idnits 2.17.1 draft-ietf-rtgwg-ipfrr-notvia-addresses-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 : ---------------------------------------------------------------------------- No issues found here. Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year == Line 513 has weird spacing: '... a Ps a ...' -- The document seems to lack a disclaimer for pre-RFC5378 work, but may have content which was first submitted before 10 November 2008. If you have contacted all the original authors and they are all willing to grant the BCP78 rights to the IETF Trust, then this is fine, and you can ignore this comment. If not, you may need to add the pre-RFC5378 disclaimer. (See the Legal Provisions document at https://trustee.ietf.org/license-info for more information.) -- The document date (October 21, 2010) is 4929 days in the past. Is this intentional? Checking references for intended status: Experimental ---------------------------------------------------------------------------- No issues found here. Summary: 0 errors (**), 0 flaws (~~), 2 warnings (==), 2 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 Network Working Group M. Shand 3 Internet-Draft S. Bryant 4 Intended status: Experimental S. Previdi 5 Expires: April 24, 2011 Cisco Systems 6 October 21, 2010 8 IP Fast Reroute Using Not-via Addresses 9 draft-ietf-rtgwg-ipfrr-notvia-addresses-06 11 Abstract 13 This draft describes a mechanism that provides fast reroute in an IP 14 network through encapsulation to "not-via" addresses. A single level 15 of encapsulation is used. The mechanism protects unicast, multicast 16 and LDP traffic against link, router and shared risk group failure, 17 regardless of network topology and metrics. 19 Requirements Language 21 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 22 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 23 document are to be interpreted as described in RFC2119 [RFC2119]. 25 Status of this Memo 27 This Internet-Draft is submitted in full conformance with the 28 provisions of BCP 78 and BCP 79. 30 Internet-Drafts are working documents of the Internet Engineering 31 Task Force (IETF). Note that other groups may also distribute 32 working documents as Internet-Drafts. The list of current Internet- 33 Drafts is at http://datatracker.ietf.org/drafts/current/. 35 Internet-Drafts are draft documents valid for a maximum of six months 36 and may be updated, replaced, or obsoleted by other documents at any 37 time. It is inappropriate to use Internet-Drafts as reference 38 material or to cite them other than as "work in progress." 40 This Internet-Draft will expire on April 24, 2011. 42 Copyright Notice 44 Copyright (c) 2010 IETF Trust and the persons identified as the 45 document authors. All rights reserved. 47 This document is subject to BCP 78 and the IETF Trust's Legal 48 Provisions Relating to IETF Documents 49 (http://trustee.ietf.org/license-info) in effect on the date of 50 publication of this document. Please review these documents 51 carefully, as they describe your rights and restrictions with respect 52 to this document. Code Components extracted from this document must 53 include Simplified BSD License text as described in Section 4.e of 54 the Trust Legal Provisions and are provided without warranty as 55 described in the Simplified BSD License. 57 Table of Contents 59 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 4 60 2. Overview of Not-via Repairs . . . . . . . . . . . . . . . . . 4 61 2.1. Use of Equal Cost Multi-Path . . . . . . . . . . . . . . . 6 62 2.2. Use of LFA repairs . . . . . . . . . . . . . . . . . . . . 6 63 3. Not-via Repair Path Computation . . . . . . . . . . . . . . . 6 64 3.1. Computing not-via repairs in routing vector protocols . . 7 65 4. Operation of Repairs . . . . . . . . . . . . . . . . . . . . . 8 66 4.1. Node Failure . . . . . . . . . . . . . . . . . . . . . . . 8 67 4.2. Link Failure . . . . . . . . . . . . . . . . . . . . . . . 8 68 4.2.1. Loop Prevention Under Node Failure . . . . . . . . . . 9 69 4.3. Multi-homed Prefixes . . . . . . . . . . . . . . . . . . . 9 70 4.4. Installation of Repair Paths . . . . . . . . . . . . . . . 10 71 5. Compound Failures . . . . . . . . . . . . . . . . . . . . . . 12 72 5.1. Shared Risk Link Groups . . . . . . . . . . . . . . . . . 12 73 5.1.1. Use of LFAs with SRLGs . . . . . . . . . . . . . . . . 16 74 5.2. Local Area Networks . . . . . . . . . . . . . . . . . . . 16 75 5.2.1. Simple LAN Repair . . . . . . . . . . . . . . . . . . 17 76 5.2.2. LAN Component Repair . . . . . . . . . . . . . . . . . 18 77 5.2.3. LAN Repair Using Diagnostics . . . . . . . . . . . . . 19 78 5.3. Multiple Independent Failures . . . . . . . . . . . . . . 19 79 5.3.1. Looping Repairs . . . . . . . . . . . . . . . . . . . 20 80 5.3.2. Outline Solution . . . . . . . . . . . . . . . . . . . 21 81 5.3.3. Looping Repairs . . . . . . . . . . . . . . . . . . . 22 82 5.3.3.1. Dropping Looping Packets . . . . . . . . . . . . . 22 83 5.3.3.2. Computing non-looping Repairs of Repairs . . . . . 23 84 5.3.3.3. N-level Mutual Loops . . . . . . . . . . . . . . . 25 85 5.3.4. Mixing LFAs and Not-via . . . . . . . . . . . . . . . 25 86 6. Optimizing not-via computations using LFAs . . . . . . . . . . 26 87 7. Multicast . . . . . . . . . . . . . . . . . . . . . . . . . . 27 88 8. Fast Reroute in an MPLS LDP Network. . . . . . . . . . . . . . 27 89 9. Encapsulation . . . . . . . . . . . . . . . . . . . . . . . . 28 90 10. Routing Extensions . . . . . . . . . . . . . . . . . . . . . . 28 91 11. Incremental Deployment . . . . . . . . . . . . . . . . . . . . 28 92 12. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 29 93 13. Security Considerations . . . . . . . . . . . . . . . . . . . 29 94 14. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . 29 95 15. References . . . . . . . . . . . . . . . . . . . . . . . . . . 29 96 15.1. Normative References . . . . . . . . . . . . . . . . . . . 29 97 15.2. Informative References . . . . . . . . . . . . . . . . . . 30 98 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . . 30 100 1. Introduction 102 When a link or a router fails, only the neighbors of the failure are 103 initially aware that the failure has occurred. In a network 104 operating IP fast reroute [RFC5714], the routers that are the 105 neighbors of the failure repair the failure. These repairing routers 106 have to steer packets to their destinations despite the fact that 107 most other routers in the network are unaware of the nature and 108 location of the failure. 110 A common limitation in most IPFRR mechanisms is an inability to 111 indicate the identity of the failure and to explicitly steer the 112 repaired packet round the failure. The extent to which this 113 limitation affects the repair coverage is topology dependent. The 114 mechanism proposed here is to encapsulate the packet to an address 115 that explicitly identifies the network component that the repair must 116 avoid. This produces a repair mechanism, which, provided the network 117 is not partitioned by the failure, will always achieve a repair. 119 2. Overview of Not-via Repairs 121 When a link or a router fails, only the neighbors of the failure are 122 initially aware that the failure has occurred. In a network 123 operating IP fast reroute [RFC5714], the routers that are the 124 neighbors of the failure repair the failure. These repairing routers 125 have to steer packets to their destinations despite the fact that 126 most other routers in the network are unaware of the nature and 127 location of the failure. 