idnits 2.17.00 (12 Aug 2021) /tmp/idnits23748/draft-thubert-rtgwg-arc-01.txt: Checking boilerplate required by RFC 5378 and the IETF Trust (see https://trustee.ietf.org/license-info): ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/1id-guidelines.txt: ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/checklist : ---------------------------------------------------------------------------- ** There are 2 instances of too long lines in the document, the longest one being 3 characters in excess of 72. Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year == Line 144 has weird spacing: '...one for relia...' == The document doesn't use any RFC 2119 keywords, yet seems to have RFC 2119 boilerplate text. -- The document date (October 18, 2013) is 3137 days in the past. Is this intentional? Checking references for intended status: Proposed Standard ---------------------------------------------------------------------------- (See RFCs 3967 and 4897 for information about using normative references to lower-maturity documents in RFCs) No issues found here. Summary: 1 error (**), 0 flaws (~~), 3 warnings (==), 1 comment (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 RTGWG P. Thubert, Ed. 3 Internet-Draft Cisco 4 Intended status: Standards Track P. Bellagamba 5 Expires: April 19, 2014 Cisco Systems 6 October 18, 2013 8 Available Routing Constructs 9 draft-thubert-rtgwg-arc-01 11 Abstract 13 This draft introduces the concept of ARC, a two-edged routing 14 construct that forms its own fault isolation and recovery domain. 15 The new paradigm can be leveraged to improve the network utilization 16 and resiliency for unicast and multicast traffic in multiple 17 environments. 19 Requirements Language 21 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 22 "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and 23 "OPTIONAL" in this document are to be interpreted as described in RFC 24 2119 [RFC2119]. 26 Status of this Memo 28 This Internet-Draft is submitted in full conformance with the 29 provisions of BCP 78 and BCP 79. 31 Internet-Drafts are working documents of the Internet Engineering 32 Task Force (IETF). Note that other groups may also distribute 33 working documents as Internet-Drafts. The list of current Internet- 34 Drafts is at http://datatracker.ietf.org/drafts/current/. 36 Internet-Drafts are draft documents valid for a maximum of six months 37 and may be updated, replaced, or obsoleted by other documents at any 38 time. It is inappropriate to use Internet-Drafts as reference 39 material or to cite them other than as "work in progress." 41 This Internet-Draft will expire on April 19, 2014. 43 Copyright Notice 45 Copyright (c) 2013 IETF Trust and the persons identified as the 46 document authors. All rights reserved. 48 This document is subject to BCP 78 and the IETF Trust's Legal 49 Provisions Relating to IETF Documents (http://trustee.ietf.org/ 50 license-info) in effect on the date of publication of this document. 51 Please review these documents carefully, as they describe your rights 52 and restrictions with respect to this document. Code Components 53 extracted from this document must include Simplified BSD License text 54 as described in Section 4.e of the Trust Legal Provisions and are 55 provided without warranty as described in the Simplified BSD License. 57 Table of Contents 59 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 2 60 2. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . 3 61 3. ARC Set representations . . . . . . . . . . . . . . . . . . . 6 62 4. Applicability . . . . . . . . . . . . . . . . . . . . . . . . 16 63 4.1. Load Balancing . . . . . . . . . . . . . . . . . . . . . . 16 64 4.1.1. Routing Hierarchies . . . . . . . . . . . . . . . . . 16 65 5. Lowest ARC First . . . . . . . . . . . . . . . . . . . . . . . 16 66 5.1. Init . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 67 5.2. Growing Trees . . . . . . . . . . . . . . . . . . . . . . 17 68 5.3. Being Safe . . . . . . . . . . . . . . . . . . . . . . . . 17 69 5.4. Bending An ARC . . . . . . . . . . . . . . . . . . . . . . 18 70 5.5. Orienting Links . . . . . . . . . . . . . . . . . . . . . 19 71 5.6. Looping or recursing . . . . . . . . . . . . . . . . . . . 19 72 6. Forwarding Along An ARC Set . . . . . . . . . . . . . . . . . 20 73 6.1. Control Plane Recovery . . . . . . . . . . . . . . . . . . 20 74 6.2. Data Plane Recovery . . . . . . . . . . . . . . . . . . . 21 75 7. Manageability . . . . . . . . . . . . . . . . . . . . . . . . 