idnits 2.17.00 (12 Aug 2021) /tmp/idnits61059/draft-ietf-rift-applicability-08.txt: Checking boilerplate required by RFC 5378 and the IETF Trust (see https://trustee.ietf.org/license-info): ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/1id-guidelines.txt: ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/checklist : ---------------------------------------------------------------------------- No issues found here. Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year -- The document date (7 November 2021) is 195 days in the past. Is this intentional? Checking references for intended status: Informational ---------------------------------------------------------------------------- == Outdated reference: A later version (-15) exists of draft-ietf-rift-rift-13 Summary: 0 errors (**), 0 flaws (~~), 2 warnings (==), 1 comment (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 RIFT WG Yuehua. Wei, Ed. 3 Internet-Draft Zheng. Zhang 4 Intended status: Informational ZTE Corporation 5 Expires: 11 May 2022 Dmitry. Afanasiev 6 Yandex 7 P. Thubert 8 Cisco Systems 9 Jaroslaw. Kowalczyk 10 Orange Polska 11 7 November 2021 13 RIFT Applicability 14 draft-ietf-rift-applicability-08 16 Abstract 18 This document discusses the properties, applicability and operational 19 considerations of RIFT in different network scenarios. It intends to 20 provide a rough guide how RIFT can be deployed to simplify routing 21 operations in Clos topologies and their variations. 23 Status of This Memo 25 This Internet-Draft is submitted in full conformance with the 26 provisions of BCP 78 and BCP 79. 28 Internet-Drafts are working documents of the Internet Engineering 29 Task Force (IETF). Note that other groups may also distribute 30 working documents as Internet-Drafts. The list of current Internet- 31 Drafts is at https://datatracker.ietf.org/drafts/current/. 33 Internet-Drafts are draft documents valid for a maximum of six months 34 and may be updated, replaced, or obsoleted by other documents at any 35 time. It is inappropriate to use Internet-Drafts as reference 36 material or to cite them other than as "work in progress." 38 This Internet-Draft will expire on 11 May 2022. 40 Copyright Notice 42 Copyright (c) 2021 IETF Trust and the persons identified as the 43 document authors. All rights reserved. 45 This document is subject to BCP 78 and the IETF Trust's Legal 46 Provisions Relating to IETF Documents (https://trustee.ietf.org/ 47 license-info) in effect on the date of publication of this document. 48 Please review these documents carefully, as they describe your rights 49 and restrictions with respect to this document. Code Components 50 extracted from this document must include Simplified BSD License text 51 as described in Section 4.e of the Trust Legal Provisions and are 52 provided without warranty as described in the Simplified BSD License. 54 Table of Contents 56 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 3 57 2. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . 3 58 3. Problem Statement of Routing in Modern IP Fabric Fat Tree 59 Networks . . . . . . . . . . . . . . . . . . . . . . . . 5 60 4. Applicability of RIFT to Clos IP Fabrics . . . . . . . . . . 5 61 4.1. Overview of RIFT . . . . . . . . . . . . . . . . . . . . 5 62 4.2. Applicable Topologies . . . . . . . . . . . . . . . . . . 8 63 4.2.1. Horizontal Links . . . . . . . . . . . . . . . . . . 8 64 4.2.2. Vertical Shortcuts . . . . . . . . . . . . . . . . . 9 65 4.2.3. Generalizing to any Directed Acyclic Graph . . . . . 9 66 4.2.4. Reachability of Internal Nodes in the Fabric . . . . 11 67 4.3. Use Cases . . . . . . . . . . . . . . . . . . . . . . . . 11 68 4.3.1. Data Center Topologies . . . . . . . . . . . . . . . 11 69 4.3.2. Metro Fabrics . . . . . . . . . . . . . . . . . . . . 12 70 4.3.3. Building Cabling . . . . . . . . . . . . . . . . . . 13 71 4.3.4. Internal Router Switching Fabrics . . . . . . . . . . 13 72 4.3.5. CloudCO . . . . . . . . . . . . . . . . . . . . . . . 13 73 5. Operational Considerations . . . . . . . . . . . . . . . . . 15 74 5.1. South Reflection . . . . . . . . . . . . . . . . . . . . 16 75 5.2. Suboptimal Routing on Link Failures . . . . . . . . . . . 16 76 5.3. Black-Holing on Link Failures . . . . . . . . . . . . . . 18 77 5.4. Zero Touch Provisioning (ZTP) . . . . . . . . . . . . . . 19 78 5.5. Mis-cabling Examples . . . . . . . . . . . . . . . . . . 20 79 5.6. Positive vs. Negative Disaggregation . . . . . . . . . . 22 80 5.7. Mobile Edge and Anycast . . . . . . . . . . . . . . . . . 24 81 5.8. IPv4 over IPv6 . . . . . . . . . . . . . . . . . . . . . 26 82 5.9. In-Band Reachability of Nodes . . . . . . . . . . . . . . 26 83 5.10. Dual Homing Servers . . . . . . . . . . . . . . . . . . . 28 84 5.11. Fabric With A Controller . . . . . . . . . . . . . . . . 28 85 5.11.1. Controller Attached to ToFs . . . . . . . . . . . . 29 86 5.11.2. Controller Attached to Leaf . . . . . . . . . . . . 29 87 5.12. Internet Connectivity Within Underlay . . . . . . . . . . 29 88 5.12.1. Internet Default on the Leaf . . . . . . . . . . . . 30 89 5.12.2. Internet Default on the ToFs . . . . . . . . . . . . 30 90 5.13. Subnet Mismatch and Address Families . . . . . . . . . . 30 91 5.14. Anycast Considerations . . . . . . . . . . . . . . . . . 30 92 5.15. IoT Applicability . . . . . . . . . . . . . . . . . . . . 31 93 5.16. Key Management . . . . . . . . . . . . . . . . . . . . . 32 94 6. Security Considerations . . . . . . . . . . . . . . . . . . . 32 95 7. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 33 96 8. Contributors . . . . . . . . . . . . . . . . . . . . . . . . 33 97 9. Normative References . . . . . . . . . . . . . . . . . . . . 33 98 10. Informative References . . . . . . . . . . . . . . . . . . . 35 99 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 36 101 1. Introduction 103 This document discusses the properties and applicability of "Routing 104 in Fat Trees" [RIFT] in different deployment scenarios and highlights 105 the operational simplicity of the technology compared to traditional 106 routing solutions. It also documents special considerations when 107 RIFT is used with or without overlays and/or controllers, and how 108 RIFT identifies topology mis-cablings and reroutes around node and 109 link failures. 111 2. Terminology 113 Clos/Fat Tree: 115 This document uses the terms Clos and Fat Tree interchangeably 116 whereas it always refers to a folded spine-and-leaf topology with 117 possibly multiple Points of Delivery (PoDs) and one or multiple Top 118 of Fabric (ToF) planes. 120 Directed Acyclic Graph (DAG): 122 A finite directed graph with no directed cycles (loops). If links in 123 a Clos are considered as either being all directed towards the top or 124 vice versa, each of such two graphs is a DAG. 126 Disaggregation: 128 Process in which a node decides to advertise more specific prefixes 129 Southwards, either positively to attract the corresponding traffic, 130 or negatively to repel it. Disaggregation is performed to prevent 131 black-holing and suboptimal routing to the more specific prefixes. 133 TIE: 135 This is an acronym for a "Topology Information Element". TIEs are 136 exchanged between RIFT nodes to describe parts of a network such as 137 links and address prefixes. A TIE has always a direction and a type. 138 North TIEs (sometimes abbreviated as N-TIEs) are used when dealing 139 with TIEs in the northbound representation and South-TIEs (sometimes 140 abbreviated as S-TIEs) for the southbound equivalent. TIEs have 141 different types such as node and prefix TIEs. 143 Node TIE: 145 This stands as acronym for a "Node Topology Information Element", 146 which contains all adjacencies the node discovered and information 147 about the node itself. Node TIE should NOT be confused with a North 148 TIE since "node" defines the type of TIE rather than its direction. 149 Consequently North Node TIEs and South Node TIEs exist. 151 Prefix TIE: 153 This is an acronym for a "Prefix Topology Information Element" and it 154 contains all prefixes directly attached to this node in case of a 155 North TIE and in case of South TIE the necessary default routes and 156 disaggregated routes the node advertises southbound. 158 South Reflection: 160 Often abbreviated just as "reflection", it defines a mechanism where 161 South Node TIEs are "reflected" from the level south back up north to 162 allow nodes in the same level without East- West links to "see" each 163 other's node Topology Information Elements (TIEs). 165 LIE: 167 This is an acronym for a "Link Information Element" exchanged on all 168 the system's links running RIFT to form ThreeWay adjacencies and 169 carry information used to perform Zero Touch Provisioning (ZTP) of 170 levels. 