lieu of the route 268 reflector client itself, and use the shortest distance from these 269 ABRs to the nexthop. This provides an approximation to the desired 270 functionality. Rather than a client picking the closest path, the 271 client would be picking the exit point closest to the client region 272 as defined by area or level. In cases where one or more nexthops are 273 in the same region as the client, one of those nexthops would be 274 preferred, with tie breaking within those nexthops performed from the 275 route reflector's position in the network. 277 It is assumed that reachability through a set of ABRs is always 278 advertised through identical prefixes from those ABRs. If a nexthop 279 is reachable through multiple ABRs but the ABRs advertise 280 reachability through prefixes of different length, then only the ABR 281 advertising the longest prefix will be considered as a viable path to 282 the nexthop. 284 BGP best path selection and its distribution has a natural 285 consequence of limiting the amount of state in the network. That is 286 not in itself a drawback. BGP speakers will rarely need to receive 287 all available BGP paths. In network deployments with multiple 288 upstream peerings or with very dense peering schemes, the number of 289 available BGP paths for a given BGP prefix can be high. Real network 290 deployments with the number of paths for a prefix ranging from 10s to 291 100s have been observed. It would be wasteful to propagate all of 292 those paths to all clients, such that each client can select paths 293 according to the position of the nexthop relative to the client. 295 Whenever a BGP route reflector would need to decide what path or 296 paths need to be selected for advertisement to one of its clients, 297 the route reflector would need to virtually position itself in its 298 client IGP network location in order to choose the right set of paths 299 based on the IGP metric to the next hops from the client's 300 perspective. 302 This technique applies in deployments with or without diverse paths 303 or the various path selection modes contemplated in add-paths. 305 In the network architectures consisting of more then single pair of 306 route reflectors it is required that all reflectors are fully meshed 307 and have ability to learn and maintain all external BGP paths. In 308 the event of constructing a hierarchy of reflectors to relax the full 309 RR mesh requirements ORR should not be run between such route 310 reflectors. 312 3.1. Client's perspective best path selection algorithm 314 For each centralized route reflector the proposal assumes that the 315 route reflector participates in a common IGP with its clients. There 316 are two scenarios to consider - flat versus hierarchical IGP network. 318 3.1.1. Flat IGP network 320 Reflectors run SPF from the client IGP node point of view such 321 that the cost of BGP nexthops from the client can be determined if 322 necessary. For the purpose of BGP path selection the interesting 323 product of this calculation is the ability to determine the IGP 324 distance from a client to a BGP next hop. This distance to a 325 nexthop would be interesting in cases where that next hop is for a 326 path which is contending with otherwise equally preferred paths. 327 This approach works in tunneled as well as conventional hop-by-hop 328 IP forwarding cores. 330 When the path selection tie breaker for a prefix is the IGP metric 331 to the BGP nexthops of the contending paths, then the route 332 reflector will determine the order of preference of the contending 333 paths by considering the distance from the client to the path 334 nexthops in order to decide what path/s to advertise to a client 335 (or group of clients where feasible). It should be noted that an 336 operator may wish to provide a distance tolerance value, such that 337 beyond a certain granularity, differences between IGP metric are 338 invisible to the path selection algorithm. This will allow a 339 route reflector some leeway in selecting between paths such that 340 rather than pick one path over another on the basis of a 341 difference in distance which is operationally irrelevant, the 342 route reflector can choose to optimize for update generation 343 grouping. Furthermore, this tolerance will reduce the likelihood 344 of generation of BGP updates when the IGP topology changes in a 345 way which is not operationally relevant. In the case that a path 346 is selected from a set for a given prefix while ignoring 347 differences in distance within the tolerance figure, then that 348 same path must always be preferred for all clients where the paths 349 are within the tolerance figure 351 3.1.2. Hierarchical IGP network 353 Hierarchy introduces two challenges: 355 The first challenge is that the RR IGP view may differ from a 356 client IGP view by