129 A common limitation in most IPFRR mechanisms is an inability to 130 indicate the identity of the failure and to explicitly steer the 131 repaired packet round the failure. The extent to which this 132 limitation affects the repair coverage is topology dependent. The 133 mechanism proposed here is to encapsulate the packet to an address 134 that explicitly identifies the network component that the repair must 135 avoid. This produces a repair mechanism, which, provided the network 136 is not partitioned by the failure, will always achieve a repair. 138 A 139 | Bp is the address to use to get 140 | a packet to B not-via P 141 | 142 S----------P----------B. . . . . . . . . .D 143 \ | Bp^ 144 \ | | 145 \ | | 146 \ C | 147 \ | 148 ----------------+ 149 Repair to Bp 151 Figure 1: Not-via repair of router failure 153 Assume that S has a packet for some destination D that it would 154 normally send via P and B, and that S suspects that P has failed. S 155 encapsulates the packet to Bp. The path from S to Bp is the shortest 156 path from S to B not going via P. If the network contains a path from 157 S to B that does not transit router P, i.e. the network is not 158 partitioned by the failure of P, then the packet will be successfully 159 delivered to B. When the packet addressed to Bp arrives at B, B 160 removes the encapsulation and forwards the repaired packet towards 161 its final destination. 163 Note that if the path from B to the final destination includes one or 164 more nodes that are included in the repair path, a packet may back 165 track after the encapsulation is removed. However, because the 166 decapsulating router is always closer to the packet destination than 167 the encapsulating router, the packet will not loop. 169 For complete protection, all of P's neighbors will require a not-via 170 address that allows traffic to be directed to them without traversing 171 P. This is shown in Figure 2. 173 A 174 |Ap 175 | 176 Sp Pa|Pb 177 S----------P----------B 178 Ps|Pc Bp 179 | 180 Cp| 181 C 183 Figure 2: The set of Not-via P Addresses 185 2.1. Use of Equal Cost Multi-Path 187 A router can use an equal cost multi-path (ECMP) repair in place of a 188 not-via repair. 190 A router computing a not-via repair path MAY subject the repair to 191 ECMP. 193 2.2. Use of LFA repairs 195 The not-via approach provides complete repair coverage and therefore 196 may be used as the sole repair mechanism. There are, however, 197 advantages in using not-via in combination with loop free alternates 198 (LFA) and or downstream paths as documented in [RFC5286]. 200 LFAs are computed on a per destination basis and in general, only a 201 subset of the destinations requiring repair will have a suitable LFA 202 repair. In this case, those destinations which are repairable by 203 LFAs are so repaired and the remainder of the destinations are 204 repaired using the not-via encapsulation. This has the advantage of 205 reducing the volume of traffic that requires encapsulation. On the 206 other hand, the path taken by an LFA repair may be less optimal than 207 that of the equivalent not-via repair for traffic destined to nodes 208 close to the far end of the failure, but may be more optimal for some 209 other traffic. The description in this document assumes that LFAs 210 will be used where available, but the distribution of repairs between 211 the two mechanisms is a local implementation choice. 213 3. Not-via Repair Path Computation 215 The not-via repair mechanism requires that all routers on the path 216 from S to B (Figure 1) have a route to Bp. They can calculate this 217 by failing node P, running an SPF, and finding the shortest route to 218 B. 220 A router has no simple way of knowing whether it is on the shortest 221 path for any particular repair. It is therefore necessary for every 222 router to calculate the path it would use in the event of any 223 possible router failure. Each router therefore "fails" every router 224 in the network, one at a time, and calculates its own best route to 225 each of the neighbors of that router. In other words, with reference 226 to Figure 1, some router X will consider each router in turn to be P, 227 fail P, and then calculate its own route to each of the not-via P 228 addresses advertised by the neighbors of P. i.e. X calculates its 229 route to Sp, Ap, Bp, and Cp, in each case, not via P. 231 To calculate the repair paths a router has to calculate n-1 SPFs 232 where n is the number of routers in the network. This is expensive 233 to compute. However, the problem is amenable to a solution in which 234 each router (X) proceeds as follows. X first calculates the base 235 topology with all routers functional and determines its normal path 236 to all not-via addresses. This can be performed as part of the 237 normal SPF computation. For each router P in the topology, X then 238 performs the following actions:- 240 1. Removes router P from the topology. 242 2. Performs an incremental SPF [ISPF] on the modified topology. The 243 iSPF process involves detaching the sub-tree affected by the 244 removal of router P, and then re-attaching the detached nodes. 245 However, it is not necessary to run the iSPF to completion. It 246 is sufficient to run the iSPF up to the point where all of the 247 nodes advertising not-via P addresses have been re-attached to 248 the SPT, and then terminate it. 250 3. Reverts to the base topology. 252 This algorithm is significantly less expensive than a set of full 253 SPFs. Thus, although a router has to calculate the repair paths for 254 n-1 failures, the computational effort is much less than n-1 SPFs. 256 Experiments on a selection of real world network topologies with 257 between 40 and 400 nodes suggest that the worst-case computational 258 complexity using the above optimizations is equivalent to performing 259 between 5 and 13 full SPFs. Further optimizations are described in 260 section 6. 262 3.1. Computing not-via repairs in routing vector protocols 264 While this document focuses on link state routing protocols, it is 265 equally possible to compute not-via repairs in distance vector (e.g. 266 RIP) or path vector (e.g. BGP) routing protocols. This can be 267 achieved with very little protocol modification by advertising the 268 not-via address in the normal way, but ensuring that the information 269 about a not-via address Ps is not propagated through the node S. In 270 the case of link protection this simply means that the advertisement 271 from P to S is suppressed, with the result that S and all other nodes 272 compute a route to Ps which doesn't traverse S, as required. 274 In the case of node protection, where P is the protected node, and N 275 is some neighbor, the advertisement of Np must be suppressed not only 276 across the link N->P, but also across any link to P. The simplest way 277 of achieving this is for P itself to perform the suppression of any 278 address of the form Xp. 280 4. Operation of Repairs 282 This section explains the basic operation of the not-via repair of 283 node and link failure. 285 4.1. Node Failure 287 When router P fails (Figure 2) S encapsulates any packet that it 288 would send to B via P to Bp, and then sends the encapsulated packet 289 on the shortest path to Bp. S follows the same procedure for routers 290 A and C in Figure 2. The packet is decapsulated at the repair target 291 (A, B or C) and then forwarded normally to its destination. The 292 repair target can be determined as part of the normal SPF by 293 recording the "next-next-hop" for each destination in addition to the 294 normal next-hop. 296 Notice that with this technique only one level of encapsulation is 297 needed, and that it is possible to repair ANY failure regardless of 298 link metrics and any asymmetry that may be present in the network. 