22 76 8. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 22 77 9. Security Considerations . . . . . . . . . . . . . . . . . . . 22 78 10. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . 22 79 11. References . . . . . . . . . . . . . . . . . . . . . . . . . . 22 80 11.1. Normative References . . . . . . . . . . . . . . . . . . 22 81 11.2. Informative References . . . . . . . . . . . . . . . . . 22 82 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . . 22 84 1. Introduction 86 Traditional routing and forwarding uses the concept of path as the 87 basic routing paradigm to get a packet from a source to a destination 88 by following an ordered sequence of arrows between intermediate 89 nodes. In this serial design, a path is broken as soon as a single 90 arrow is, and getting around a breakage can require path 91 recomputation, network reconvergence, and incur delays to till 92 service is restored. 94 Multiple paths can be bound together for instance to form a Directed 95 Acyclic Graph (DAG) to a destination, but that technique can be 96 difficult to balance and cannot provide a full path redundancy even 97 in the case of a biconnected graph. For instance, if the node that 98 is closest to the DAG destination has only one link to that 99 destination, then it does not have a alternate path to get to that 100 destination. 102 It is also possible to compute an alternate routing topology for fast 103 rerouting to a given destination, in which case some signalling, 104 tagging or labelling can be put in place to indicate whether a packet 105 follows the normal path or was rerouted over an alternate topology. 106 Once a packet is rerouted, it is bound to the alternate topology so 107 only one breakage can be handled with looplessness guarantees in most 108 practical situations. 110 This draft introduces the concept of an Available Routing Construct 111 (ARC) as a routing construct made of a bidirectional sequence of 112 nodes and links with 2 outgoing edges, so that, upon a single 113 breakage, each lively node in along ARC can still reach one of the 114 outgoing edges. 116 The routing graph to reach a certain destination is expressed as a 117 cascade of ARCs, each ARC providing its own independent domain of 118 fault isolation and recovery. Unicast traffic may enter an ARC via 119 any node but it may only leave the ARC through one of its two edges. 120 One node along the ARC is designated as the cursor. In normal 121 unicast operations, the traffic inside an ARC flows away from the 122 cursor towards an edge. Upon a failure, packets may bounce on the 123 breakage point and flow the other way along the ARC to take the other 124 exit. 126 Aa a result an ARC is resilient to any single failure, and the 127 recovery can be driven either from the data plane or the control 128 plane. A second failure occurring within a same ARC will isolate an 129 ARC segment. This can be further corrected from the control plane by 130 reversing all the incoming Edges in a process that might recurse till 131 an exit is found. When ARC reversal is applied, an ARC topology is 132 resilient to some cases of Shared Risk Link Group (SRLG) failures. 134 This draft presents the concept and provides an intuition of how ARCs 135 can simplify the operation and improve the network utilization and 136 resiliency for all sorts of traffic in multiple environments, but 137 defers to further documents to elaborate on the algorithms and 138 optimizations in the different application domains. 140 For instance, ARCs can also be used in datacenters for the purpose of 141 fast-reroute, or within a service provider network to simplify load 142 balancing operations or leverage optimally the ring topologies 143 [RFC5921]. An ARC topology can be flooded over itself and serve as a 144 backbone for reliable multicasting operations. 146 2. Terminology 148 The draft uses the following terminology: 150 ARC: Available Routing Construct. An ARC is a loopless ordered set 151 of nodes and links whereby traffic may enter via any node in the 152 ARC but may only leave the ARC through either one of the ARC 153 edges. 155 Comb: An ARC generalization: a Comb is a n-edged loopless set of 156 nodes and links with n >= 2; traffic may enter via any node in the 157 Comb but may only exit the Comb through one of its n edges. A 158 Comb comes with a walk operation that enables to attempt to exit 159 via every edge and to discover when all have been tried. 