virtue of one or the other having a summarized 357 view versus the other. Summarization, by its nature, loses 358 information. Consider the example where a client within a PoP 359 sees two prefixes with two metrics for two egress points within 360 the PoP, but where the RR only sees a single summary covering 361 reachability to both nexthops as injected by the ABR. For 362 clarification purposes in the case of ISIS by ABR we refer to L1/ 363 L2 node. However it needs to be observed that inter area networks 364 running LDP are required to disable summarisation of all FEC 365 advertised in LDP (typically all loopbacks) unless [RFC5283] is 366 deployed. Such deployments are not likely to suffer summarization 367 difficulties. 369 The second challenge is that in cases where the client is in a 370 different level of hierarchy from the RR, the RR can not build a 371 Shortest Path First (SPF) tree with the client node as root, 372 simply because the topology derived by the IGP will not include 373 the client node. It will instead only include reachability to the 374 client from one or more ABRs. In order to overcome this problem, 375 the RR could compute an SPF tree from the ABRs in the area. The 376 RR would then determine the shortest distance from a client which 377 lives behind the ABRs, to a nexthop, by adding the advertised 378 distances from an ABR to the client and the distance from the ABR 379 to a nexthop, for each ABR, and picking the minimum. This assumes 380 that IGP metrics on links are symmetric; i.e. that the distance 381 from the ABR to the client or nexthop is equal to the distance 382 from the client or nexthop to the ABR. 384 There are cases where the above approach does not help. If RR is 385 trying to arbitrate amongst a set of paths for a client which is 386 in the same hierarchy as some of those paths, and in a different 387 hierarchy to the RR, the opaqueness of the region containing the 388 client at the RR defeats the selection process. It is impossible 389 to determine the relative position of the RR client and the paths 390 within the client region. 392 The solution for hierarchical IGP networks also assumes that if 393 RRs are present and are responsible for calculation of BGP best 394 path to clients they are either placed in each local area 395 coinciding with area containing clients or they are placed in the 396 core (area 0/level 2) of the network. 398 3.2. Aside: Configuration-based flexible route reflector placement 400 The ability to exploit topology information available in the IGP in 401 ways described above can also be used to virtually place the RR at 402 different points in the network for purposes other than hot potato 403 routing. 405 A route reflector can be globally configured to "pretend" its logical 406 location is one of any of the other nodes within a given IGP area/ 407 level flooding scope regardless of its physical connectivity. 409 Such flexibility provides a useful tool for reflector virtualization, 410 and supports moving or replacing physical route reflectors without 411 any effect on routing. Such a change can be permanent or it could be 412 performed during network maintenance in order to minimize network 413 impact. 415 A possible variation would allow the virtual placement of RR to be 416 effected on a per-AF or AF plus update/peer group granularity. It 417 should be noted that this approach provides for splitting one 418 centralized route reflector such that it is virtually positioned at 419 various network locations, with the network location depending upon 420 of address family or address family plus update/peer group. 422 Virtual slicing of a centralized route reflector relaxes the need to 423 propagate all BGP paths between RRs in a alternative conventional 424 distributed RR deployment. It is expected that such RRs would be 425 deployed in redundant sets, and that those RRs would not need to be 426 physically collocated, while still benefiting from the possibility of 427 being logically collocated, and therefore not compromising any of the 428 best path selection symmetry. 430 3.3. Route reflector client grouping 432 It may be appropriate to allow the operator, or the route reflector 433 itself, to group clients together using IGP distance between clients 434 to determine grouping. All the operation discussed above which 435 relied upon computing best path for each client, and measuring 436 distances from each client to different nexthops, would instead be 437 performed for each group of clients. Configurable thresholds can be 438 used to determine which IGP metric changes should be visible to BGP, 439 and trigger best paths recomputation. The latter would be beneficial 440 in existing BGP RR code too. 442 Alternatively route reflector client grouping could be accomplished 443 statically by the operator by coloring clients belonging to a common 444 group (for example being part of the same POP). In order to 445 accomplish such marking it is proposed that BGP OPEN message be 446 augmented with an optional parameter indicating the Group ID given 447 peer belongs to. 449 3.3.1. Route Reflector Client Group ID 451 This is an Optional Parameter in BGP OPEN message that is used by a 452 BGP speaker to convey to its route reflectors the Group ID value. 