299 The only exception to this is where the failure was a single point of 300 failure that partitioned the network, in which case ANY repair is 301 clearly impossible. 303 4.2. Link Failure 305 The normal mode of operation of the network would be to assume router 306 failure. However, where some destinations are only reachable through 307 the failed router, it is desirable that an attempt be made to repair 308 to those destinations by assuming that only a link failure has 309 occurred. 311 To perform a link repair, S encapsulates to Ps (i.e. it instructs the 312 network to deliver the packet to P not-via S). All of the neighbors 313 of S will have calculated a path to Ps in case S itself had failed. 314 S could therefore give the packet to any of its neighbors (except, of 315 course, P). However, S should preferably send the encapsulated 316 packet on the shortest available path to P. This path is calculated 317 by running an SPF with the link SP failed. Note that this may again 318 be an incremental calculation, which can terminate when address Ps 319 has been reattached. 321 4.2.1. Loop Prevention Under Node Failure 323 It is necessary to consider the behavior of IPFRR solutions when a 324 link repair is attempted in the presence of node failure. In its 325 simplest form the not-via IPFRR solution prevents the formation of 326 loops forming as a result of mutual repair, by never providing a 327 repair path for a not-via address. The repair of packets with not- 328 via addresses is considered in more detail in Section 5.3. Referring 329 to Figure 2, if A was the neighbor of P that was on the link repair 330 path from S to P, and P itself had failed, the repaired packet from S 331 would arrive at A encapsulated to Ps. A would have detected that the 332 AP link had failed and would normally attempt to repair the packet. 333 However, no repair path is provided for any not-via address, and so A 334 would be forced to drop the packet, thus preventing the formation of 335 loop. 337 4.3. Multi-homed Prefixes 339 A multi-homed Prefix (MHP) is a prefix that is reachable via more 340 than one router in the network. Some of these may be repairable 341 using LFAs as described in [RFC5286]. Only those without such a 342 repair need be considered here. 344 When IPFRR router S (Figure 3) discovers that P has failed, it needs 345 to send packets addressed to the MHP X, which is normally reachable 346 through P, to an alternate router, which is still able to reach X. 348 X X X 349 | | | 350 | | | 351 | Sp |Pb | 352 Z...............S----------P----------B...............Y 353 Ps|Pc Bp 354 | 355 Cp| 356 C 358 Figure 3: Multi-homed Prefixes 360 S should choose the closest router that can reach X during the 361 failure as the alternate router. S determines which router to use as 362 the alternate while running the SPF with P failed. This is 363 accomplished by the normal process of re-attaching a leaf node to the 364 core topology (this is sometimes known as a "partial SPF"). 366 First, consider the case where the shortest alternate path to X is 367 via Z. S can reach Z without using the failed router P. However, S 368 cannot just send the packet towards Z, because the other routers in 369 the network will not be aware of the failure of P, and may loop the 370 packet back to S. S therefore encapsulates the packet to Z (using a 371 normal address for Z). When Z receives the encapsulated packet it 372 removes the encapsulation and forwards the packet to X. 374 Now consider the case where the shortest alternate path to X is via 375 Y, which S reaches via P and B. To reach Y, S must first repair the 376 packet to B using the normal not-via repair mechanism. To do this S 377 encapsulates the packet for X to Bp. When B receives the packet it 378 removes the encapsulation and discovers that the packet is intended 379 for MHP X. The situation now reverts to the previous case, in which 380 the shortest alternate path does not require traversal of the 381 failure. B therefore follows the algorithm above and encapsulates 382 the packet to Y (using a normal address for Y). Y removes the 383 encapsulation and forwards the packet to X. 385 It may be that the cost of reaching X using local delivery from the 386 alternate router (i.e. Z or Y) is greater than the cost of reaching 387 X via P. Under those circumstances, the alternate router would 388 normally forward to X via P, which would cause the IPFRR repair to 389 loop. To prevent the repair from looping the alternate router must 390 locally deliver a packet received via a repair encapsulation. This 391 may be specified by using a special address with the above semantics. 392 Note that only one such address is required per node. Notice that 393 using the not-via approach, only one level of encapsulation was 394 needed to repair MHPs to the alternate router. 396 4.4. Installation of Repair Paths 398 The following algorithm is used by node S (Figure 3) to pre- 399 calculate and install repair paths in the FIB, ready for immediate 400 use in the event of a failure. It is assumed that the not-via repair 401 paths have already been calculated as described above. 403 For each neighbor P, consider all destinations which are reachable 404 via P in the current topology:- 406 1. For all destinations with an ECMP or LFA repair (as described in 407 [RFC5286]) install that repair. 409 2. For each destination (DR) that remains, identify in the current 410 topology the next-next-hop (H) (i.e. the neighbor of P that P 411 will use to send the packet to DR). This can be determined 412 during the normal SPF run by recording the additional 413 information. If S has a path to the not-via address Hp (H not 414 via P), install a not-via repair to Hp for the destination DR. 416 3. Identify all remaining destinations (M) which can still be 417 reached when node P fails. These will be multi-homed prefixes 418 that are not repairable by LFA, for which the normal attachment 419 node is P, or a router for which P is a single point of failure, 420 and have an alternative attachment point that is reachable after 421 P has failed. One way of determining these destinations would be 422 to run an SPF rooted at S with node P removed, but an 423 implementation may record alternative attachment points during 424 the normal SPF run. In either case, the next best point of 425 attachment can also be determined for use in step (4) below. 427 4. For each multi-homed prefix (M) identified in step (3):- 429 A. Identify the new attachment node (as shown in Figure 3). 430 This may be:- 432 a. Y, where the next hop towards Y is P, or 434 b. Z, where the next hop towards Z is not P. 436 If the attachment node is Z, install the repair for M as a 437 tunnel to Z' (where Z' is the address of Z that is used to 438 force local forwarding). 440 B. For the subset of prefixes (M) that remain (having attachment 441 point Y), install the repair path previously installed for 442 destination Y. 444 For each destination (DS) that remains, install a not-via repair 445 to Ps (P not via S). Note, these are destinations for which node 446 P is a single point of failure, and can only be repaired by 447 assuming that the apparent failure of node P was simply a failure 448 of the S-P link. Note that, if available, a downstream path to P 449 may be used for such a repair. This cannot generate a persistent 450 loop in the event of the failure of node P, but if one neighbor 451 of P uses a not-via repair and another uses a downstream path, it 452 is possible for a packet sent on the downstream path to be 453 returned to the sending node inside a not-via encapsulation. 