161 Cursor: A virtual point along an ARC that can be located on a node or 162 on a link between 2 nodes. In normal operations, the traffic 163 along the ARC flows away from its Cursor. If the cursor is a 164 node, then traffic can be distributed on both sides. The Cursor 165 may be moved to change the way traffic is load balanced along an 166 ARC. It may also be placed at the location of a failure to direct 167 traffic away from that point. 169 ARC Node: A Node that belongs to an ARC. 171 Edge ARC Node: An ARC Node at an edge of its ARC. An Edge ARC Node 172 is a node via wich traffic can exit the ARC. 174 Edge Link: A directed link outgoing from an Edge ARC Node. Traffic 175 can only exit from an ARC via an Edge Link. An Edge Link does not 176 accept traffic into an ARC. 178 Intermediate ARC Node: A node that is not at an edge of an ARC. A 179 Intermediate ARC Node node that can receive traffic and forward 180 traffic between its adjacent nodes. 182 Intermediate Link: A link between two Intermediate ARC Nodes. An 183 Intermediate Link is reversible, meaning that traffic is allowed 184 in both directions though an individual packet is constrained in 185 the way its direction is reversed. For stable links such as wired 186 links, the typical constraint is that the direction of a packet 187 may be reversed at most once along a given ARC. 189 Collapsed ARC: An ARC that is formed of a single node. This node is 190 altogether the cursor and both Edge Nodes. This implies that the 191 node has at least 2 outgoing links to 2 different Safe Nodes. 193 | 194 | 195 V 196 C+EAN 197 /|\ 198 / | \ 199 | V | 200 V V 202 E: Edge ARC Node -| collapsed in a single node 203 C: Cursor -| 205 Infrastructure ARC: An ARC that is formed of more than one node, 206 which also means that the Edge Nodes are different nodes. 208 | \ | | 209 | \ | | | 210 V V V | 211 _->IAN<---->IAN<---->IAN<---->IAN<-_ | 212 / + C \ | 213 / \| 214 V V 215 EAN EAN 216 | /|\ 217 | / | \ 218 | | V | 219 V V V 221 IAN: Intermediate ARC Node -| 222 EAN: Edge ARC Node |- All are Safe Nodes 223 C: Cursor -| 225 DAG: Directed Acyclic Graph. 227 ARC Set (or Cascade): A DAG with ARCs as vertices. In the DAG, an 228 edge between ARC A and ARC B corresponds to a link from an Edge 229 ARC Node in ARC A and an arbitrary ARC Node in ARC B. Note that by 230 definition, an ARC has at least 2 outgoing Edge Links, one per 231 Edge Node, and maybe more if an Edge Node has multiple outgoing 232 Edge Links. All vertices in the DAG have 2 forwarding solutions, 233 even the ARC closest to the destination. 235 Omega: the abstract destination (== root) of an ARC Set. 237 ARC Height: An arbitrary distance from Omega that is associated to an 238 ARC. The Height of an ARC must be more than the Height of any of 239 the ARCs it terminates into. The order of ARC formation by a 240 given algorithm can be used as a Height whereby an ARC is always 241 strictly higher or lower than another. 243 Buttressing ARC: A split ARC that is merged into another ARC at one 244 edge. An ARC and one or more Buttressing ARCs form a Comb 245 construct that is resilient to additional breakages. A 246 Buttressing ARC may be applied to an ARC or a Comb iff traffic 247 outgoing the Buttressing ARC Edge always reaches in an ARC that is 248 lower than this ARC, or Omega. 250 | \ | | 251 | \ | | | 252 V V V | 253 _->IAN<---->IAN<---->IAN<---->IAN<-_----->IAN<-_ 254 / + C \ | \ 255 / \| \ 256 V V V 257 EAN EAN EAN 258 | /|\ | 259 | / | \ | 260 | | V | | 261 V V V V 263 Safe Node: A node is Safe if there is no single point of failure - 264 apart from the node itself - on its way to Omega. From this 265 definition, a node is Safe if it has at least two non-congruent 266 paths to two different other Safe Nodes. It results that a Safe 267 node that is not Omega has at least two completely disjunct paths 268 to Omega. When an ARC has been successfully constructed, all its 269 nodes become safe with respect to the Omega for which the ARC was 270 constructed. By extension for a collapsed path Omega is deemed to 271 be Safe, that is any node that pertains in Omega is a Safe Node. 273 ?-S: A node N is deemed dependent on a node S or S-dependent (denoted 274 as ?-S) if S is the last single point of failure along N's 275 shortest path to Omega. 