453 Such value will allow automatic and predictable peer grouping on the 454 route reflectors as deemed necessary from operator's network 455 architecture. 457 The parameter contains precisely one set of [Group_ID Code, Group_ID 458 Length, Group_ID Value] encoded as shown below: 460 +----------------------------+ 461 | Group ID Code (1 octet) | 462 +----------------------------+ 463 | Group ID Length (1 octet) | 464 +----------------------------+ 465 | Group ID Value (4 octets) | 466 +----------------------------+ 468 The use and meaning of these fields are as follows: 470 Group ID Code: 472 Group ID Code is a one octet field that identifies Group ID 473 optional parameter of BGP OPEN message. Value TBD by IANA 474 Recommended value: 3. 476 Group ID Length: 478 Group ID Length is a one octet field that contains the length 479 of the Group ID Value field in octets. It is fixed and equals 480 to 4. 482 Group ID Value: 484 Group ID Value is a fixed length field of size equal to 485 four octets that contains the numerical value of group given 486 BGP speaker should be part of on the route reflector. 488 Two special values are reserved: 490 0x00000000 - No grouping preference 491 0xFFFFFFFF - Do not group this BGP speaker 493 An implementation may allow automatic population of 494 GROUP_ID value using IGP area identifier. 496 Route reflectors or EBGP speakers receiving such Group IDs from their 497 respective BGP peers as part of the BGP OPEN procedure MAY use them 498 when constructing update or peer groups in addition to any of the 499 existing grouping mechanism already available. An implementation may 500 allow operator to explicitly allow or disallow honoring such grouping 501 or provide means for manual overwrite via explicit configuration. 503 3.4. Discussion 505 This is not the first instance where a router participating in an IGP 506 is required to build the SPF tree using a root other than itself. 507 Determination of loop free alternate paths as described in [RFC5714] 508 is one such example. 510 Determining the shortest path and associated cost between any two 511 arbitrary points in a network based on the IGP topology learned by a 512 router is expected to add some extra cost in terms of CPU resource. 513 However SPF tree generation code is now implemented efficiently in a 514 number of implementations, and therefor this is not expected to be a 515 major drawback. The number of SPTs computed in the general non- 516 hierarchical case is expected to be of the order of the number of 517 clients of an RR whenever a topology change is detected. Advanced 518 optimizations like partial and incremental SPF may also be exploited. 519 By the nature of route reflection, the number of clients can be split 520 arbitrarily by the deployment of more route reflectors for a given 521 number of clients. While this is not expected to be necessary in 522 existing networks with best in class route reflectors available 523 today, this avenue to scaling up the route reflection infrastructure 524 would be available. If we consider the overall network wide cost/ 525 benefit factor, the only alternative to achieve the same level of 526 optimality would require significantly increasing state on the edges 527 of the network, which, in turn, will consume CPU and memory resources 528 on all BGP speakers in the network. Building this client perspective 529 into the route reflectors seems appropriate. 531 3.5. Advantages 533 The solution described provides a model for integrating the client 534 perspective into the best path computation for RRs. More 535 specifically, the choice or BGP path factors in the IGP metric 536 between the client and the nexthop, rather than the distance from the 537 RR to the nexthop. The documented method does not require any BGP or 538 IGP protocol changes as required changes are contained within the RR 539 implementation. 541 This solution can be deployed in traditional hop-by-hop forwarding 542 networks as well as in end-to-end tunneled environments. In the 543 networks where there are multiple route reflectors and hop-by-hop 544 forwarding without encapsulation, such optimizations should be 545 enabled on all route reflectors. Otherwise clients may receive an 546 inconsistent view of the network and in turn lead to intra-domain 547 forwarding loops. 