172 Shortest-Path First (SPF): 174 A well-known graph algorithm attributed to Dijkstra that establishes 175 a tree of shortest paths from a source to destinations on the graph. 176 SPF acronym is used due to its familiarity as general term for the 177 node reachability calculations that RIFT can employ to ultimately 178 calculate routes of which Dijkstra algorithm is a possible one. 180 North SPF (N-SPF): 182 A reachability calculation that is progressing northbound, as example 183 SPF that is using South Node TIEs only. Normally it progresses a 184 single hop only and installs default routes. 186 South SPF (S-SPF): 188 A reachability calculation that is progressing southbound, as example 189 SPF that is using North Node TIEs only. 191 3. Problem Statement of Routing in Modern IP Fabric Fat Tree Networks 193 Clos [CLOS] topologies (called commonly a fat tree/network in modern 194 IP fabric considerations as homonym to the original definition of the 195 term Fat Tree [FATTREE]) have gained prominence in today's 196 networking, primarily as a result of the paradigm shift towards a 197 centralized data-center based architecture that deliver a majority of 198 computation and storage services. 200 Current routing protocols were geared towards a network with an 201 irregular topology with isotropic properties, and low degree of 202 connectivity. When applied to Fat Tree topologies: 204 * They tend to need extensive configuration or provisioning during 205 bring up and adding or removing Rift nodes from the fabric. 207 * All nodes including spine and leaf nodes learn the entire network 208 topology and routing information, which is in fact, not needed on 209 the leaf nodes during normal operation. 211 * They flood significant amounts of duplicate link state information 212 between spine and leaf nodes during topology updates and 213 convergence events, requiring that additional CPU and link 214 bandwidth be consumed. This may impact the stability and 215 scalability of the fabric, make the fabric less reactive to 216 failures, and prevent the use of cheaper hardware at the lower 217 levels (i.e. spine and leaf nodes). 219 4. Applicability of RIFT to Clos IP Fabrics 221 Further content of this document assumes that the reader is familiar 222 with the terms and concepts used in OSPF [RFC2328] and IS-IS 223 [ISO10589-Second-Edition] link-state protocols. The sections of RIFT 224 [RIFT] outline the requirements of routing in IP fabrics and RIFT 225 protocol concepts. 227 4.1. Overview of RIFT 229 RIFT is a dynamic routing protocol that is tailored for use in Clos, 230 Fat-Tree, and other anisotropic topologies. A core property of RIFT 231 is that its operation is sensitive to the structure of the fabric - 232 it is anisotropic. RIFT acts as a link-state protocol when "pointing 233 north" - advertising southwards routes to northwards peer routers 234 (parents) through flooding and database synchronization- but operates 235 hop-by-hop like a distance-vector protocol when "pointing south" - 236 typically advertising a fabric default route directed towards the Top 237 of Fabric (ToF, aka superspine) to southwards peer routers 238 (children). 240 The fabric default is typically the default route, as described in 241 Section 4.2.3.8 "Southbound Default Route Origination" of RIFT 242 [RIFT]. The ToF nodes may alternatively originate more specific 243 prefixes (P') southbound instead of the default route. In such a 244 scenario, all addresses carried within the RIFT domain must be 245 contained within P', and it is possible for a leaf that acts as 246 gateway to the internet to advertise the default route instead. 248 RIFT floods flat link-state information northbound only so that each 249 level obtains the full topology of levels south of it. That 250 information is never flooded east-west or back south again. So a top 251 tier node has full set of prefixes from the Shortest Path First (SPF) 252 calculation. 254 In the southbound direction, the protocol operates like a "fully 255 summarizing, unidirectional" path-vector protocol or rather a 256 distance-vector with implicit split horizon. Routing information, 257 normally just the default route, propagates one hop south and is "re- 258 advertised" by nodes at next lower level. 260 +---------------+ +----------------+ 261 | ToF | | ToF | LEVEL 2 262 + ++------+--+--+-+ ++-+--+----+-----+ 263 | | | | | | | | | ^ 264 + | | | +-------------------------+ | 265 Distance | +-------------------+ | | | | | 266 Vector | | | | | | | | + 267 South | | | | +--------+ | | | Link+State 268 + | | | | | | | | Flooding 269 | | | +----------------+ | | | North 270 v | | | | | | | | + 271 ++---+-+ +------+ +-+----+ ++----++ | 272 |SPINE | |SPINE | | SPINE| | SPINE| | LEVEL 1 273 + ++----++ ++---+-+ +-+--+-+ ++----++ | 274 + | | | | | | | | | ^ N 275 Distance | +-------+ | | +--------+ | | | E 276 Vector | | | | | | | | | +------> 277 South | +-------+ | | | +------+ | | | | 278 + | | | | | | | | | + 279 v ++--++ +-+-++ ++--++ ++--++ + 280 |LEAF| |LEAF| |LEAF| |LEAF| LEVEL 0 281 +----+ +----+ +----+ +----+ 283 Figure 1: RIFT overview 285 A spine node has only information necessary for its level, which is 286 all destinations south of the node based on SPF calculation, default 287 route, and potential disaggregated routes. 289 RIFT combines the advantage of both link-state and distance-vector: 291 * Fastest possible convergence 293 * Automatic detection of topology 295 * Minimal routes/info on Top-of-Rack (ToR) switches, aka leaf nodes 297 * High degree of ECMP 299 * Fast de-commissioning of nodes 301 * Maximum propagation speed with flexible prefixes in an update 303 So there are two types of link-state database which are "north 304 representation" North Topology Information Elements (N-TIEs) and 305 "south representation" South Topology Information Elements (S-TIEs). 306 The N-TIEs contain a link-state topology description of lower levels 307 and S-TIEs carry simply default and disaggregated routes for the 308 lower levels. 310 RIFT also eliminates major disadvantages of link-state and distance- 311 vector with: 313 * Reduced and balanced flooding 315 * Level constrained automatic neighbor detection 317 To achieve this, RIFT builds on the art of IGPs, not only OSPF and 318 IS-IS but also MANET and IoT, to provide unique features: 320 * Automatic (positive or negative) route disaggregation of 321 northwards routes upon fallen leaves 323 * Recursive operation in the case of negative route disaggregation 325 * Anisotropic routing that extends a principle seen in RPL [RFC6550] 326 to wide superspines 328 * Optimal flooding reduction that derives from the concept of a 329 "multipoint relay" (MPR) found in OLSR [RFC3626] and balances the 330 flooding load over northbound links and nodes. 332 Additional advantages that are unique to RIFT are listed below, the 333 details of which can be found in RIFT [RIFT]. 335 * True ZTP(Zero Touch Provisioning) 337 * Minimal blast radius on failures 339 * Can utilize all paths through fabric without looping 341 * Simple leaf implementation that can scale down to servers 343 * Key-Value store 345 * Horizontal links used for protection only 347 4.2. Applicable Topologies 349 Albeit RIFT is specified primarily for "proper" Clos or Fat Tree 350 topologies, the protocol natively supports Points of Delivery (PoD) 351 concepts, which, strictly speaking, are not found in the original 352 Clos concept. 354 Further, the specification explains and supports operations of multi- 355 plane Clos variants where the protocol recommends the use of inter- 356 plane rings at the Top-of-Fabric level to allow the reconciliation of 357 topology view of different planes to make the negative disaggregation 358 viable in case of failures within a plane. These observations hold 359 not only in case of RIFT but also in the generic case of dynamic 360 routing on Clos variants with multiple planes and failures in bi- 361 sectional bandwidth, especially on the leafs. 363 4.2.1. Horizontal Links 365 RIFT is not limited to pure Clos divided into PoD and multi-planes 366 but supports horizontal (East-West) links below the top of fabric 367 level. Those links are used only for last resort northbound 368 forwarding when a spine loses all its northbound links or cannot 369 compute a default route through them. 371 A full-mesh connectivity between nodes on the same level can be 372 employed and that allows N-SPF to provide for any node losing all its 373 northbound adjacencies (as long as any of the other nodes in the 374 level are northbound connected) to still participate in northbound 375 forwarding. 