277 3. ARC Set representations 279 An ARC Set can be represented in a number of fashions: 281 Graph View: 283 H2<==>H<==>H1<---I--->J1<==>J--->K1<===>K 284 | | | | | 285 | | | | | 286 V V V V V 287 D1<==>D<==>D3 E1<==>E F1<==>F<==>F2 G 288 | | | | | | / \ 289 | | | | | | / \ 290 V V V V V V V V 291 B1<==>B2<==>B3<==>B--->A<==>A1<------C2<==>C<==>C4 292 | | | | 293 | | | | 294 | V V | 295 +--------------------> Omega <-------------------+ 297 This representation is similar to a classical routing graph with 298 the pecularity that some Links are marked reversible. An ARC is 299 represented as a sequence of reversible links. The node that 300 holds the cursor is also indicated somehow. 302 ARC View: 304 +========I========+ 305 | | 306 | +====J====+ 307 | | | 308 +====H====+ | +=====K=====+ 309 | | | | | 310 +====D====+ +====E====+ +====F====+ +====G====+ 311 | | | | | | | | 312 +=========B=========+ | | +=========C=========+ 313 | | | | | | 314 | +======A=======+ | 315 | | | | 316 ------------------------------------------------------------------Omega 318 This representation is similar to a classical routing graph with 319 the pecularity that some Links are marked reversible. An ARC is 320 represented as a sequence of reversible links. 322 Collapsed DAG view: 324 +====+ +====+ +====+ +====+ 325 | H | <--- | I | ---> | J | ---> | K | 326 | \__ | ___/ | 327 | \ | / | 328 V _| V |_ V 329 +====+ +====+ +====+ +====+ 330 | D | | E | | F | <--- | G | 331 \ \ __/ \__ __/ \__ / / 332 \ \ / \ / \ / / 333 _| _| |_ _| |_ _| |_ |_ 334 +====+ +====+ +====+ 335 | B | ---> | A | <--- | C | 336 | | | | 337 V V V V 338 ------------------------------------------------------------------Omega 339 A DAG representation whereby an ARC is abstracted as a vertice and 340 links between ARCs are shown as directed edges. This way, the 341 reversible links are omitted and the graph is simplified. It can 342 be noted that even the vertice closest to Omega has 2 non- 343 congruent forwarding solutions, that is Heir Links to Omega. 345 4. Applicability 347 This section has to be refined. ARCs probably apply to both unicast 348 and multicast and the authors expect further documents to explain how 349 that is done. The examples below are provided as an indication but 350 is not limiting the field of applications. 352 4.1. Load Balancing 354 In normal conditions, only the cursor may distribute its traffic 355 between the two Edge Nodes. If an Edge Node is still congested after 356 the cursor forwards all its traffic towards the other Edge Node, then 357 the cursor can be moved towards the congested Edge in order to derive 358 even more traffic towards the other Edge. If both Edges are 359 congested, then a backpressure can be applied on the incoming ARCs so 360 that they move their own traffic towards their own alternate Edge. 361 The process may recurse. 363 4.1.1. Routing Hierarchies 365 The ARC methods may be used to build and/or leverage routing 366 hierarchies, allowing high availability at multiple hierarchical 367 levels. In one hand, the view of an ARC Set can be simplified by 368 abstracting an ARC as a node in a DAG. The view of the routing 369 topology is thus simplified, as illustrated in Figure 6. In the 370 other hand, ARCs may be used inside a subtopology, such as a ring, to 371 enable forwarding inside a ring towards a next ring. Then, 372 abstracting a full ring as a node, ARCs can be applied to a graph of 373 rings, providing another level of redundancy and an abstract end to 374 end path computation that is represented as a cascade of ARCs of 375 rings. 377 5. Lowest ARC First 379 The open Lowest ARC First(oLAF) algorithm is presented below in such 380 a way as to help the reader figure how an ARC Set can be obtained but 381 not in a computer-optimized fashion that is left to be determined. 382 oLAF is based on Dijkstra's algorithm for Shortest Path First (SPF) 383 computation, and is designed in such a fashion that the reverse SPF 384 tree towards a destination is conserved and preferred for forwarding 385 along the resulting ARC Set. 387 We make the computation on behalf of Omega, that is an abstraction, 388 but could represent the node or the set of nodes that we want to 389 reach with an ARC Set. If Omega is instantiated as an actual 390 destination node, then that node may be a fine location for an ARC 391 Computing Engine. 393 5.1. Init 394 So we start with an proverbial Initial Set of Nodes that are 395 interconnected by Links, and Omega that is the destination that we 396 want to reach with an ARC Set. 398 If there is no Heir, we're done. If there is a single Heir then the 399 graph is monoconnected, so we restart the computation taking that 400 Heir off the Set of Nodes and making it Omega. 