454 Since packets destined to not-via addresses are not repaired, the 455 packet will be dropped after executing a single turn loop. 457 5. Compound Failures 459 The following types of failures involve more than one component: 461 1. Shared Risk Link Groups 463 2. Local Area Networks 465 3. Multiple Independent Failures 467 The considerations that apply in each of the above situations are 468 described in the following sections. 470 5.1. Shared Risk Link Groups 472 A Shared Risk Link Group (SRLG) is a set of links whose failure can 473 be caused by a single action such as a conduit cut or line card 474 failure. When repairing the failure of a link that is a member of an 475 SRLG, it must be assumed that all the other links that are also 476 members of the SRLG have also failed. Consequently, any repair path 477 must be computed to avoid not just the adjacent link, but also all 478 the links which are members of the same SRLG. 480 In Figure 4 below, the links S-P and A-B are both members of SRLG 481 "a". The semantics of the not-via address Ps changes from simply "P 482 not-via the link S-P" to be "P not-via the link S-P or any other link 483 with which S-P shares an SRLG" In Figure 4 this is the links that are 484 members of SRLG "a". I.e. links S-P and A-B. Since the information 485 about SRLG membership of all links is available in the Link State 486 Database, all nodes computing routes to the not-via address Ps can 487 infer these semantics, and perform the computation by failing all the 488 links in the SRLG when running the iSPF. 490 Note that it is not necessary for S to consider repairs to any other 491 nodes attached to members of the SRLG (such as B). It is sufficient 492 for S to repair to the other end of the adjacent link (P in this 493 case). 495 a Ps 496 S----------P---------D 497 | | 498 | a | 499 A----------B 500 | | 501 | | 502 C----------E 504 Figure 4: Shared Risk Link Group 506 In some cases, it may be that the links comprising the SRLG occur in 507 series on the path from S to the destination D, as shown in Figure 5. 508 In this case, multiple consecutive repairs may be necessary. S will 509 first repair to Ps, then P will repair to Dp. In both cases, because 510 the links concerned are members of SRLG "a" the paths are computed to 511 avoid all members of SRLG "a". 513 a Ps a Dp 514 S----------P---------D 515 | | | 516 | a | | 517 A----------B | 518 | | | 519 | | | 520 C----------E---------F 522 Figure 5: Shared Risk Link Group members in series 524 While the use of multiple repairs in series introduces some 525 additional overhead, these semantics avoid the potential 526 combinatorial explosion of not-via addresses that could otherwise 527 occur. 529 Note that although multiple repairs are used, only a single level of 530 encapsulation is required. This is because the first repair packet 531 is decapsulated before the packet is re-encapsulated using the not- 532 via address corresponding to the far side of the next link which is a 533 member of the same SRLG. In some cases the de-capsulation and re- 534 encapsulation takes place (at least notionally) at a single node, 535 while in other cases, these functions may be performed by different 536 nodes. This scenario is illustrated in Figure 6 below. 538 a Ps a Dg 539 S----------P---------G--------D 540 | | | | 541 | a | | | 542 A----------B | | 543 | | | | 544 | | | | 545 C----------E---------F--------H 547 Figure 6: Shared Risk Link Group members in series 549 In this case, S first encapsulates to Ps, and node P decapsulates the 550 packet and forwards it "native" to G using its normal FIB entry for 551 destination D. G then repairs the packet to Dg. 553 It can be shown that such multiple repairs can never form a loop 554 because each repair causes the packet to move closer to its 555 destination. 557 It is often the case that a single link may be a member of multiple 558 SRLGs, and those SRLGs may not be isomorphic. This is illustrated in 559 Figure 7 below. 561 ab Ps a Dg 562 S----------P---------G--------D 563 | | | | 564 | a | | | 565 A----------B | | 566 | | | | 567 | b | | b | 568 C----------E---------F--------H 569 | | 570 | | 571 J----------K 573 Figure 7: Multiple Shared Risk Link Groups 575 The link SP is a member of SRLGs "a" and "b". When a failure of the 576 link SP is detected, it must be assumed that BOTH SRLGs have failed. 577 Therefore the not-via path to Ps must be computed by failing all 578 links which are members of SRLG "a" or SRLG "b". I.e. the semantics 579 of Ps is now "P not-via any links which are members of any of the 580 SRLGs of which link SP is a member". This is illustrated in Figure 8 581 below. 583 ab Ps a Dg 584 S----/-----P---------G---/----D 585 | | | | 586 | a | | | 587 A----/-----B | | 588 | | | | 589 | b | | b | 590 C----/-----E---------F---/----H 591 | | 592 | | 593 J----------K 595 Figure 8: Topology used for repair computation for link S-P 597 In this case, the repair path to Ps will be S-A-C-J-K-E-B-P. It may 598 appear that there is no path to D because GD is a member of SRLG "a" 599 and FH is a member of SRLG "b". This is true if BOTH SRLGs "a" and 600 "b" have in fact failed. But that would be an instance of multiple 601 uncorrelated failures which are out of scope for this design. In 602 practice it is likely that there is only a single failure, i.e. 603 either SRLG "a" or SRLG "b" has failed, but not both. These two 604 possibilities are indistinguishable from the point of view of the 605 repairing router S and so it must repair on the assumption that both 606 are unavailable. However, each link repair is considered 607 independently. The repair to Ps delivers the packet to P which then 608 forwards the packet to G. When the packet arrives at G, if SRLG "a" 609 has failed it will be repaired around the path G-F-H-D. This is 610 illustrated in Figure 9 below. If, on the other hand, SRLG "b" has 611 failed, link GD will still be available. In this case the packet 612 will be delivered as normal across the link GD. 614 ab Ps a Dg 615 S----/-----P---------G---/----D 616 | | | | 617 | a | | | 618 A----/-----B | | 619 | | | | 620 | b | | b | 621 C----------E---------F--------H 622 | | 623 | | 624 J----------K 626 Figure 9: Topology used for repair computation for link G-D 628 A repair strategy that assumes the worst-case failure for each link 629 can often result in longer repair paths than necessary. In cases 630 where only a single link fails, rather than the full SRLG, this 631 strategy may occasionally fail to identify a repair even though a 632 viable repair path exists in the network. The use of sub-optimal 633 repair paths is an inevitable consequence of this compromise 634 approach. The failure to identify any repair is a serious 635 deficiency, but is a rare occurrence in a robustly designed network. 636 This problem can be addressed by:- 638 1. Reporting that the link in question is irreparable, so that the 639 network designer can take appropriate action. 641 2. Modifying the design of the network to avoid this possibility. 643 3. Using some form of SRLG diagnostic (for example, by running BFD 644 over alternate repair paths) to determine which SRLG member(s) 645 has actually failed and using this information to select an 646 appropriate pre-computed repair path. However, aside from the 647 complexity of performing the diagnostics, this requires multiple 648 not-via addresses per interface, which has poor scaling 649 properties. 651 4. Using the machanism described in Section 5.3 653 5.1.1. Use of LFAs with SRLGs 655 Section 4.1 above describes the repair of links which are members of 656 one or more SRLGs. LFAs can be used for the repair of such links 657 provided that any other link with which S-P shares an SRLG is avoided 658 when computing the LFA. This is described for the simple case of 659 "local-SRLGs" in [RFC5286]. 