377 Note that a "ring" of horizontal links at any level below ToF does 378 not provide a "ring-based protection" scheme since the SPF 379 computation would have to deal necessarily with breaking of "loops" 380 in Dijkstra sense--an application for which RIFT is not intended. 382 4.2.2. Vertical Shortcuts 384 Through relaxations of the specified adjacency forming rules, RIFT 385 implementations can be extended to support vertical "shortcuts". The 386 RIFT specification itself does not provide the exact details since 387 the resulting solution suffers from either much larger blast radius 388 with increased flooding volumes or in case of maximum aggregation 389 routing, bow-tie problems. 391 4.2.3. Generalizing to any Directed Acyclic Graph 393 RIFT is an anisotropic routing protocol, meaning that it has a sense 394 of direction (northbound, southbound, east-west) and that it operates 395 differently depending on the direction. 397 * Northbound, RIFT operates as a link-state protocol, whereby the 398 control packets are reflooded first all the way north and only 399 interpreted later. All the individual fine grained routes are 400 advertised. 402 * Southbound, RIFT operates as a distance-vector protocol, whereby 403 the control packets are flooded only one-hop, interpreted, and the 404 consequence of that computation is what gets flooded one more hop 405 south. In the most common use-cases, a ToF node can reach most of 406 the prefixes in the fabric. If that is the case, the ToF node 407 advertises the fabric default and negatively disaggregates the 408 prefixes that it cannot reach. On the other hand, a ToF node that 409 can reach only a small subset of the prefixes in the fabric will 410 preferably advertise those prefixes and refrain from aggregating. 412 In the general case, what gets advertised south are: 414 1. A fabric default that aggregates all the prefixes that are 415 reachable within the fabric, and that could be a default route 416 or a prefix that is dedicated to this particular fabric. 418 2. The loopback addresses of the northbound nodes, e.g., for 419 inband management. 421 3. The disaggregated prefixes for the dynamic exceptions to the 422 fabric default, advertised to route around the black hole that 423 may form. 425 * East-West routing can optionally be used, with specific 426 restrictions. It is used when a sibling has access to the fabric 427 default but this node does not. 429 Since a Directed Acyclic Graph (DAG) provides a sense of north (the 430 direction of the DAG) and of south (the reverse), it can be used to 431 apply RIFT--an edge in the DAG that has only incoming vertices is a 432 ToF node. 434 There are a number of caveats though: 436 * The DAG structure must exist before RIFT starts, so there is a 437 need for a companion protocol to establish the logical DAG 438 structure. 440 * A generic DAG does not have a sense of east and west. The 441 operation specified for east-west links and the southbound 442 reflection between nodes are not applicable. Also ZTP(Zero Touch 443 Provisioning) will derive a sense of depth that will eliminate 444 some links. Variations of ZTP(Zero Touch Provisioning) could be 445 derived to meet specific objectives, e.g., make it so that most 446 routers have at least 2 parents to reach the ToF. 448 * RIFT applies to any Destination-Oriented DAG (DODAG) where there's 449 only one ToF node and the problem of disaggregation does not 450 exist. In that case, RIFT operates very much like RPL [RFC6550], 451 but using Link State for southbound routes (downwards in RPL's 452 terms). For an arbitrary DAG with multiple destinations (ToFs) 453 the way disaggregation happens has to be considered. 455 * Positive disaggregation expects that most of the ToF nodes reach 456 most of the leaves, so disaggregation is the exception as opposed 457 to the rule. When this is no more true, it makes sense to turn 458 off disaggregation and route between the ToF nodes over a ring, a 459 full mesh, transit network, or a form of area zero. There again, 460 this operation is similar to RPL operating as a single DODAG with 461 a virtual root. 463 * In order to aggregate and disaggregate routes, RIFT requires that 464 all the ToF nodes share the full knowledge of the prefixes in the 465 fabric. 467 * This can be achieved with a ring as suggested by "RIFT" [RIFT], by 468 some preconfiguration, or using a synchronization with a common 469 repository where all the active prefixes are registered. 471 4.2.4. Reachability of Internal Nodes in the Fabric 473 RIFT does not require that nodes have reachable addresses in the 474 fabric, though it is clearly desirable for operational purposes. 475 Under normal operating conditions this can be easily achieved by 476 injecting the node's loopback address into North and South Prefix 477 TIEs or other implementation specific mechanisms. 479 Special considerations arise when a node loses all northbound 480 adjacencies, but is not at the top of the fabric. If a spine node 481 loses all northbound links, the spine node doesn't advertise default 482 route. But if the level of the spine node is auto-determined by ZTP, 483 it will "fall down" as despicted in Figure 8. 485 4.3. Use Cases 487 4.3.1. Data Center Topologies 489 4.3.1.1. Data Center Fabrics 491 RIFT is suited for applying in data center (DC) IP fabrics underlay 492 routing, vast majority of which seem to be currently (and for the 493 foreseeable future) Clos architectures. It significantly simplifies 494 operation and deployment of such fabrics as described in Section 5 495 for environments compared to extensive proprietary provisioning and 496 operational solutions. 498 4.3.1.2. Adaptations to Other Proposed Data Center Topologies 499 . +-----+ +-----+ 500 . | | | | 501 .+-+ S0 | | S1 | 502 .| ++---++ ++---++ 503 .| | | | | 504 .| | +------------+ | 505 .| | | +------------+ | 506 .| | | | | 507 .| ++-+--+ +--+-++ 508 .| | | | | 509 .| | A0 | | A1 | 510 .| +-+--++ ++---++ 511 .| | | | | 512 .| | +------------+ | 513 .| | +-----------+ | | 514 .| | | | | 515 .| +-+-+-+ +--+-++ 516 .+-+ | | | 517 . | L0 | | L1 | 518 . +-----+ +-----+ 520 Figure 2: Level Shortcut 522 RIFT is not strictly limited to Clos topologies. The protocol only 523 requires a sense of "compass rose directionality" either achieved 524 through configuration or derivation of levels. So, conceptually, 525 shortcuts between levels could be included. Figure 2 depicts an 526 example of a shortcut between levels. In this example, sub-optimal 527 routing will occur when traffic is sent from L0 to L1 via S0's 528 default route and back down through A0 or A1. In order to avoid 529 that, only default routes from A0 or A1 are used, all leaves would be 530 required to install each others routes. 532 While various technical and operational challenges may require the 533 use of such modifications, discussion of those topics are outside the 534 scope of this document. 536 4.3.2. Metro Fabrics 538 The demand for bandwidth is increasing steadily, driven primarily by 539 environments close to content producers (server farms connection via 540 DC fabrics) but in proximity to content consumers as well. Consumers 541 are often clustered in metro areas with their own network 542 architectures that can benefit from simplified, regular Clos 543 structures and hence from RIFT. 545 4.3.3. Building Cabling 547 Commercial edifices are often cabled in topologies that are either 548 Clos or its isomorphic equivalents. The Clos can grow rather high 549 with many levels. That presents a challenge for traditional routing 550 protocols (except BGP and by now largely phased-out PNNI) which do 551 not support an arbitrary number of levels which RIFT does naturally. 552 Moreover, due to the limited sizes of forwarding tables in network 553 elements of building cabling, the minimum FIB size RIFT maintains 554 under normal conditions is cost-effective in terms of hardware and 555 operational costs. 557 4.3.4. Internal Router Switching Fabrics 559 It is common in high-speed communications switching and routing 560 devices to use fabrics when a crossbar is not feasible due to cost, 561 head-of-line blocking or size trade-offs. Normally such fabrics are 562 not self-healing or rely on 1:/+1 protection schemes but it is 563 conceivable to use RIFT to operate Clos fabrics that can deal 564 effectively with interconnections or subsystem failures in such 565 module. RIFT is not IP specific and hence any link addressing 566 connecting internal device subnets is conceivable. 