402 Else, if Omega is a single Node, or if Omega is composed of multiple 403 nodes but we are willing to accept that both ends of an ARC terminate 404 in a same node in Omega, then we create virtual Omega nodes, a 405 minimum of two and at most one per Heir, and we make them the new 406 Omega. Note: we need at least two destinations because both ends of 407 an ARC cannot terminate in a same node. 409 Now we can start building an ARC Set towards the resulting Omega. 411 In this process, we create so-called Dependent Sets of nodes, each 412 owned by a Safe Node S, DSet(S). DSet(S) contains nodes that are not 413 determined to be Safe at the current stage of the computation and for 414 which S, the owner Safe Node, is the last single point of failure on 415 the shortest path tree to Omega. It results that a given node can be 416 at most in one DSet, and that a Safe Node belongs to its own DSet. 418 For each node S in Omega we create a DSet(S) in which we place S. 420 5.2. Growing Trees 422 And then the process goes like this: 424 We select the node in the Set of Nodes that is closest to Omega using 425 the cost towards Omega as if we were building a traditional reverse 426 SPF tree and we place the selected node in the same Dependent Set as 427 its parent in the reverse SPF tree. Note that for a Heir, the parent 428 might be a real node in Omega, or a virtual Omega node. 430 If we kept it at that, we would be building subtrees that are hanging 431 off a Safe Node and together would represent the reverse shortest 432 path tree towards Omega, each subtree being grown separately inside 433 DSet(S) where S is the (virtual) Safe node that is the root of the 434 subtree. 436 5.3. Being Safe 438 But once we have placed the selected node in a DSet, we consider its 439 neighbors one by one. If at least one of the neighbors is already in 440 a different DSet than this node, we select the neighbor that provides 441 the shortest alternate path to Omega for the selected node. 443 Doing so, we have isolated two paths: 445 o one along its own shortest path that is contained within its own 446 Dependent Set and that leads to the owner Safe Node of this set. 448 o and one via the selected neighbor, along its own shortest path 449 within the selected neighbor's Dependent Set and that leads to the 450 owner Safe Node of that other set. 452 Because the two sets are different and have no intersection, these 453 paths are non-congruent. And because the two non-congruent paths 454 lead to two different Safe Nodes, this node is Safe. 456 It might happen that: 458 o the selected node's parent is already a Safe Node, in which case 459 the selected node is the Edge AN on its shortest path side. 461 o It might also happen that the selected neighbor is already a Safe 462 Node, in which case selected node is the Edge AN on its alternate 463 side. 465 If both conditions are met for a same AN, then that AN forms a 466 collapsed ARC by itself. 468 5.4. Bending An ARC 470 Now we form an ARC as follows: 472 o A height is attributed to this ARC that must be strictly more than 473 that of the ARCs it terminates into, if any. The order in which 474 the ARCs are built may be used in some cases. 476 o The ARC terminates in the two Safe Nodes that are the owners of 477 the two DSets. The normal behaviour is to make a Edge Link the 478 link to the Safe Node. 480 o If the Safe Node at one end forms a collapsed ARC by itself, it 481 may be absorbed in the ARC in order to build a multi-edged ARC. 483 o If one of the two Safe Nodes pertains in a ARC or a Comb construct 484 that is higher than the other end, then this ARC may be merged at 485 the Safe Node with its original ARC, in order to form a Comb 486 construct whereby this ARC is a Buttressing ARC of the Comb. The 487 resulting Comb conserves the height on the original ARC or Comb 488 that it extends. 490 o The ARC is built by adjoining the two non-congruent paths that we 491 isolated for the selected node. 493 o The selected node is the node farthest from Omega in the resulting 494 ARC, so we make it the cursor. 496 o The link between the selected node and the selected neighbor would 497 not have been used in a classical reverse SPF tree. Here, we have 498 determined that this link is in fact critical to connect 2 zones 499 of the network (the DSets) that can act as a back up for one 500 another in case of the failure of their respective single points 501 of failure (the Safe Nodes). 