549 With this approach, an ISP can effect a hot potato routing policy 550 even if route reflection has been moved from the forwarding plane to 551 the core and hop-by-hop switching has been replaced by end to end 552 MPLS or IP encapsulation. 554 As per above, the approach reduces the amount of state which needs to 555 be pushed to the edge in order to perform hot potato routing. The 556 memory and CPU resource required at the edge to provide hot potato 557 routing using this approach is lower than what would be required in 558 order to achieve the same level of optimality by pushing and 559 retaining all available paths (potentially 10s) per each prefix at 560 the edge. 562 The proposal allows for a fast and safe transition to BGP control 563 plane route reflection without compromising an operator's closest 564 exit operational principle. Hot potato routing is important to most 565 ISPs. The inability to perform hot potato routing effectively stops 566 migrations to centralized route reflection and edge-to-edge LSP/IP 567 encapsulation for traffic to IPv4 and IPv6 prefixes. 569 4. Angular distance approximation for BGP warm potato routing 571 This section describes an alternative solution to the use of IGP 572 topology information to virtually position the RR at the client 573 location in the network. This solution involves modeling the network 574 topology as a set of elements (regions, PoPs or routers) arranged in 575 a circle. Route reflector clients and inter-domain exit points would 576 then be statically assigned to those elements such that one can 577 compute the angular distance between route-reflector clients and the 578 various exit points in order to infer the distance between any two 579 elements. This measure of distance can be used as an effective 580 alternative to the IGP distance as a tie breaker in the path 581 selection algorithm if necessary. 583 4.1. Problem statement 585 This solution addresses the problem described in earlier sections, 586 while attempting to minimize computational overhead. The aim of the 587 proposed solution is to enable a route reflector to provide a route 588 reflector client with an exit point for a prefix which is 'closest' 589 to the client rather than the route-reflector, without having to 590 distribute all paths to that client, or having to derive each 591 client's view of the network topology. The measure of closest is 592 based on a simplistic description of network topology provided by the 593 operator. 595 Consider the following example of an ISP network topology drawn to 596 reflect the location of the nodes and POPs: 598 N4 POP4 600 CLIENT B 601 POP4 POP1 N1 603 CORE 604 RR(s) POP2 N2 606 N5 POP3 POP2 N3 608 CLIENT A 609 POP3 611 N - represents the different exit points for a given prefix. POP2 is 612 a geographically large PoP with two paths; N2 and N3. 614 In a deployment where the centralized RRs tie break on the basis of 615 their IGP-based view of the network, N1 above would be advertised to 616 all clients on the basis that it is closest to the RR. Path N4 would 617 be a more appropriate choice for client B. Similarly, N5 would be 618 more appropriate for client A since path N5 is closer to client A 619 then path N1. 621 4.2. Proposed solution 623 The proposed solution revolves around the operator establishing the 624 angular position of the route-reflector clients and inter-domain exit 625 points in the network. The route reflector then picks the path to 626 advertise to a client based on the client's angular position versus 627 the angular position of the inter-domain exit points originating the 628 paths. The operator can choose the granularity of angular position 629 appropriate to the desired goals. On one hand, the coarseness of the 630 angular position will effect the operator overhead; versus the 631 optimality of routing on the other. The finest granularity possible 632 will be the relative position of originating clients. 634 Note that this solution has nothing to do with actual IGP link 635 metrics and resulting topology in the network. 637 It can be shown that for each network topology, elements such as AS 638 exit points can be mapped on to a circle. By putting POPs, Regions 639 or individual clients onto the hypothetical circle we can identify an 640 angular location for each element relative to some fixed direction; 641 for example defining the angular north of the circle at 0 degrees. 643 The angular position of elements in the network can be conveyed to a 644 route reflector in a number of ways: 646 Assignment of angular position of each RR client through 647 configuration on the route reflector itself; per client 648 configuration on RR 650 Assignment of angular position of an RR client at each client, 651 then propagating it to RRs. 653 The proposed angular distance approximation is compatible with both 654 flat and hierarchical IGP deployments. 