661 5.2. Local Area Networks 663 LANs are a special type of SRLG and are solved using the SRLG 664 mechanisms outlined above. With all SRLGs there is a trade-off 665 between the sophistication of the fault detection and the size of the 666 SRLG. Protecting against link failure of the LAN link(s) is 667 relatively straightforward, but as with all fast reroute mechanisms, 668 the problem becomes more complex when it is desired to protect 669 against the possibility of failure of the nodes attached to the LAN 670 as well as the LAN itself. 672 +--------------Q------C 673 | 674 | 675 | 676 A--------S-------(N)-------------P------B 677 | 678 | 679 | 680 +--------------R------D 682 Figure 10: Local Area Networks 684 Consider the LAN shown in Figure 10. For connectivity purposes, we 685 consider that the LAN is represented by the pseudonode (N). To 686 provide IPFRR protection, S must run a connectivity check to each of 687 its protected LAN adjacencies P, Q, and R, using, for example BFD 688 [RFC5880]. 690 When S discovers that it has lost connectivity to P, it is unsure 691 whether the failure is: 693 o its own interface to the LAN, 695 o the LAN itself, 697 o the LAN interface of P, 699 o the node P. 701 5.2.1. Simple LAN Repair 703 A simple approach to LAN repair is to consider the LAN and all of its 704 connected routers as a single SRLG. Thus, the address P not via the 705 LAN (Pl) would require P to be reached not-via any router connected 706 to the LAN. This is shown in Figure 11. 708 Ql Cl 709 +-------------Q--------C 710 | Qc 711 | 712 As Sl | Pl Bl 713 A--------S-------(N)------------P--------B 714 Sa | Pb 715 | 716 | Rl Dl 717 +-------------R--------D 718 Rd 720 Figure 11: Local Area Networks - LAN SRLG 722 In this case, when S detected that P had failed it would send traffic 723 reached via P and B to B not-via the LAN or any router attached to 724 the LAN (i.e. to Bl). Any destination only reachable through P would 725 be addressed to P not-via the LAN or any router attached to the LAN 726 (except of course P). 728 Whilst this approach is simple, it assumes that a large portion of 729 the network adjacent to the failure has also failed. This will 730 result in the use of sub-optimal repair paths and in some cases the 731 inability to identify a viable repair. 733 5.2.2. LAN Component Repair 735 In this approach, possible failures are considered at a finer 736 granularity, but without the use of diagnostics to identify the 737 specific component that has failed. Because S is unable to diagnose 738 the failure it must repair traffic sent through P and B, to B not- 739 via P,N (i.e. not via P and not via N), on the conservative 740 assumption that both the entire LAN and P have failed. Destinations 741 for which P is a single point of failure must as usual be sent to P 742 using an address that avoids the interface by which P is reached from 743 S, i.e. to P not-via N. Similarly for routers Q and R. 745 Notice that each router that is connected to a LAN must, as usual, 746 advertise one not-via address for each neighbor. In addition, each 747 router on the LAN must advertise an extra address not via the 748 pseudonode (N). 750 Notice also that each neighbor of a router connected to a LAN must 751 advertise two not-via addresses, the usual one not via the neighbor 752 and an additional one, not via either the neighbor or the pseudonode. 753 The required set of LAN address assignments is shown in Figure 12 754 below. Each router on the LAN, and each of its neighbors, is 755 advertising exactly one address more than it would otherwise have 756 advertised if this degree of connectivity had been achieved using 757 point-to-point links. 759 Qs Qp Qc Cqn 760 +--------------Q---------C 761 | Qr Qn Cq 762 | 763 Asn Sa Sp Sq | Ps Pq Pb Bpn 764 A--------S-------(N)-------------P---------B 765 As Sr Sn | Pr Pn Bp 766 | 767 | Rs Rp Pd Drn 768 +--------------R---------D 769 Rq Rn Dr 771 Figure 12: Local Area Networks 773 5.2.3. LAN Repair Using Diagnostics 775 A more specific LAN repair can be undertaken by using diagnostics. 776 In order to explicitly diagnose the failed network component, S 777 correlates the connectivity reports from P and one or more of the 778 other routers on the LAN, in this case, Q and R. If it lost 779 connectivity to P alone, it could deduce that the LAN was still 780 functioning and that the fault lay with either P, or the interface 781 connecting P to the LAN. It would then repair to B not via P (and P 782 not-via N for destinations for which P is a single point of failure) 783 in the usual way. If S lost connectivity to more than one router on 784 the LAN, it could conclude that the fault lay only with the LAN, and 785 could repair to P, Q and R not-via N, again in the usual way. 787 5.3. Multiple Independent Failures 789 IPFRR repair of multiple simultaneous failures which are not members 790 of a known SRLG is complicated by the problem that the use of 791 multiple concurrent repairs may result in looping repair paths. As 792 described in Section 4.2.1, the simplest method of preventing such 793 loops, is to ensure that packets addressed to a not-via address are 794 not repaired but instead are dropped. It is possible that a network 795 may experience multiple simultaneous failures. This may be due to 796 simple statistical effects, but the more likely cause is 797 unanticipated SRLGs. When multiple failures which are not part of an 798 anticipated group are detected, repairs are abandoned and the network 799 reverts to normal convergence. Although safe, this approach is 800 somewhat draconian, since there are many circumstances were multiple 801 repairs do not induce loops. 803 This section describes the properties of multiple unrelated failures 804 and proposes some methods that may be used to address this problem. 806 5.3.1. Looping Repairs 808 Let us assume that the repair mechanism is based on solely on not-via 809 repairs. LFA or downstream routes may be incorporated, and will be 810 dealt with later. 812 A------//------B------------D 813 / \ 814 / \ 815 F G 816 \ / 817 \ / 818 X------//------Y 820 Figure 13: The General Case of Multiple Failures 822 The essential case is as illustrated in Figure 13. Note that 823 depending on the repair case under consideration, there may be paths 824 present in Figure 13, that are in addition to those shown in the 825 figure. For example there may be paths between A and B, and/or 826 between X and Y. These paths are omitted for graphical clarity. 828 There are three cases to consider: 830 1) Consider the general case of a pair of protected links A-B and 831 X-Y as shown in the network fragment shown Figure 13. If the 832 repair path for A-B does not traverse X-Y and the repair path for 833 X-Y does not traverse A-B, this case is completely safe and will 834 not cause looping or packet loss. 836 A more common variation of this case is shown in Figure 14, which 837 shows two failures in different parts of the network in which a 838 packet from A to D traverses two concatenated repairs. 840 A------//------B------------X------//------Y------D 841 | | | | 842 | | | | 843 M--------------+ N--------------+ 845 Figure 14: Concatenated Repairs 847 2) In Figure 13, the repair for A-B traverses X-Y, but the repair 848 for X-Y does not traverse A-B. This case occurs when the not-via 849 path from A to B traverses link X-Y, but the not-via path from X 850 to Y traverses some path not shown in Figure 13. Without the 851 multi-failure mechanism described in this section the repaired 852 packet for A-B would be dropped when it reached X-Y, since the 853 repair of repaired packets would be forbidden. However, if this 854 packet were allowed to be repaired, the path to D would be 855 complete and no harm would be done, although two levels of 856 encapsulation would be required. 