503 o Because the ARC can be used in both directions, each AN along the 504 ARC has two non-congruent paths to the Safe Nodes that the ARC 505 terminates into. So it is a Safe Node. We create individual 506 DSets for each AN and we move the AN to its own DSet. 508 5.5. Orienting Links 510 For each ARC Node along the ARC: 512 o any link (there can be zero for a collapsed ARC, one for an Edge 513 AN or two of them for a Intermediate AN) between this AN and a 514 next AN along this ARC is made an Intermediate Link, that is, 515 reversible. The normal direction, away from the cursor, preserves 516 the shortest path. 518 o If this AN is an Edge AN for this ARC, than all links off this 519 node that terminate in a Safe Node are made Edge Links, that is, 520 outgoing but not reversible. 522 o All the other links left undertermined. 524 The nodes left in the Dependent Sets but the owner Safe Node are 525 still not Safe. They are moved back to the original Set of Nodes to 526 enable forming additional ARCs which might depend on this ARC in the 527 ARC Set. 529 5.6. Looping or recursing 531 We are done processing the particular node we had picked in the 532 original Set of Nodes. If the Set of Nodes as it stands now is not 533 empty, we continue from Section 5.2. 535 If the Set of Nodes went empty, we are done with this pass and we 536 consider the Dependent Sets that we have put together. In a 537 biconnected graph, there should be one set per node and one node per 538 set, denoting that every node is a Safe Node. 540 If some portion of the graph is monoconnected, then each 541 monoconnected portion forms the Dependent Set of the Safe Node that 542 is its single point of failure. In order to be maximally redundant, 543 we need to form the ARCs again, within the Dependent Set. 545 To do so, we remove the Safe Node from the Dependent set and make it 546 Omega. We make the resulting DSet our Set of Nodes and run the 547 algorithm again. 549 This may recurse a number of times if the graph has monoconnected 550 zones within others. 552 6. Forwarding Along An ARC Set 554 Under normal conditions, the traffic flows away from the cursor of 555 the current ARC and cascades into the next ARC on that side of the 556 cursor, with the Height of the current ARC decreasing monotonically 557 from ARC to ARC till Omega is reached. 559 The same goes for a generic Comb construct. When Buttressing ARCs 560 are applied on a main ARC or other Buttressing ARCs, the final 561 construct assumes the shape of a tree. The tree may be walked in 562 different manners but the shortest path requires to start going down 563 the current ARC or Buttressing ARC to its Edge. 565 In case of Label forwarding, the same recursivity is applied and a 566 multiple ARC label path is constructed. Each ARC has is own set of 567 label path per Omega, each ARC Set label path being merged into the 568 lower ARC label set, thus at the interconnection point. At minimum, 569 ARC label path should be built from the cursor toward each edge, but 570 this would require label path recompilation upon cursor move, the 571 proposed approach is then to build for the normal flow to an Omega 572 one pair of label path from edge to edge. 574 As this label construct maps the ARC topology with local significant 575 label, the Label Distribution Protocol (LDP) could be reused to 576 announce label association to neighbors on the ARC. 578 Upon a breakage inside an ARC, until a corrective action takes place, 579 some traffic will be lost. The corrective action might be either 580 operated at the control plane or the data plane, if immediate action 581 and near-zero packet loss is required. 583 6.1. Control Plane Recovery 585 Upon a first breakage in an ARC, the cursor is moved to the breakage 586 point, either a node or a link, so that traffic flows away from the 587 cursor again. 589 Upon a second breakage within a same ARC, a segment of the ARC is now 590 isolated. Both breakage points become sinks till an additional 591 corrective action, such as modifying the ARC Set, takes place. All 592 incoming links in the isolated segment are blocked , causing the 593 traffic to exit at the other end of the incoming ARCs. 595 Blocking an Edge Link in the incoming ARC may create an isolated 596 segment in the incoming ARC as well if it is a second breakage there 597 too, or if both edges of the incoming ARC tterminate in the broken 598 segment. In that case the process recurses and the broken zone can 599 be determined as the collection of the isolated segments. 