656 In the example illustrated above the route reflector might learn or 657 be configured with the following set of paths and corresponding 658 angular positions: 660 Prefix X/Y N1 N2 N3 N4 N5 662 Location 663 in degrees 60 85 120 290 260 665 If the absolute angular position of clients A and B were as follows: 667 Client A: 260 degrees 669 Client B: 290 degrees 671 Then the corresponding angular distances for those clients versus the 672 exit points can be calculated as follows: 674 Prefix X/Y N1 N2 N3 N4 N5 676 Client A 200 175 140 30 0 678 Client B 230 205 170 0 30 680 With an RR running the BGP best path algorithm modified to use the 681 angular distance from the client to the nexthops, rather than its IGP 682 distance to the nexthops as tie breaker, each client is provided with 683 its closest path with the measure of closeness reflecting the angular 684 position as configured by the operator. 686 The model used by the operator in order to determine the angular 687 position of a client or exit point, might involve grouping elements 688 together by region or PoP, or might involve no grouping at all. 689 Implementations should allow the operator to pick the appropriate 690 granularity. 692 4.3. Centralized vs distributed route reflectors 694 In an environment where the RR clusters are distributed (yet 695 centralized enough to make hot potato routing hard), and each RR 696 cluster serves a subset of clients, it becomes necessary to propagate 697 the angular position of the clients between route reflectors. This 698 can be achieved as follows: 700 Deploy add-paths between route reflectors in order to maximize 701 path diversity within the cluster. 703 A non AS transitive BGP community of type (TBA by IANA) can be 704 used to encode and propagate angular position between 0 and 359 of 705 a client. This community is only relevant to the route reflectors 706 of a given BGP domain and should be stripped either at the ASBR 707 boundary or when propagating updates to BGP peers which are not 708 route reflectors. 710 The angular position marking could also be added by clients and 711 advertised to the route reflector. This would require some 712 configuration effort. 714 5. Client's perspective policy based best path selection 716 There is some deployment scenarios where a service provider wants to 717 achieve a stronger control on traffic exiting the AS (for capacity 718 planning) rather than using hot potato routing based on IGP metric. 720 | | | | 721 | | | | 722 GW1 GW2 GW3 GW4 724 RR1 RR2 726 R1 R2 R3 728 Considering the figure above, all gateways have iBGP sessions to RR1 729 and RR2, and R1 R2 R3 have iBGP sessions as well to RR1 and RR2. 730 Gateway routers are meshed to an external network (for example, a 731 transit service provider). 733 We would like to achieve a strong control on the gateway used 734 (primary and backup) for each router (or each set of routers) in the 735 network (taking into account that routers do not support ADD PATHs). 736 For example, R1 using GW1 as primary and GW2 as backup; R2 using GW2 737 as primary and GW3 as backup; R3 using GW3 as primary and GW4 as 738 backup. 740 Basically, today a prefix P1 is received on each gateway from the 741 external network. Each gateway will send the prefix to both route 742 reflectors. Each route-reflector will receive four paths for P1 and 743 choose the best one based on his own decision process. Note that RR1 744 and RR2 may choose a different path as best. Each route-reflector 745 sends his best path towards R1, R2 and R3. Each router will receive 746 the same paths from the route-reflectors for P1 (at max, only two 747 gateways are visible from Rx routers). So default behavior does not 748 fit our requirements in term of traffic flows. 750 Using current BGP mechanisms available, we could achieve our 751 requirements using two solutions : 753 o Modify the BGP meshing: for example, R1 meshed directly to GW1 and 754 GW2 and apply inbound policies on R1; R2 meshed directly to GW2 755 and GW3 and apply inbound policies on R2 ... 757 o Adding more route-reflectors (one RR per gateway used as primary) 758 and applying inbound policies on RRs to make each RR choosing a 759 different primary gateway and apply policies on routers to select 760 his own primary gateway. 762 These solutions have many drawbacks: first one is not flexible (re- 763 meshing needed when we want to change gateway of a router), second 764 one requires a lot of CAPEX. 766 We would like to introduce a solution where a single currently 767 deployed route-reflector chassis may take a different best path 768 decision for different set of clients based on preferences. 770 It should be noted that in simple scenarios (example: two RRs and two 771 gateways), RFC6774 would be able to fulfill service provider needs. 772 The solution proposed here would permit to handle more complex 773 scenarios and fine gateway choice per client or groups of clients. 