568 4.3.5. CloudCO 570 The Cloud Central Office (CloudCO) is a new stage of telecom Central 571 Office. It takes the advantage of Software Defined Networking (SDN) 572 and Network Function Virtualization (NFV) in conjunction with general 573 purpose hardware to optimize current networks. The following figure 574 illustrates this architecture at a high level. It describes a single 575 instance or macro-node of cloud CO that provides a number of Value 576 Added Services (VAS), a Broadband Access Abstraction (BAA), and 577 virtualized nerwork services. An Access I/O module faces a Cloud CO 578 access node, and the Customer Premises Equipments (CPEs) behind it. 579 A Network I/O module is facing the core network. The two I/O modules 580 are interconnected by a leaf and spine fabric [TR-384]. 582 +---------------------+ +----------------------+ 583 | Spine | | Spine | 584 | Switch | | Switch | 585 +------+---+------+-+-+ +--+-+-+-+-----+-------+ 586 | | | | | | | | | | | | 587 | | | | | +-------------------------------+ | 588 | | | | | | | | | | | | 589 | | | | +-------------------------+ | | | 590 | | | | | | | | | | | | 591 | | +----------------------+ | | | | | | | | 592 | | | | | | | | | | | | 593 | +---------------------------------+ | | | | | | | 594 | | | | | | | | | | | | 595 | | | +-----------------------------+ | | | | | 596 | | | | | | | | | | | | 597 | | | | | +--------------------+ | | | | 598 | | | | | | | | | | | | 599 +--+ +-+---+--+ +-+---+--+ +--+----+--+ +-+--+--+ +--+ 600 |L | | Leaf | | Leaf | | Leaf | | Leaf | |L | 601 |S | | Switch | | Switch | | Switch | | Switch| |S | 602 ++-+ +-+-+-+--+ +-+-+-+--+ +--+-+--+--+ ++-+--+-+ +-++ 603 | | | | | | | | | | | | | | 604 | +-+-+-+--+ +-+-+-+--+ +--+-+--+--+ ++-+--+-+ | 605 | |Compute | |Compute | | Compute | |Compute| | 606 | |Node | |Node | | Node | |Node | | 607 | +--------+ +--------+ +----------+ +-------+ | 608 | || VAS5 || || vDHCP|| || vRouter|| ||VAS1 || | 609 | |--------| |--------| |----------| |-------| | 610 | |--------| |--------| |----------| |-------| | 611 | || VAS6 || || VAS3 || || v802.1x|| ||VAS2 || | 612 | |--------| |--------| |----------| |-------| | 613 | |--------| |--------| |----------| |-------| | 614 | || VAS7 || || VAS4 || || vIGMP || ||BAA || | 615 | |--------| |--------| |----------| |-------| | 616 | +--------+ +--------+ +----------+ +-------+ | 617 | | 618 ++-----------+ +---------++ 619 |Network I/O | |Access I/O| 620 +------------+ +----------+ 622 Figure 3: An example of CloudCO architecture 624 The Spine-Leaf architecture deployed inside CloudCO meets the network 625 requirements of adaptable, agile, scalable and dynamic. 627 5. Operational Considerations 629 RIFT presents the opportunity for organizations building and 630 operating IP fabrics to simplify their operation and deployments 631 while achieving many desirable properties of a dynamic routing on 632 such a substrate: 634 * RIFT only floods routing information to the devices that 635 absolutely need it. RIFT design follows minimum blast radius and 636 minimum necessary epistemological scope philosophy which leads to 637 good scaling properties while delivering maximum reactiveness. 639 * RIFT allows for extensive Zero Touch Provisioning within the 640 protocol. In its most extreme version RIFT does not rely on any 641 specific addressing and for IP fabric can operate using IPv6 ND 642 [RFC4861] only. 644 * RIFT has provisions to detect common IP fabric mis-cabling 645 scenarios. 647 * RIFT negotiates automatically BFD per link. This allows for IP 648 and micro-BFD [RFC7130] to replace Link Aggregation Groups (LAGs) 649 which do hide bandwidth imbalances in case of constituent 650 failures. Further automatic link validation techniques similar to 651 [RFC5357] could be supported as well. 653 * RIFT inherently solves many difficult problems associated with the 654 use of traditional routing topologies with dense meshes and high 655 degrees of ECMP by including automatic bandwidth balancing, flood 656 reduction and automatic disaggregation on failures while providing 657 maximum aggregation of prefixes in default scenarios. 659 * RIFT reduces FIB size towards the bottom of the IP fabric where 660 most nodes reside and allows with that for cheaper hardware on the 661 edges and introduction of modern IP fabric architectures that 662 encompass e.g. server multi-homing. 664 * RIFT provides valley-free routing and with that is loop free. 665 This allows the use of any such valley-free path in bi-sectional 666 fabric bandwidth between two destination irrespective of their 667 metrics which can be used to balance load on the fabric in 668 different ways. 670 * RIFT includes a key-value distribution mechanism which allows for 671 many future applications such as automatic provisioning of basic 672 overlay services or automatic key roll-overs over whole fabrics. 674 * RIFT is designed for minimum delay in case of prefix mobility on 675 the fabric. In conjunction with [RFC8505], RIFT can differentiate 676 anycast advertisements from mobility events and retain only the 677 most recent advertisement in the latter case. 679 * Many further operational and design points collected over many 680 years of routing protocol deployments have been incorporated in 681 RIFT such as fast flooding rates, protection of information 682 lifetimes and operationally easily recognizable remote ends of 683 links and node names. 685 5.1. South Reflection 687 South reflection is a mechanism that South Node TIEs are "reflected" 688 back up north to allow nodes in same level without east-west links to 689 "see" each other. 691 For example, Spine111\Spine112\Spine121\Spine122 reflects Node S-TIEs 692 from ToF21 to ToF22 separately. Respectively, 693 Spine111\Spine112\Spine121\Spine122 reflects Node S-TIEs from ToF22 694 to ToF21 separately. So ToF22 and ToF21 see each other's node 695 information as level 2 nodes. 697 In an equivalent fashion, as the result of the south reflection 698 between Spine121-Leaf121-Spine122 and Spine121-Leaf122-Spine122, 699 Spine121 and Spine 122 knows each other at level 1. 701 5.2. Suboptimal Routing on Link Failures 702 +--------+ +--------+ 703 | ToF21 | | ToF22 | LEVEL 2 704 ++--+-+-++ ++-+--+-++ 705 | | | | | | | + 706 | | | | | | | linkTS8 707 +------------+ | +-+linkTS3+-+ | | | +-------------+ 708 | | | | | | + | 709 | +---------------------------+ | linkTS7 | 710 | | | | + + + | 711 | | | +-------+linkTS4+------------+ | 712 | | | + + | | | 713 | | | +-------------+--+ | | 714 | | | | | linkTS6 | | 715 +-+----+-+ +-+----+-+ ++--------+ +-+----+-+ 716 |Spine111| |Spine112| |Spine121 | |Spine122| LEVEL 1 717 +-+---+--+ +-+----+-+ +-+---+---+ +-+----+-+ 718 | | | | | | | | 719 | +-------------+ | + ++XX+linkSL6+---+ + 720 | | | | linkSL5 | | linkSL8 721 | +-----------+ | | + +---+linkSL7+-+ | + 722 | | | | | | | | 723 +-+---+-+ +--+--+-+ +-+---+-+ +--+--+-+ 724 |Leaf111| |Leaf112| |Leaf121| |Leaf122| LEVEL 0 725 +-+-----+ +-+-----+ +-----+-+ +-+-----+ 726 + + + + 727 Prefix111 Prefix112 Prefix121 Prefix122 729 Figure 4: Suboptimal routing upon link failure use case 731 As shown in Figure 4, as the result of the south reflection between 732 Spine121-Leaf121-Spine122 and Spine121-Leaf122-Spine122, Spine121 and 733 Spine 122 knows each other at level 1. 735 Without disaggregation mechanism, when linkSL6 fails, the packet from 736 leaf121 to prefix122 will probably go up through linkSL5 to linkTS3 737 then go down through linkTS4 to linkSL8 to Leaf122 or go up through 738 linkSL5 to linkTS6 then go down through linkTS8 and linkSL8 to 739 Leaf122 based on pure default route. It's the case of suboptimal 740 routing or bow-tieing. 742 With disaggregation mechanism, when linkSL6 fails, Spine122 will 743 detect the failure according to the reflected node S-TIE from 744 Spine121. Based on the disaggregation algorithm provided by RIFT, 745 Spine122 will explicitly advertise prefix122 in Disaggregated Prefix 746 S-TIE PrefixesElement(prefix122, cost 1). The packet from leaf121 to 747 prefix122 will only be sent to linkSL7 following a longest-prefix 748 match to prefix 122 directly then go down through linkSL8 to Leaf122 749 . 751 5.3. Black-Holing on Link Failures 753 +--------+ +--------+ 754 | ToF 21 | | ToF 22 | LEVEL 2 755 ++-+--+-++ ++-+--+-++ 756 | | | | | | | + 757 | | | | | | | linkTS8 758 +--------------+ | +-+linkTS3+X+ | | | +--------------+ 759 linkTS1 | | | | | + | 760 + +-----------------------------+ | linkTS7 | 761 | | + | + + + | 762 | | linkTS2 +-------+linkTS4+X+----------+ | 763 | + + + + | | | 764 | linkTS5 +-+ +------------+--+ | | 765 | + | | | linkTS6 | | 766 +-+----+-+ +-+----+-+ ++-------+ +-+-----++ 767 |Spine111| |Spine112| |Spine121| |Spine122| LEVEL 1 768 +-+---+--+ ++----+--+ +-+---+--+ +-+----+-+ 769 | | | | | | | | 770 + +---------------+ | + +---+linkSL6+---+ + 771 linkSL1 | | | linkSL5 | | linkSL8 772 + +--+linkSL3+--+ | | + +---+linkSL7+-+ | + 773 | | | | | | | | 774 +-+---+-+ +--+--+-+ +-+---+-+ +--+--+-+ 775 |Leaf111| |Leaf112| |Leaf121| |Leaf122| LEVEL 0 776 +-+-----+ +-+-----+ +-----+-+ +-----+-+ 777 + + + + 778 Prefix111 Prefix112 Prefix121 Prefix122 780 Figure 5: Black-holing upon link failure use case 782 This scenario illustrates a case when double link failure occurs and 783 with that black-holing can happen. 