858 3) The repair for A-B traverses X-Y AND the repair for X-Y 859 traverses A-B. In this case unrestricted repair would result in 860 looping packets and increasing levels of encapsulation. 862 The challenge in applying IPFRR to a network that is undergoing 863 multiple failures is, therefore, to identify which of these cases 864 exist in the network and react accordingly. 866 5.3.2. Outline Solution 868 When A is computing the not-via repair path for A-B (i.e. the path 869 for packets addressed to Ba, read as "B not-via A") it is aware of 870 the list of nodes which this path traverses. This can be recorded by 871 a simple addition to the SPF process, and the not-via addresses 872 associated with each forward link can be determined. If the path 873 were A, F, X, Y, G, B, (Figure 13) the list of not-via addresses 874 would be: Fa, Xf, Yx, Gy, Bg. Under standard not-via operation, A 875 would populate its FIB such that all normal addresses normally 876 reachable via A-B would be encapsulated to Ba when A-B fails, but 877 traffic addressed to any not-via address arriving at A would be 878 dropped. The new procedure modifies this such that any traffic for a 879 not-via address normally reachable over A-B is also encapsulated to 880 Ba unless the not-via address is one of those previously identified 881 as being on the path to Ba, for example Yx, in which case the packet 882 is dropped. 884 The above procedure allows cases 1 and 2 above to be repaired, while 885 preventing the loop which would result from case 3. 887 Note that this is accomplished by pre-computing the required FIB 888 entries, and does not require any detailed packet inspection. The 889 same result could be achieved by checking for multiple levels of 890 encapsulation and dropping any attempt to triple encapsulate. 891 However, this would require more detailed inspection of the packet, 892 and causes difficulties when more than 2 "simultaneous" failures are 893 contemplated. 895 So far we have permitted benign repairs to coexist, albeit sometimes 896 requiring multiple encapsulation. Note that in many cases there will 897 be no performance impact since unless both failures are on the same 898 node, the two encapsulations or two decapsulations will be performed 899 at different nodes. There is however the issue of the MTU impact of 900 multiple encapsulations. 902 In the following sub-section we consider the various strategies that 903 may be applied to case 3 - mutual repairs that would loop. 905 5.3.3. Looping Repairs 907 In case 3, the simplest approach is to simply not install repairs for 908 repair paths that might loop. In this case, although the potentially 909 looping traffic is dropped, the traffic is not repaired. If we 910 assume that a hold-down is applied before reconvergence in case the 911 link failure was just a short glitch, and if a loop free convergence 912 mechanism further delays convergence, then the traffic will be 913 dropped for an extended period. In these circumstances it would be 914 better to "abandon all hope" (AAH) 915 [] and immediately 916 invoke normal re-convergence. 918 Note that it is not sufficient to expedite the issuance of an LSP 919 reporting the failure, since this may be treated as a permitted 920 simultaneous failure by the oFIB algorithm. It is therefore 921 necessary to explicitly trigger an oFIB AAH. 923 5.3.3.1. Dropping Looping Packets 925 One approach to case 3 is to allow the repair, and to experimentally 926 discover the incompatibility of the repairs if and when they occur. 927 With this method we permit the repair in case 3 and trigger AAH when 928 a packet drop count on the not-via address has been incremented. 929 Alternatively, it is possible to wait until the LSP describing the 930 change is issued normally (i.e. when X announces the failure of X-Y). 931 When the repairing node A, which has precomputed that X-Y failures 932 are mutually incompatible with its own repairs receives this LSP it 933 can then issue the AAH. This has the disadvantage that it doesn't 934 overcome the hold-down delay, but it requires no "data-driven" 935 operation, and it still has the required effect of abandoning the 936 oFIB which is probably the longer of the delays (although with 937 signalled oFIB this should be sub-second). 939 Whilst both of the experimental approaches described above are 940 feasible, they tend to induce AAH in the presence of otherwise 941 feasible repairs, and they are contrary to the philosophy of repair 942 pre-determination that has been applied to existing IPFRR solutions. 944 5.3.3.2. Computing non-looping Repairs of Repairs 946 An alternative approach to simply dropping the looping packets, or to 947 detecting the loop after it has occurred, is to use secondary SRLGs. 948 With a link state routing protocol it is possible to precompute the 949 incompatibility of the repairs in advance and to compute an 950 alternative SRLG repair path. Although this does considerably 951 increase the computational complexity it may be possible to compute 952 repair paths that avoid the need to simply drop the offending 953 packets. 955 This approach requires us to identify the mutually incompatible 956 failures, and advertise them as "secondary SRLGs". When computing 957 the repair paths for the affected not-via addresses these links are 958 simultaneously failed. Note that the assumed simultaneous failure 959 and resulting repair path only applies to the repair path computed 960 for the conflicting not-via addresses, and is not used for normal 961 addresses. This implies that although there will be a longer repair 962 path when there is more than one failure, if there is a single 963 failure the repair path length will be "normal". 965 Ideally we would wish to only invoke secondary SRLG computation when 966 we are sure that the repair paths are mutually incompatible. 967 Consider the case of node A in Figure 13. A first identifies that 968 the repair path for A-B is via F-X-Y-G-B. It then explores this path 969 determining the repair path for each link in the path. Thus, for 970 example, it performs a check at X by running an SPF rooted at X with 971 the X-Y link removed to determine whether A-B is indeed on X's repair 972 path for packets addressed to Yx. 974 Some optimizations are possible in this calculation, which appears at 975 first sight to be order hk (where h is the average hop length of 976 repair paths and k is the average number of neighbours of a router). 977 When A is computing its set of repair paths, it does so for all its k 978 neighbours. In each case it identifies a list of node pairs 979 traversed by each repair. These lists may often have one or more 980 node pairs in common, so the actual number of link failures which 981 require investigation is the union of these sets. It is then 982 necessary to run an SPF rooted at the first node of each pair (the 983 first node because the pairings are ordered representing the 984 direction of the path), with the link to the second node removed. 985 This SPF, while not an incremental, can be terminated as soon as the 986 not-via address is reached. For example, when running the SPF rooted 987 at X, with the link X-Y removed, the SPF can be terminated when Yx is 988 reached. Once the path has been found, the path is checked to 989 determine if it traverses any of A's links in the direction away from 990 A. Note that, because the node pair XY may exist in the list for more 991 than one of A's links (i.e. it lies on more than one repair path), it 992 is necessary to identify the correct list, and hence link which has a 993 mutually looping repair path. That link of A is then advertised by A 994 as a secondary SRLG paired with the link X-Y. Also note that X will 995 be running this algorithm as well, and will identify that XY is 996 paired with A-B and so advertise it. This could perhaps be used as a 997 further check. 