775 5.1. Proposal 777 Our proposal is to reuse the concept introduced in [I.D.ietf-idr-ix- 778 bgp-route-server] in an iBGP context. To perform per client best 779 path selection, the router should maintain a per client BGP local-RIB 780 (or Adj-RIB-Out) associated with inbound policies implemented between 781 Adj-RIB-In and client LOC-RIB. 783 It would not be very scalable to use a per client policy (considering 784 hundreds of peers on a route-reflector), therefor our proposal is to 785 group clients sharing common policies inside a client group to 786 minimize computation/memory overhead. Client grouping could be done 787 statically (by configuration) or dynamically using the solution 788 described in section 3.3.1 of this document. Client grouping would 789 be performed with a per AFI/SAFI granularity as gateway/client 790 mapping may change in each AFI/SAFI context. A route-reflector 791 should be able to implement multiple client groups (with associated 792 inbound policies) as well as a default client group for clients that 793 does not require any specific policy decision: in this case, the 794 overall BGP best path computation would be used. 796 5.2. Example 797 GW1 GW2 GW3 798 \ | / 799 \ | / 800 RR1 801 / | \ 802 R1 R2 R3 804 In the above figure GW1, GW2, GW3 and R3 are standard ibgp route- 805 reflector clients. R1 and R2 want to use a special gateway 806 combination (primary GW3, backup GW2, last resort GW1). R1 and R2 807 are configured in a specific client group CG1 on the route-reflector 808 while other peers are in the default client group. CG1 is associated 809 with a policy achieving the expected GW preference for R1 and R2, and 810 letting other paths without any change. 812 All routes received by RR1 (ebgp, ibgp, ibgp rr client, ibgp rr 813 client routing context) must be evaluated using overall BGP best path 814 computation as well as in client group, the client group policy will 815 accept or not the route to be evaluated by the local decision 816 process. 818 o Paths from GW1, GW2, GW3 are compared within default client group 819 leading to one GW (for example GW1) to be selected as best and 820 installed in global LOC-RIB. GW1 path will be advertised to GW2, 821 GW3 and R3 as they are in default CG. In CG1, preference of GW 822 paths has been modified, leading to GW3 being the best path and 823 installed in client group LOC-RIB. GW3 path will be advertised to 824 R1 and R2, as R1 and R2 are part of CG1. 826 o Paths from R3 are compared within default client group and 827 advertised to GW1, GW2, GW3. Those paths are also compared within 828 CG1 (as accepted by policy) and advertised to R1 and R2. 830 o Paths from R1 are compared within default client group and 831 advertised to GW1, GW2, GW3 and R3. Those paths are also compared 832 within GG1 (as accepted by policy) and advertised to R2. 834 o Paths from R2 are compared within default client group and 835 advertised to GW1, GW2, GW3 and R3. Those paths are also compared 836 within CG1 (as accepted by policy) and advertised to R1. 838 5.3. Avoiding routing loops 840 Compared to the IGP approaches described in this document, the policy 841 based route-reflection should be limited to end-to-end encapsulation 842 environments to avoid intra-domain forwarding loops. Using end-to- 843 end encapsulation permit Edge routers to transport the traffic to the 844 targeted/preferred ASBR without any loop in the core. 846 To avoid a potential rerouting of the ASBR into the core (and 847 possible loop between Edges and ASBR), we must enforce forwarding at 848 the ASBR to the eBGP peer. This could be done by : 850 o implementing policies on ASBR to prefer eBGP path and install it 851 in FIB. 853 o implementing tunneling of traffic until the outside interface 854 (ASBR action to switch to outside interface). 856 The exact choice of encapsulation and techniques to prevent transport 857 loops (including potential loops at gateways) is left to the operator 858 choice and its specification is outside of the scope of this 859 document. 861 6. Deployment considerations 863 The solutions are primarily intended for end-to-end tunneled 864 environments, i.e. where traffic is label switched or IP tunneled 865 across the core. If unencapsulated hop-by-hop forwarding is used, 866 either misconfigurations or conflicts between these optimizations and 867 classical BGP path selection rules could lead to intra-domain 868 forwarding loops. Under certain circumstances the solutions can also 869 be deployable without end-to-end tunneling. In particular the best 870 path selection based on the client's IGP best-path selection is 871 guaranteed not to cause any forwarding loops (other than micro loops 872 associated with reconvergence) when deployed in a flat IGP area 873 provided that no distance tolerance value is used so that the path 874 choice is truly made on a per-client basis. 