785 Without disaggregation mechanism, when linkTS3 and linkTS4 both fail, 786 the packet from leaf111 to prefix122 would suffer 50% black-holing 787 based on pure default route. The packet supposed to go up through 788 linkSL1 to linkTS1 then go down through linkTS3 or linkTS4 will be 789 dropped. The packet supposed to go up through linkSL3 to linkTS2 790 then go down through linkTS3 or linkTS4 will be dropped as well. 791 It's the case of black-holing. 793 With disaggregation mechanism, when linkTS3 and linkTS4 both fail, 794 ToF22 will detect the failure according to the reflected node S-TIE 795 of ToF21 from Spine111\Spine112. Based on the disaggregation 796 algorithm provided by RIFT, ToF22 will explicitly originate an S-TIE 797 with prefix 121 and prefix 122, that is flooded to spines 111, 112, 798 121 and 122. 800 The packet from leaf111 to prefix122 will not be routed to linkTS1 or 801 linkTS2. The packet from leaf111 to prefix122 will only be routed to 802 linkTS5 or linkTS7 following a longest-prefix match to prefix122. 804 5.4. Zero Touch Provisioning (ZTP) 806 RIFT is designed to require a very minimal configuration to simplify 807 its operation and avoid human errors; based on that minimal 808 information, Zero Touch Provisioning (ZTP) autoconfigures the key 809 operational parameters of all the RIFT nodes, including the SystemID 810 of the node that must be unique in the RIFT network and the level of 811 the node in the Fat Tree, which determines which peers are northwards 812 "parents" and which are southwards "children". 814 ZTP is always on, but its decisions can be overridden when a network 815 administrator prefers to impose its own configuration. In that case, 816 it is the responsibility of the administrator to ensure that the 817 configured parameters are correct, in other words that the SystemID 818 of each node is unique, and that the administratively set levels 819 truly reflect the relative position of the nodes in the fabric. It 820 is recommended to let ZTP configure the network, and when not, it is 821 recommended to configure the level of all the nodes to avoid an 822 undesirable interaction between ZTP and the manual configuration. 824 ZTP requires that the administrator points out the Top-of-Fabric 825 (ToF) nodes to set the baseline from which the fabric topology is 826 derived. The Top-of-Fabric nodes are configured with TOP_OF_FABRIC 827 flag which are initial 'seeds' needed for other ZTP nodes to derive 828 their level in the topology. ZTP computes the level of each node 829 based on the Highest Available Level (HAL) of the potential parent(s) 830 nearest that baseline, which represents the superspine. In a 831 fashion, RIFT can be seen as a distance-vector protocol that computes 832 a set of feasible successors towards the superspine and auto- 833 configures the rest of the topology. 835 The autoconfiguration mechanism computes a global maximum of levels 836 by diffusion. The derivation of the level of each node happens then 837 based on Link Information Elements (LIEs) received from its neighbors 838 whereas each node (with possibly exceptions of configured leaves) 839 tries to attach at the highest possible point in the fabric. This 840 guarantees that even if the diffusion front reaches a node from 841 "below" faster than from "above", it will greedily abandon already 842 negotiated level derived from nodes topologically below it and 843 properly peer with nodes above. 845 The achieved equilibrium can be disturbed massively by all nodes with 846 highest level either leaving or entering the domain (with some finer 847 distinctions not explained further). It is therefore recommended 848 that each node is multi-homed towards nodes with respective HAL 849 offerings. Fortunately, this is the natural state of things for the 850 topology variants considered in RIFT. 852 A RIFT node may also be configured to confine it to the leaf role 853 with the LEAF_ONLY flag. A leaf node can also be configured to 854 support leaf-2-leaf procedures with the LEAF_2_LEAF flag. In either 855 case the node cannot be TOP_OF_FABRIC and its level cannot be 856 configured. RIFT will fully determine the node's level after it is 857 attached to the topology and ensure that the node is at the "bottom 858 of the hierarchy" (southernmost). 860 5.5. Mis-cabling Examples 862 +----------------+ +-----------------+ 863 | ToF21 | +------+ ToF22 | LEVEL 2 864 +-------+----+---+ | +----+---+--------+ 865 | | | | | | | | | 866 | | | +----------------------------+ | 867 | +---------------------------+ | | | | 868 | | | | | | | | | 869 | | | | +-----------------------+ | | 870 | | +------------------------+ | | | 871 | | | | | | | | | 872 +-+---+--+ +-+---+--+ | +--+---+-+ +--+---+-+ 873 |Spine111| |Spine112| | |Spine121| |Spine122| LEVEL 1 874 +-+---+--+ ++----+--+ | +--+---+-+ +-+----+-+ 875 | | | | | | | | | 876 | +---------+ | link-M | +---------+ | 877 | | | | | | | | | 878 | +-------+ | | | | +-------+ | | 879 | | | | | | | | | 880 +-+---+-+ +--+--+-+ | +-+---+-+ +--+--+-+ 881 |Leaf111| |Leaf112+-----+ |Leaf121| |Leaf122| LEVEL 0 882 +-------+ +-------+ +-------+ +-------+ 884 Figure 6: A single plane mis-cabling example 886 Figure 6 shows a single plane mis-cabling example. It's a perfect 887 Fat Tree fabric except link-M connecting Leaf112 to ToF22. 889 The RIFT control protocol can discover the physical links 890 automatically and be able to detect cabling that violates Fat Tree 891 topology constraints. It reacts accordingly to such mis-cabling 892 attempts, at a minimum preventing adjacencies between nodes from 893 being formed and traffic from being forwarded on those mis-cabled 894 links. Leaf112 will in such scenario use link-M to derive its level 895 (unless it is leaf) and can report links to Spine111 and Spine112 as 896 mis-cabled unless the implementations allows horizontal links. 898 Figure 7 shows a multiple plane mis-cabling example. Since Leaf112 899 and Spine121 belong to two different PoDs, the adjacency between 900 Leaf112 and Spine121 can not be formed. link-W would be detected and 901 prevented. 903 +-------+ +-------+ +-------+ +-------+ 904 |ToF A1| |ToF A2| |ToF B1| |ToF B2| LEVEL 2 905 +-------+ +-------+ +-------+ +-------+ 906 | | | | | | | | 907 | | | +-----------------+ | | | 908 | +--------------------------+ | | | | 909 | +------+ | | | +------+ | 910 | | +-----------------+ | | | | | 911 | | | +--------------------------+ | | 912 | A | | B | | A | | B | 913 +-----+--+ +-+---+--+ +--+---+-+ +--+-----+ 914 |Spine111| |Spine112| +---+Spine121| |Spine122| LEVEL 1 915 +-+---+--+ ++----+--+ | +--+---+-+ +-+----+-+ 916 | | | | | | | | | 917 | +---------+ | | | +---------+ | 918 | | | | link-W | | | | 919 | +-------+ | | | | +-------+ | | 920 | | | | | | | | | 921 +-+---+-+ +--+--+-+ | +-+---+-+ +--+--+-+ 922 |Leaf111| |Leaf112+------+ |Leaf121| |Leaf122| LEVEL 0 923 +-------+ +-------+ +-------+ +-------+ 924 +--------PoD#1----------+ +---------PoD#2---------+ 926 Figure 7: A multiple plane mis-cabling example 928 RIFT provides an optional level determination procedure in its Zero 929 Touch Provisioning mode. Nodes in the fabric without their level 930 configured determine it automatically. This can have possibly 931 counter-intuitive consequences however. One extreme failure scenario 932 is depicted in Figure 8 and it shows that if all northbound links of 933 spine11 fail at the same time, spine11 negotiates a lower level than 934 Leaf11 and Leaf12. 936 To prevent such scenario where leafs are expected to act as switches, 937 LEAF_ONLY flag can be set for Leaf111 and Leaf112. Since level -1 is 938 invalid, Spine11 would not derive a valid level from the topology in 939 Figure 8. It will be isolated from the whole fabric and it would be 940 up to the leafs to declare the links towards such spine as mis- 941 cabled. 