999 The ordering of the pairs in the lists is important. i.e. X-Y and 1000 Y-X are dealt with separately. If and only if the repairs are 1001 mutually incompatible, we need to advertise the pair of links as a 1002 secondary SRLG, and then ALL nodes compute repair paths around both 1003 failures using an additional not-via address with the semantics not- 1004 via A-B AND not-via X-Y. 1006 A further possibility is that because we are going to the trouble of 1007 advertising these SRLG sets, we could also advertise the new repair 1008 path and only get the nodes on that path to perform the necessary 1009 computation. Note also that once we have reached Q space with 1010 respect to the two failures we need no longer continue the 1011 computation, so we only need to notify the nodes on the path that are 1012 not in Q-space. 1014 One cause of mutually looping repair paths is the existence of nodes 1015 with only two links, or sections of the network which are only bi- 1016 connected. In these cases, repair is clearly impossible - the 1017 failure of both links partitions the network. It would be 1018 advantageous to be able to identify these cases, and inhibit the 1019 fruitless advertisement of the secondary SRLG information. This 1020 could be achieved by the node detecting the requirement for a 1021 secondary SRLG, first running the not-via computation with both links 1022 removed. If this does not result in a path, it is clear that the 1023 network would be partitioned by such a failure, and so no 1024 advertisement is required. 1026 5.3.3.3. N-level Mutual Loops 1028 [Editors' Note: This section needs to be reviewed before final 1029 publication] 1031 It is tempting to conclude that the mechanism described above can be 1032 applied to the general case of N failures. If we use the approach of 1033 assuming that repairs are not mutual, and catching the loops and 1034 executing AAH when they occur, then we can attempt repairs in the 1035 case of N failures. 1037 If we use the approach of avoiding potentially mutual repairs and 1038 creating secondary SRLG, then we have to explore N levels of repair, 1039 where N is the number of simultaneous failures we wish to repair. 1041 5.3.4. Mixing LFAs and Not-via 1043 So far in this section we have assumed that all repairs use not-via 1044 tunnels. However, in practise we may wish to use LFAs or downstream 1045 routes where available. This complicates the issue, because their 1046 use results in packets which are being repaired, but NOT addressed to 1047 not-via addresses. If BOTH links are using downstream routes there 1048 is no possibility of looping, since it is impossible to have a pair 1049 of nodes which are both downstream of each other [RFC5286]. 1051 Loops can however occur when LFAs are used. An obvious example is 1052 the well known node repair problem with LFAs [RFC5286]. If one link 1053 is using a downstream route, while the other is using a not-via 1054 tunnel, the potential mechanism described above would work provided 1055 it were possible to determine the nodes on the path of the downstream 1056 route. Some methods of computing downstream routes do not provide 1057 this path information. If the path information is however available, 1058 the link using a downstream route will have a discard FIB entry for 1059 the not-via address of the other link. The consequence is that 1060 potentially looping packets will be discarded when they attempt to 1061 cross this link. 1063 In the case where the mutual repairs are both using not-via repairs, 1064 the loop will be broken when the packet arrives at the second 1065 failure. However packets are unconditionally repaired by means of a 1066 downstream routes, and thus when the mutual pair consists of a 1067 downstream route and a not-via repair, the looping packet will only 1068 be dropped when it gets back to the first failure. i.e. it will 1069 execute a single turn of the loop before being dropped. 1071 There is a further complication with downstream routes, since 1072 although the path may be computed to the far side of the failure, the 1073 packet may "peel off" to its destination before reaching the far side 1074 of the failure. In this case it may traverse some other link which 1075 has failed and was not accounted for on the computed path. If the 1076 A-B repair (Figure 1) is a downstream route and the X-Y repair is a 1077 not-via repair, we can have the situation where the X-Y repair 1078 packets encapsulated to Yx follow a path which attempts to traverse 1079 A-B. If the A-B repair path for "normal" addresses is a downstream 1080 route, it cannot be assumed that the repair path for packets 1081 addressed to Yx can be sent to the same neighbour. This is because 1082 the validity of a downstream route must be ascertained in the 1083 topology represented by Yx, i.e. that with the link X-Y failed. This 1084 is not the same topology that was used for the normal downstream 1085 calculation, and use of the normal downstream route for the 1086 encapsulated packets may result in an undetected loop. If it is 1087 computationally feasible to check the downstream route in this 1088 topology (i.e. for any not-via address Qp which traverses A-B we must 1089 perform the downstream calculation for that not-via address in the 1090 topology with link Q-P failed.), then the downstream repair for Yx 1091 can safely be used. These packets cannot re-visit X-Y, since by 1092 definition they will avoid that link. Alternatively, the packet 1093 could be always repaired in a not-via tunnel. i.e. even though the 1094 normal repair for traffic traversing A-B would be to use a downstream 1095 route, we could insist that such traffic addressed to a not-via 1096 address MUST use a tunnel to Ba. Such a tunnel would only be 1097 installed for an address Qp if it were established that it did not 1098 traverse Q-P (using the rules described above). 1100 6. Optimizing not-via computations using LFAs 1102 If repairing node S has an LFA to the repair endpoint it is not 1103 necessary for any router to perform the incremental SPF with the link 1104 SP removed in order to compute the route to the not-via address Ps. 1105 This is because the correct routes will already have been computed as 1106 a result of the SPF on the base topology. Node S can signal this 1107 condition to all other routers by including a bit in its LSP or LSA 1108 associated with each LFA protected link. Routers computing not-via 1109 routes can then omit the running of the iSPF for links with this bit 1110 set. 1112 When running the iSPF for a particular link AB, the calculating 1113 router first checks whether the link AB is present in the existing 1114 SPT. If the link is not present in the SPT, no further work is 1115 required. This check is a normal part of the iSPF computation. 1117 If the link is present in the SPT, this optimization introduces a 1118 further check to determine whether the link is marked as protected by 1119 an LFA in the direction in which the link appears in the SPT. If so 1120 the iSPF need not be performed. For example, if the link appears in 1121 the SPT in the direction A->B and A has indicated that the link AB is 1122 protected by an LFA no further action is required for this link. 1124 If the receipt of this information is delayed, the correct operation 1125 of the protocol is not compromised provided that the necessity to 1126 perform a not-via computation is re-evaluated whenever new 1127 information arrives. 