601 If a segment of an ARC is getting isolated by a dual failure but that 602 ARC segment has incoming Edges then the ARC can be reversed. This 603 reversal is done by reversing of all the incoming Edges, which become 604 outgoing. The segment that was isolated now benefits from multiple 605 exits in a loop free fashion. This process might in turn isolate a 606 segment of an ARC that was incoming and the process recurses and some 607 links flap. If a real exit exits the process will stabilize, but a 608 count to infinity must be put in place to avoid a permanent flapping 609 when a whole ARC Subset is physically isolated. One may consider 610 that this process is in fact the classical link reversal technique, 611 as applied to the DAG of ARCs. 613 6.2. Data Plane Recovery 615 Upon a breakage inside an ARC, it is possible in the data plane to 616 reverse the direction of -to turn- a given packet once along the ARC 617 so the packets exits over the other Edge Link. But in order to avoid 618 loops, it is undesirable to reverse the direction of a given packet a 619 second time. 621 Note that once a given packet leaves an ARC to enter the next, it is 622 free to bounce again in the next ARC. In other Words, the domain that 623 is impacted by a turn is limited to the current ARC itself; the ARC 624 forms the event horizon wherein the notion that a turn happened may 625 cause a loop. 627 So a local strategy must be put in place inside an ARC to allow a 628 given packet to bounce once upon a breakage, and get dropped upon a 629 second breakage. 631 In the case of IP packet forwarding, a packet can be tagged when it 632 bounces inside an ARC, or when it passes the cursor, for instance by 633 reserving a TOS bit for that purpose. When the packet bounces, the 634 bit is set and when the packet leaves the ARC, the bit is reset and 635 may be used again in the next ARC. In the generic case of a Comb, a 636 strategy must be put in place to walk the structure and drop a packet 637 that tries all the Edges. it attempts to pass the cursor twice in a 638 same direction, meaning that more than a full walk was already 639 accomplished. 641 In the case of MPLS forwarding, the same result can be achieved with 642 either 3 or 4 Labels Switched Paths (LSPs) along the ARC. 644 3-Labels method: In this case we lay a primary LSP from the cursoo to 645 the Edge in each direction, and a backup LSP Edge to Edge in each 646 direction. So a node along the way has three labels, one primary 647 and two backup, one in each direction. Should the primary path 648 fail, the packet can be placed along the backup LSP in the other 649 direction. We'll note that this method contrains the location of 650 the cursor. Should the cursor move, The primary LSPs have to be 651 recomputed, at a minimum between the old and the new location of 652 the cursor where the direction is reversed. 654 4-Labels method: In this case we have two primary and two backup LSPs 655 Edge to Edge in each direction. The labels are independent of the 656 location of the cursor, so the cursor can be moved in control 657 plane with no impact on labels. 659 7. Manageability 661 This specification describes a generic model. Protocols and 662 management will come later 664 8. IANA Considerations 666 This specification does not require IANA action. 668 9. Security Considerations 670 This specification is not found to introduce new security threat. 672 10. Acknowledgements 674 The authors wishes to thank Dirk Anteunis, Stewart Bryant, IJsbrand 675 Wijnands, George Swallow, Eric Osborne, Clarence Filsfils and Eric 676 Levy-Abegnoli for their participation and continuous support to the 677 work presented here. 679 11. References 681 11.1. Normative References 683 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 684 Requirement Levels", BCP 14, RFC 2119, March 1997. 686 11.2. Informative References 688 [RFC5921] Bocci, M., Bryant, S., Frost, D., Levrau, L. and L. 689 Berger, "A Framework for MPLS in Transport Networks", RFC 690 5921, July 2010. 692 Authors' Addresses 694 Pascal Thubert, editor 695 Cisco Systems, Inc 696 Building D 697 45 Allee des Ormes - BP1200 698 MOUGINS - Sophia Antipolis, 06254 699 FRANCE 701 Phone: +33 497 23 26 34 702 Email: pthubert@cisco.com 703 Patrice Bellagamba 704 Cisco Systems 705 214 Avenue des fleurs 706 Saint-Raphael, 83700 707 FRANCE 709 Phone: +33.6.1998.4346 710 Email: pbellaga@cisco.com