943 +-------+ +-------+ +-------+ +-------+ 944 |ToF A1| |ToF A2| |ToF A1| |ToF A2| 945 +-------+ +-------+ +-------+ +-------+ 946 | | | | | | 947 | +-------+ | | | 948 + + | | ====> | | 949 X X +------+ | +------+ | 950 + + | | | | 951 +----+--+ +-+-----+ +-+-----+ 952 |Spine11| |Spine12| |Spine12| 953 +-+---+-+ ++----+-+ ++----+-+ 954 | | | | | | 955 | +---------+ | | | 956 | +-------+ | | +-------+ | 957 | | | | | | 958 +-+---+-+ +--+--+-+ +-----+-+ +-----+-+ 959 |Leaf111| |Leaf112| |Leaf111| |Leaf112| 960 +-------+ +-------+ +-+-----+ +-+-----+ 961 | | 962 | +--------+ 963 | | 964 +-+---+-+ 965 |Spine11| 966 +-------+ 968 Figure 8: Fallen spine 970 5.6. Positive vs. Negative Disaggregation 972 Disaggregation is the procedure whereby [RIFT] advertises a more 973 specific route southwards as an exception to the aggregated fabric- 974 default north. Disaggregation is useful when a prefix within the 975 aggregation is reachable via some of the parents but not the others 976 at the same level of the fabric. It is mandatory when the level is 977 the ToF since a ToF node that cannot reach a prefix becomes a black 978 hole for that prefix. The hard problem is to know which prefixes are 979 reachable by whom. 981 In the general case, [RIFT] solves that problem by interconnecting 982 the ToF nodes. So the ToF nodes can exchange the full list of 983 prefixes that exist in the fabric and figure out when a ToF node 984 lacks reachability to some prefixes. This requires additional ports 985 at the ToF, typically 2 ports per ToF node to form a ToF-spanning 986 ring. [RIFT] also defines the southbound reflection procedure that 987 enables a parent to explore the direct connectivity of its peers, 988 meaning their own parents and children; based on the advertisements 989 received from the shared parents and children, it may enable the 990 parent to infer the prefixes its peers can reach. 992 When a parent lacks reachability to a prefix, it may disaggregate the 993 prefix negatively, i.e., advertise that this parent can be used to 994 reach any prefix in the aggregation except that one. The Negative 995 Disaggregation signaling is simple and functions transitively from 996 ToF to top-of-pod (ToP) and then from ToP to Leaf. But it is hard 997 for a parent to figure which prefix it needs to disaggregate, because 998 it does not know what it does not know; it results that the use of a 999 spanning ring at the ToF is required to operate the Negative 1000 Disaggregation. Also, though it is only an implementation problem, 1001 the programmation of the FIB is complex compared to normal routes, 1002 and may incur recursions. 1004 The more classical alternative is, for the parents that can reach a 1005 prefix that peers at the same level cannot, to advertise a more 1006 specific route to that prefix. This leverages the normal longest 1007 prefix match in the FIB, and does not require a special 1008 implementation. But as opposed to the Negative Disaggregation, the 1009 Positive Disaggregation is difficult and inefficient to operate 1010 transitively. 1012 Transitivity is not needed to a grandchild if all its parents 1013 received the Positive Disaggregation, meaning that they shall all 1014 avoid the black hole; when that is the case, they collectively build 1015 a ceiling that protects the grandchild. But until then, a parent 1016 that received a Positive Disaggregation may believe that some peers 1017 are lacking the reachability and readvertise too early, or defer and 1018 maintain a black hole situation longer than necessary. 1020 In a non-partitioned fabric, all the ToF nodes see one another 1021 through the reflection and can figure if one is missing a child. In 1022 that case it is possible to compute the prefixes that the peer cannot 1023 reach and disaggregate positively without a ToF-spanning ring. The 1024 ToF nodes can also ascertain that the ToP nodes are connected each to 1025 at least a ToF node that can still reach the prefix, meaning that the 1026 transitive operation is not required. 1028 The bottom line is that in a fabric that is partitioned (e.g., using 1029 multiple planes) and/or where the ToP nodes are not guaranteed to 1030 always form a ceiling for their children, it is mandatory to use the 1031 Negative Disaggregation. On the other hand, in a highly symmetrical 1032 and fully connected fabric, (e.g., a canonical Clos Network), the 1033 Positive Disaggregation methods allows to save the complexity and 1034 cost associated to the ToF-spanning ring. 1036 Note that in the case of Positive Disaggregation, the first ToF 1037 node(s) that announces a more-specific route attracts all the traffic 1038 for that route and may suffer from a transient incast. A ToP node 1039 that defers injecting the longer prefix in the FIB, in order to 1040 receive more advertisements and spread the packets better, also keeps 1041 on sending a portion of the traffic to the black hole in the 1042 meantime. In the case of Negative Disaggregation, the last ToF 1043 node(s) that injects the route may also incur an incast issue; this 1044 problem would occur if a prefix that becomes totally unreachable is 1045 disaggregated, but doing so is mostly useless and is not recommended. 1047 5.7. Mobile Edge and Anycast 1049 When a physical or a virtual node changes its point of attachement in 1050 the fabric from a previous-leaf to a next-leaf, new routes must be 1051 installed that supersede the old ones. Since the flooding flows 1052 northwards, the nodes (if any) between the previous-leaf and the 1053 common parent are not immediately aware that the path via previous- 1054 leaf is obsolete, and a stale route may exist for a while. The 1055 common parent needs to select the freshest route advertisement in 1056 order to install the correct route via the next-leaf. This requires 1057 that the fabric determines the sequence of the movements of the 1058 mobile node. 1060 On the one hand, a classical sequence counter provides a total order 1061 for a while but it will eventually wrap. On the other hand, a 1062 timestamp provides a permanent order but it may miss a movement that 1063 happens too quickly vs. the granularity of the timing information. 1064 It is not envisioned that an average fabric supports Precision Time 1065 Protocol [IEEEstd1588] in the short term, nor that the precision 1066 available with the Network Time Protocol [RFC5905] (in the order of 1067 100 to 200ms) may not be necessarily enough to cover, e.g., the fast 1068 mobility of a Virtual Machine. 1070 Section 4.3.3. "Mobility" of [RIFT] specifies an hybrid method that 1071 combines a sequence counter from the mobile node and a timestamp from 1072 the network taken at the leaf when the route is injected. If the 1073 timestamps of the concurrent advertisements are comparable (i.e., 1074 more distant than the precision of the timing protocol), then the 1075 timestamp alone is used to determine the relative freshness of the 1076 routes. Otherwise, the sequence counter from the mobile node, if 1077 available, is used. One caveat is that the sequence counter must not 1078 wrap within the precision of the timing protocol. Another is that 1079 the mobile node may not even provide a sequence counter, in which 1080 case the mobility itself must be slower than the precision of the 1081 timing. 1083 Mobility must not be confused with anycast. In both cases, a same 1084 address is injected in RIFT at different leaves. In the case of 1085 mobility, only the freshest route must be conserved, since mobile 1086 node changed its point of attachment for a leaf to the next. In the 1087 case of anycast, the node may be either multihomed (attached to 1088 multiple leaves in parallel) or reachable beyond the fabric via 1089 multiple routes that are redistributed to different leaves; either 1090 way, in the case of anycast, the multiple routes are equally valid 1091 and should be conserved. Without further information from the 1092 redistributed routing protocol, it is impossible to sort out a 1093 movement from a redistribution that happens asynchronously on 1094 different leaves. [RIFT] expects that anycast addresses are 1095 advertised within the timing precision, which is typically the case 1096 with a low-precision timing and a multihomed node. Beyond that time 1097 interval, RIFT interprets the lag as a mobility and only the freshest 1098 route is retained. 1100 When using IPv6 [RFC8200], RIFT suggests to leverage "Registration 1101 Extensions for IPv6 over Low-Power Wireless Personal Area Network 1102 (6LoWPAN) Neighbor Discovery (ND)" [RFC8505] as the IPv6 ND 1103 interaction between the mobile node and the leaf. This provides not 1104 only a sequence counter but also a lifetime and a security token that 1105 may be used to protect the ownership of an address [RFC8928]. When 1106 using [RFC8505], the parallel registration of an anycast address to 1107 multiple leaves is done with the same sequence counter, whereas the 1108 sequence counter is incremented when the point of attachement 1109 changes. This way, it is possible to differentiate a mobile node 1110 from a multihomed node, even when the mobility happens within the 1111 timing precision. It is also possible for a mobile node to be 1112 multihomed as well, e.g., to change only one of its points of 1113 attachement. 