1129 This optimization is not particularly beneficial to nodes close to 1130 the repair since, as has been observed above, the computation for 1131 nodes on the LFA path is trivial. However, for nodes upstream of the 1132 link SP for which S-P is in the path to P, there is a significant 1133 reduction in the computation required. 1135 7. Multicast 1137 Multicast traffic can be repaired in a similar way to unicast. The 1138 multicast forwarder is able to use the not-via address to which the 1139 multicast packet was addressed as an indication of the expected 1140 receive interface and hence to correctly run the required RPF check. 1142 In some cases, all the destinations, including the repair endpoint, 1143 are repairable by an LFA. In this case, all unicast traffic may be 1144 repaired without encapsulation. Multicast traffic still requires 1145 encapsulation, but for the nodes on the LFA repair path the 1146 computation of the not-via forwarding entry is unnecessary since, by 1147 definition, their normal path to the repair endpoint is not via the 1148 failure. 1150 A more complete description of multicast operation is for further 1151 study. 1153 8. Fast Reroute in an MPLS LDP Network. 1155 Not-via addresses are IP addresses and LDP [RFC5036] will distribute 1156 labels for them in the usual way. The not-via repair mechanism may 1157 therefore be used to provide fast re-route in an MPLS network by 1158 first pushing the label which the repair endpoint uses to forward the 1159 packet, and then pushing the label corresponding to the not-via 1160 address needed to effect the repair. Referring once again to 1161 Figure 1, if S has a packet destined for D that it must reach via P 1162 and B, S first pushes B's label for D. S then pushes the label that 1163 its next hop to Bp needs to reach Bp. 1165 Note that in an MPLS LDP network it is necessary for S to have the 1166 repair endpoint's label for the destination. When S is effecting a 1167 link repair it already has this. In the case of a node repair, S 1168 either needs to set up a directed LDP session with each of its 1169 neighbor's neighbors, or it needs to use the next-next hop label 1170 distribution mechanism proposed in [I-D.shen-mpls-ldp-nnhop-label]. 1172 9. Encapsulation 1174 Any IETF specified IP in IP encapsulation may be used to carry a not- 1175 via repair. IP in IP [RFC2003], GRE [RFC1701] and L2TPv3 [RFC3931], 1176 all have the necessary and sufficient properties. The requirement is 1177 that both the encapsulating router and the router to which the 1178 encapsulated packet is addressed have a common ability to process the 1179 chosen encapsulation type. When an MPLS LDP network is being 1180 protected, the encapsulation would normally be an additional MPLS 1181 label. In an MPLS enabled IP network an MPLS label may be used in 1182 place of an IP in IP encapsulation in the case above. 1184 10. Routing Extensions 1186 IPFRR requires IGP extensions. Each IPFRR router that is directly 1187 connected to a protected network component must advertise a not-via 1188 address for that component. This must be advertised in such a way 1189 that the association between the protected component (link, router or 1190 SRLG) and the not-via address can be determined by the other routers 1191 in the network. 1193 It is necessary that not-via capable routers advertise in the IGP 1194 that they will calculate not-via routes. 1196 It is necessary for routers to advertise the type of encapsulation 1197 that they support (MPLS, GRE, L2TPv3 etc). However, the deployment 1198 of mixed IP encapsulation types within a network is discouraged. 1200 11. Incremental Deployment 1202 Incremental deployment is supported by excluding routers that are not 1203 calculating not-via routes (as indicated by their capability 1204 information flooded with their link state information) from the base 1205 topology used for the computation of repair paths. In that way 1206 repairs may be steered around islands of routers that are not IPFRR 1207 capable. Routers that are protecting a network component need to 1208 have the capability to encapsulate and decapsulate packets. However, 1209 routers that are on the repair path only need to be capable of 1210 calculating not-via paths and including the not-via addresses in 1211 their FIB i.e. these routers do not need any changes to their 1212 forwarding mechanism. 1214 12. IANA Considerations 1216 There are no IANA considerations that arise from this draft. 1218 13. Security Considerations 1220 The repair endpoints present vulnerability in that they might be used 1221 as a method of disguising the delivery of a packet to a point in the 1222 network. The primary method of protection should be through the use 1223 of a private address space for the not-via addresses. These 1224 addresses MUST NOT be advertised outside the area, and SHOULD be 1225 filtered at the network entry points. In addition, a mechanism might 1226 be developed that allowed the use of the mild security available 1227 through the use of a key [RFC1701] [RFC3931]. With the deployment of 1228 such mechanisms, the repair endpoints would not increase the security 1229 risk beyond that of existing IP tunnel mechanisms. An attacker may 1230 attempt to overload a router by addressing an excessive traffic load 1231 to the de-capsulation endpoint. Typically, routers take a 50% 1232 performance penalty in decapsulating a packet. The attacker could 1233 not be certain that the router would be impacted, and the extremely 1234 high volume of traffic needed, would easily be detected as an 1235 anomaly. If an attacker were able to influence the availability of a 1236 link, they could cause the network to invoke the not-via repair 1237 mechanism. A network protected by not-via IPFRR is less vulnerable 1238 to such an attack than a network that undertook a full convergence in 1239 response to a link up/down event. 1241 14. Acknowledgements 1243 The authors would like to acknowledge contributions made by Alia 1244 Atlas and John Harper. 1246 15. References 1248 15.1. Normative References 1250 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 1251 Requirement Levels", BCP 14, RFC 2119, March 1997. 1253 15.2. Informative References 1255 [I-D.shen-mpls-ldp-nnhop-label] 1256 Shen, N., "Discovering LDP Next-Nexthop Labels", 1257 draft-shen-mpls-ldp-nnhop-label-02 (work in progress), 1258 May 2005. 1260 [ISPF] McQuillan, J., Richer, I., and E. Rosen, "ARPANET Routing 1261 Algorithm Improvements"", BBN Technical Report 3803, 1978. 1263 [RFC1701] Hanks, S., Li, T., Farinacci, D., and P. Traina, "Generic 1264 Routing Encapsulation (GRE)", RFC 1701, October 1994. 1266 [RFC2003] Perkins, C., "IP Encapsulation within IP", RFC 2003, 1267 October 1996. 1269 [RFC3931] Lau, J., Townsley, M., and I. Goyret, "Layer Two Tunneling 1270 Protocol - Version 3 (L2TPv3)", RFC 3931, March 2005. 1272 [RFC5036] Andersson, L., Minei, I., and B. Thomas, "LDP 1273 Specification", RFC 5036, October 2007. 1275 [RFC5286] Atlas, A. and A. Zinin, "Basic Specification for IP Fast 1276 Reroute: Loop-Free Alternates", RFC 5286, September 2008. 1278 [RFC5714] Shand, M. and S. Bryant, "IP Fast Reroute Framework", 1279 RFC 5714, January 2010. 1281 [RFC5880] Katz, D. and D. Ward, "Bidirectional Forwarding Detection 1282 (BFD)", RFC 5880, June 2010. 1284 Authors' Addresses 1286 Mike Shand 1287 Cisco Systems 1288 250, Longwater Avenue. 1289 Reading, Berks RG2 6GB 1290 UK 1292 Email: mshand@cisco.com 1293 Stewart Bryant 1294 Cisco Systems 1295 250, Longwater Avenue. 1296 Reading, Berks RG2 6GB 1297 UK 1299 Email: stbryant@cisco.com 1301 Stefano Previdi 1302 Cisco Systems 1303 Via Del Serafico, 200 1304 00142 Rome, 1305 Italy 1307 Email: sprevidi@cisco.com