876 Regarding potential intra-domain forwarding loops at ASBR level, this 877 could be solved by enforcing external route preference or by 878 performing tunnel to external interface switching action on ASBRs. 880 Regarding client's IGP best-path selection, it should be self evident 881 that this solution does not interfere with policies enforced above 882 IGP tie breaking in the BGP best path algorithm. 884 The solution applies to NLRIs of all address families which can be 885 route reflected. 887 It should be noted that customized per-client or group of clients 888 best path selection is already in use today in the context of 889 Internet Exchange Point (IXP) route servers. In an IXP route server 890 the client best path is selected as a result of different policies 891 rather than IGP metric distance to BGP next hop. 893 A possible scalability impact of optimizing path selection to take 894 account of the RR client position or operator's policy based 895 preference is that different RR clients receive different paths, and 896 therefore update/peer group efficiency diminishes. This cost is 897 imposed by the requirement to optimize the egress path from the 898 client's perspective. It is also likely that groups of clients will 899 end up receiving the same best path/s, in which case, inefficiency of 900 update generation will be minimized. It should be noted that in the 901 cases described under flexible router placement where placement is 902 determined on a per update/peer group basis or per route reflector, 903 the scale benefits of peer groupings are retained. 905 7. Security considerations 907 No new security issues are introduced to the BGP protocol by this 908 specification. 910 8. IANA Considerations 912 IANA is requested to allocate a type code for the Standard BGP 913 Community to be used for inter cluster propagation of angular 914 position of the clients. 916 IANA is requested to allocate a new type code from BGP OPEN Optional 917 Parameter Types registry to be used for Group_ID propagation. 919 9. Acknowledgments 921 Authors would like to thank Eric Rosen, Clarence Filsfils, Uli 922 Bornhauser Russ White, Jakob Heitz and Mike Shand for their valuable 923 input. 925 10. References 927 10.1. Normative References 929 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 930 Requirement Levels", BCP 14, RFC 2119, March 1997. 932 [RFC4271] Rekhter, Y., Li, T., and S. Hares, "A Border Gateway 933 Protocol 4 (BGP-4)", RFC 4271, January 2006. 935 [RFC4360] Sangli, S., Tappan, D., and Y. Rekhter, "BGP Extended 936 Communities Attribute", RFC 4360, February 2006. 938 [RFC5492] Scudder, J. and R. Chandra, "Capabilities Advertisement 939 with BGP-4", RFC 5492, February 2009. 941 10.2. Informative References 943 [I-D.ietf-idr-add-paths] 944 Walton, D., Retana, A., Chen, E., and J. Scudder, 945 "Advertisement of Multiple Paths in BGP", draft-ietf-idr- 946 add-paths-09 (work in progress), October 2013. 948 [RFC1997] Chandrasekeran, R., Traina, P., and T. Li, "BGP 949 Communities Attribute", RFC 1997, August 1996. 951 [RFC1998] Chen, E. and T. Bates, "An Application of the BGP 952 Community Attribute in Multi-home Routing", RFC 1998, 953 August 1996. 955 [RFC4384] Meyer, D., "BGP Communities for Data Collection", BCP 114, 956 RFC 4384, February 2006. 958 [RFC4456] Bates, T., Chen, E., and R. Chandra, "BGP Route 959 Reflection: An Alternative to Full Mesh Internal BGP 960 (IBGP)", RFC 4456, April 2006. 962 [RFC4893] Vohra, Q. and E. Chen, "BGP Support for Four-octet AS 963 Number Space", RFC 4893, May 2007. 965 [RFC5283] Decraene, B., Le Roux, JL., and I. Minei, "LDP Extension 966 for Inter-Area Label Switched Paths (LSPs)", RFC 5283, 967 July 2008. 969 [RFC5668] Rekhter, Y., Sangli, S., and D. Tappan, "4-Octet AS 970 Specific BGP Extended Community", RFC 5668, October 2009. 972 [RFC5714] Shand, M. and S. Bryant, "IP Fast Reroute Framework", RFC 973 5714, January 2010. 975 [RFC6774] Raszuk, R., Fernando, R., Patel, K., McPherson, D., and K. 976 Kumaki, "Distribution of Diverse BGP Paths", RFC 6774, 977 November 2012. 979 Authors' Addresses 980 Robert Raszuk 981 NTT I3 982 101 S Ellsworth Avenue Suite 350 983 San Mateo, CA 94401 984 US 986 Email: robert@raszuk.net 988 Christian Cassar 989 Cisco Systems 990 10 New Square Park 991 Bedfont Lakes, FELTHAM TW14 8HA 992 UK 994 Email: ccassar@cisco.com 996 Erik Aman 997 TeliaSonera 998 Marbackagatan 11 999 Farsta SE-123 86 1000 Sweden 1002 Email: erik.aman@teliasonera.com 1004 Bruno Decraene 1005 Orange 1006 38-40 rue du General Leclerc 1007 Issy les Moulineaux cedex 9 92794 1008 France 1010 Email: bruno.decraene@orange.com 1012 Stephane Litkowski 1013 Orange 1014 9 rue du chene germain 1015 Cesson Sevigne 35512 1016 France 1018 Email: stephane.litkowski@orange.com