1115 5.8. IPv4 over IPv6 1117 RIFT allows advertising IPv4 prefixes over IPv6 RIFT network. IPv6 1118 Address Family (AF) configures via the usual Neighbor Discovery (ND) 1119 mechanisms and then V4 can use V6 nexthops analogous to [RFC8950]. 1120 It is expected that the whole fabric supports the same type of 1121 forwarding of address families on all the links. RIFT provides an 1122 indication whether a node is v4 forwarding capable and 1123 implementations are possible where different routing tables are 1124 computed per address family as long as the computation remains loop- 1125 free. 1127 +-----+ +-----+ 1128 +---+---+ | ToF | | ToF | 1129 ^ +--+--+ +-----+ 1130 | | | | | 1131 | | +-------------+ | 1132 | | +--------+ | | 1133 + | | | | 1134 V6 +-----+ +-+---+ 1135 Forwarding |Spine| |Spine| 1136 + +--+--+ +-----+ 1137 | | | | | 1138 | | +-------------+ | 1139 | | +--------+ | | 1140 | | | | | 1141 v +-----+ +-+---+ 1142 +---+---+ |Leaf | | Leaf| 1143 +--+--+ +--+--+ 1144 | | 1145 IPv4 prefixes| |IPv4 prefixes 1146 | | 1147 +---+----+ +---+----+ 1148 | V4 | | V4 | 1149 | subnet | | subnet | 1150 +--------+ +--------+ 1152 Figure 9: IPv4 over IPv6 1154 5.9. In-Band Reachability of Nodes 1156 RIFT doesn't precondition that nodes of the fabric have reachable 1157 addresses. But the operational reasons to reach the internal nodes 1158 may exist. Figure 10 shows an example that the network management 1159 station (NMS) attaches to leaf1. 1161 +-------+ +-------+ 1162 | ToF1 | | ToF2 | 1163 ++---- ++ ++-----++ 1164 | | | | 1165 | +----------+ | 1166 | +--------+ | | 1167 | | | | 1168 ++-----++ +--+---++ 1169 |Spine1 | |Spine2 | 1170 ++-----++ ++-----++ 1171 | | | | 1172 | +----------+ | 1173 | +--------+ | | 1174 | | | | 1175 ++-----++ +--+---++ 1176 | Leaf1 | | Leaf2 | 1177 +---+---+ +-------+ 1178 | 1179 |NMS 1181 Figure 10: In-Band reachability of node 1183 If NMS wants to access Leaf2, it simply works. Because loopback 1184 address of Leaf2 is flooded in its Prefix North TIE. 1186 If NMS wants to access Spine2, it simply works too. Because spine 1187 node always advertises its loopback address in the Prefix North TIE. 1188 NMS may reach Spine2 from Leaf1-Spine2 or Leaf1-Spine1-ToF1/ 1189 ToF2-Spine2. 1191 If NMS wants to access ToF2, ToF2's loopback address needs to be 1192 injected into its Prefix South TIE. This TIE must be seen by all 1193 nodes at the level below - the spine nodes in Figure 10 - that must 1194 form a ceiling for all the traffic coming from below (south). 1195 Otherwise, the traffic from NMS may follow the default route to the 1196 wrong ToF Node, e.g., ToF1. 1198 In case of failure between ToF2 and spine nodes, ToF2's loopback 1199 address must be disaggregated recursively all the way to the leaves. 1200 In a partitioned ToF, even with recursive disaggregation a ToF node 1201 is only reachable within its plane. 1203 A possible alternative to recursive disaggregation is to use a ring 1204 that interconnects the ToF nodes to transmit packets between them for 1205 their loopback addresses only. The idea is that this is mostly 1206 control traffic and should not alter the load balancing properties of 1207 the fabric. 1209 5.10. Dual Homing Servers 1211 Each RIFT node may operate in Zero Touch Provisioning (ZTP) mode. It 1212 has no configuration (unless it is a Top-of-Fabric at the top of the 1213 topology or the must operate in the topology as leaf and/or support 1214 leaf-2-leaf procedures) and it will fully configure itself after 1215 being attached to the topology. 1217 +---+ +---+ +---+ 1218 |ToF| |ToF| |ToF| ToF 1219 +---+ +---+ +---+ 1220 | | | | | | 1221 | +----------------+ | | 1222 | +----------------+ | 1223 | | | | | | 1224 +----------+--+ +--+----------+ 1225 | ToR1 | | ToR2 | Spine 1226 +--+------+---+ +--+-------+--+ 1227 +---+ | | | | | | +---+ 1228 | +-----------------+ | | | 1229 | | | +-------------+ | | 1230 + | + | | +-----------------+ | 1231 X | X | +--------x-----+ | X | 1232 + | + | | | + | 1233 +---+ +---+ +---+ +---+ 1234 | | | | | | | | 1235 +---+ +---+ ...............+---+ +---+ 1236 SV(1) SV(2) SV(n-1) SV(n) Leaf 1238 Figure 11: Dual-homing servers 1240 In the single plane, the worst condition is disaggregation of every 1241 other servers at the same level. Suppose the links from ToR1 (Top of 1242 Rack) to all the leaves become not available. All the servers' 1243 routes are disaggregated and the FIB of the servers will be expanded 1244 with n-1 more specific routes. 1246 Sometimes, people may prefer to disaggregate from ToR to servers from 1247 start on, i.e. the servers have couple tens of routes in FIB from 1248 start on beside default routes to avoid breakages at rack level. 1249 Full disaggregation of the fabric could be achieved by configuration 1250 supported by RIFT. 1252 5.11. Fabric With A Controller 1254 There are many different ways to deploy the controller. One 1255 possibility is attaching a controller to the RIFT domain from ToF and 1256 another possibility is attaching a controller from the leaf. 1258 +------------+ 1259 | Controller | 1260 ++----------++ 1261 | | 1262 | | 1263 +----++ ++----+ 1264 ------- | ToF | | ToF | 1265 | +--+--+ +-----+ 1266 | | | | | 1267 | | +-------------+ | 1268 | | +--------+ | | 1269 | | | | | 1270 +-----+ +-+---+ 1271 RIFT domain |Spine| |Spine| 1272 +--+--+ +-----+ 1273 | | | | | 1274 | | +-------------+ | 1275 | | +--------+ | | 1276 | | | | | 1277 | +-----+ +-+---+ 1278 ------- |Leaf | | Leaf| 1279 +-----+ +-----+ 1281 Figure 12: Fabric with a controller 1283 5.11.1. Controller Attached to ToFs 1285 If a controller is attaching to the RIFT domain from ToF, it usually 1286 uses dual-homing connections. The loopback prefix of the controller 1287 should be advertised down by the ToF and spine to leaves. If the 1288 controller loses link to ToF, make sure the ToF withdraw the prefix 1289 of the controller. 1291 5.11.2. Controller Attached to Leaf 1293 If the controller is attaching from a leaf to the fabric, no special 1294 provisions are needed. 1296 5.12. Internet Connectivity Within Underlay 1298 If global addressing is running without overlay, an external default 1299 route needs to be advertised through RIFT fabric to achieve internet 1300 connectivity. For the purpose of forwarding of the entire RIFT 1301 fabric, an internal fabric prefix needs to be advertised in the South 1302 Prefix TIE by ToF and spine nodes. 1304 5.12.1. Internet Default on the Leaf 1306 In case that the internet gateway is a leaf, the leaf node as the 1307 internet gateway needs to advertise a default route in its Prefix 1308 North TIE. 1310 5.12.2. Internet Default on the ToFs 1312 In case that the internet gateway is a ToF, the ToF and spine nodes 1313 need to advertise a default route in the Prefix South TIE. 1315 5.13. Subnet Mismatch and Address Families 1317 +--------+ +--------+ 1318 | | LIE LIE | | 1319 | A | +----> <----+ | B | 1320 | +---------------------+ | 1321 +--------+ +--------+ 1322 X/24 Y/24 1324 Figure 13: subnet mismatch 1326 LIEs are exchanged over all links running RIFT to perform Link 1327 (Neighbor) Discovery. A node must NOT originate LIEs on an address 1328 family if it does not process received LIEs on that family. LIEs on 1329 same link are considered part of the same negotiation independent on 1330 the address family they arrive on. An implementation must be ready 1331 to accept TIEs on all addresses it used as source of LIE frames. 1333 As shown in the above figure, without further checks adjacency of 1334 node A and B may form, but the forwarding between node A and node B 1335 may fail because subnet X mismatches with subnet Y. 1337 To prevent this a RIFT implementation should check for subnet 1338 mismatch just like e.g. ISIS does. This can lead to scenarios where 1339 an adjacency, despite exchange of LIEs in both address families may 1340 end up having an adjacency in a single AF only. This is a 1341 consideration especially in Section 5.8 scenarios. 1343 5.14. Anycast Considerations 1344 + traffic 1345 | 1346 v 1347 +------+------+ 1348 | ToF | 1349 +---+-----+---+ 1350 | | | | 1351 +------------+ | | +------------+ 1352 | | | | 1353 +---+---+ +-------+ +-------+ +---+---+ 1354 | | | | | | | | 1355 |Spine11| |Spine12| |Spine21| |Spine22| LEVEL 1 1356 +-+---+-+ ++----+-+ +-+---+-+ ++----+-+ 1357 | | | | | | | | 1358 | +---------+ | | +---------+ | 1359 | +-------+ | | | +-------+ | | 1360 | | | | | | | | 1361 +-+---+-+ +--+--+-+ +-+---+-+ +--+--+-+ 1362 | | | | | | | | 1363 |Leaf111| |Leaf112| |Leaf121| |Leaf122| LEVEL 0 1364 +-+-----+ ++------+ +-----+-+ +-----+-+ 1365 + + + ^ | 1366 PrefixA PrefixB PrefixA | PrefixC 1367 | 1368 + traffic 1370 Figure 14: Anycast 1372 If the traffic comes from ToF to Leaf111 or Leaf121 which has anycast 1373 prefix PrefixA, RIFT can deal with this case well. But if the 1374 traffic comes from Leaf122, it arrives Spine21 or Spine22 at level 1. 1375 But Spine21 or Spine22 doesn't know another PrefixA attaching 1376 Leaf111. So it will always get to Leaf121 and never get to Leaf111. 1377 If the intension is that the traffic should been offloaded to 1378 Leaf111, then use policy guided prefixes defined in RIFT [RIFT]. 1380 5.15. IoT Applicability 1382 The design of RIFT inherits from RPL [RFC6550] the anisotropic design 1383 of a default route upwards (northwards); it also inherits the 1384 capability to inject external host routes at the Leaf level using 1385 Wireless ND (WiND) [RFC8505][RFC8928] between a RIFT-agnostic host 1386 and a RIFT router. Both the RPL and the RIFT protocols are meant for 1387 large scale, and WiND enables device mobility at the edge the same 1388 way in both cases. 1390 The main difference between RIFT and RPL is that with RPL, there's a 1391 single Root, whereas RIFT has many ToF nodes. This adds huge 1392 capabilities for leaf-2-leaf ECMP paths, but additional complexity 1393 with the need to disaggregate. Also RIFT uses Link State flooding 1394 northwards, and is not designed for low-power operation. 1396 Still nothing prevents that the IP devices connected at the Leaf are 1397 IoT (Internet of Things) devices, which typically expose their 1398 address using WiND - which is an upgrade from 6LoWPAN ND [RFC6775]. 1400 A network that serves high speed/ high power IoT devices should 1401 typically provide deterministic capabilities for applications such as 1402 high speed control loops or movement detection. The Fat Tree is 1403 highly reliable, and in normal condition provides an equivalent 1404 multipath operation; but the ECMP doesn't provide hard guarantees for 1405 either delivery or latency. As long as the fabric is non-blocking 1406 the result is the same; but there can be load unbalances resulting in 1407 incast and possibly congestion loss that will prevent the delivery 1408 within bounded latency. 1410 This could be alleviated with Packet Replication, Elimination and 1411 Reordering (PREOF) [RFC8655] leaf-2-leaf but PREOF is hard to provide 1412 at the scale of all flows, and the replication may increase the 1413 probability of the overload that it attempts to solve. 1415 Note that the load balancing is not RIFT's problem, but it is key to 1416 serve IoT adequately. 1418 5.16. Key Management 1420 As outlined in Section "Security Considerations" of [RIFT], either a 1421 private shared key or a public/private key pair is used to 1422 authenticate the adjacency. Both the key distribution and key 1423 synchronization methods are out of scope for this document. Both 1424 nodes in the adjacency must share the same keys, key type, and 1425 algorithm for a given key ID. Mismatched keys will not inter-operate 1426 as their security envelopes will be unverifiable. 1428 Key roll-over while the adjacency is active may be supported. The 1429 specific mechanism is well documented in [RFC6518]. 1431 6. Security Considerations 1433 This document presents applicability of RIFT. As such, it does not 1434 introduce any security considerations. However, there are a number 1435 of security concerns at [RIFT]. 1437 7. IANA Considerations 1439 This document has no IANA actions. 1441 8. Contributors 1443 The following people (listed in alphabetical order) contributed 1444 significantly to the content of this document and should be 1445 considered co-authors: 1447 Tom Verhaeg 1449 Juniper Networks 1451 Email: tverhaeg@juniper.net 1453 Tony Przygienda 1455 Juniper Networks 1457 1194 N. Mathilda Ave 1459 Sunnyvale, CA 94089 1461 US 1463 Email: prz@juniper.net 1465 9. Normative References 1467 [ISO10589-Second-Edition] 1468 International Organization for Standardization, 1469 "Intermediate system to Intermediate system intra-domain 1470 routeing information exchange protocol for use in 1471 conjunction with the protocol for providing the 1472 connectionless-mode Network Service (ISO 8473)", November 1473 2002. 1475 [TR-384] Broadband Forum Technical Report, "TR-384 Cloud Central 1476 Office Reference Architectural Framework", January 2018. 1478 [RFC2328] Moy, J., "OSPF Version 2", STD 54, RFC 2328, 1479 DOI 10.17487/RFC2328, April 1998, 1480 . 1482 [RFC4861] Narten, T., Nordmark, E., Simpson, W., and H. Soliman, 1483 "Neighbor Discovery for IP version 6 (IPv6)", RFC 4861, 1484 DOI 10.17487/RFC4861, September 2007, 1485 . 1487 [RFC5357] Hedayat, K., Krzanowski, R., Morton, A., Yum, K., and J. 1488 Babiarz, "A Two-Way Active Measurement Protocol (TWAMP)", 1489 RFC 5357, DOI 10.17487/RFC5357, October 2008, 1490 . 1492 [RFC7130] Bhatia, M., Ed., Chen, M., Ed., Boutros, S., Ed., 1493 Binderberger, M., Ed., and J. Haas, Ed., "Bidirectional 1494 Forwarding Detection (BFD) on Link Aggregation Group (LAG) 1495 Interfaces", RFC 7130, DOI 10.17487/RFC7130, February 1496 2014, . 1498 [RFC8950] Litkowski, S., Agrawal, S., Ananthamurthy, K., and K. 1499 Patel, "Advertising IPv4 Network Layer Reachability 1500 Information (NLRI) with an IPv6 Next Hop", RFC 8950, 1501 DOI 10.17487/RFC8950, November 2020, 1502 . 1504 [RFC6518] Lebovitz, G. and M. Bhatia, "Keying and Authentication for 1505 Routing Protocols (KARP) Design Guidelines", RFC 6518, 1506 DOI 10.17487/RFC6518, February 2012, 1507 . 1509 [RFC6550] Winter, T., Ed., Thubert, P., Ed., Brandt, A., Hui, J., 1510 Kelsey, R., Levis, P., Pister, K., Struik, R., Vasseur, 1511 JP., and R. Alexander, "RPL: IPv6 Routing Protocol for 1512 Low-Power and Lossy Networks", RFC 6550, 1513 DOI 10.17487/RFC6550, March 2012, 1514 . 1516 [RFC6775] Shelby, Z., Ed., Chakrabarti, S., Nordmark, E., and C. 1517 Bormann, "Neighbor Discovery Optimization for IPv6 over 1518 Low-Power Wireless Personal Area Networks (6LoWPANs)", 1519 RFC 6775, DOI 10.17487/RFC6775, November 2012, 1520 . 1522 [RFC8655] Finn, N., Thubert, P., Varga, B., and J. Farkas, 1523 "Deterministic Networking Architecture", RFC 8655, 1524 DOI 10.17487/RFC8655, October 2019, 1525 . 1527 [RIFT] Sharma, A., Thubert, P., Rijsman, B., and D. Afanasiev, 1528 "RIFT: Routing in Fat Trees", Work in Progress, Internet- 1529 Draft, draft-ietf-rift-rift-13, 12 July 2021, 1530 . 1533 10. Informative References 1535 [IEEEstd1588] 1536 IEEE standard for Information Technology, "IEEE Standard 1537 for a Precision Clock Synchronization Protocol for 1538 Networked Measurement and Control Systems", 1539 . 1541 [CLOS] Yuan, X., "On Nonblocking Folded-Clos Networks in Computer 1542 Communication Environments", IEEE International Parallel & 1543 Distributed Processing Symposium, 2011. 1545 [FATTREE] Leiserson, C. E., "Fat-Trees: Universal Networks for 1546 Hardware-Efficient Supercomputing", 1985. 1548 [RFC3626] Clausen, T., Ed. and P. Jacquet, Ed., "Optimized Link 1549 State Routing Protocol (OLSR)", RFC 3626, 1550 DOI 10.17487/RFC3626, October 2003, 1551 . 1553 [RFC5905] Mills, D., Martin, J., Ed., Burbank, J., and W. Kasch, 1554 "Network Time Protocol Version 4: Protocol and Algorithms 1555 Specification", RFC 5905, DOI 10.17487/RFC5905, June 2010, 1556 . 1558 [RFC8200] Deering, S. and R. Hinden, "Internet Protocol, Version 6 1559 (IPv6) Specification", STD 86, RFC 8200, 1560 DOI 10.17487/RFC8200, July 2017, 1561 . 1563 [RFC8505] Thubert, P., Ed., Nordmark, E., Chakrabarti, S., and C. 1564 Perkins, "Registration Extensions for IPv6 over Low-Power 1565 Wireless Personal Area Network (6LoWPAN) Neighbor 1566 Discovery", RFC 8505, DOI 10.17487/RFC8505, November 2018, 1567 . 1569 [RFC8928] Thubert, P., Ed., Sarikaya, B., Sethi, M., and R. Struik, 1570 "Address-Protected Neighbor Discovery for Low-Power and 1571 Lossy Networks", RFC 8928, DOI 10.17487/RFC8928, November 1572 2020, . 1574 Authors' Addresses 1576 Yuehua Wei (editor) 1577 ZTE Corporation 1578 No.50, Software Avenue 1579 Nanjing 1580 210012 1581 China 1583 Email: wei.yuehua@zte.com.cn 1585 Zheng Zhang 1586 ZTE Corporation 1587 No.50, Software Avenue 1588 Nanjing 1589 210012 1590 China 1592 Email: zhang.zheng@zte.com.cn 1594 Dmitry Afanasiev 1595 Yandex 1597 Email: fl0w@yandex-team.ru 1599 Pascal Thubert 1600 Cisco Systems, Inc 1601 Building D 1602 45 Allee des Ormes - BP1200 1603 06254 MOUGINS - Sophia Antipolis 1604 France 1606 Phone: +33 497 23 26 34 1607 Email: pthubert@cisco.com 1609 Jaroslaw Kowalczyk 1610 Orange Polska 1612 Email: jaroslaw.kowalczyk2@orange.com