idnits 2.17.00 (12 Aug 2021) /tmp/idnits19188/draft-huang-computing-delivery-in-routing-network-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 19 instances of too long lines in the document, the longest one being 6 characters in excess of 72. Miscellaneous warnings: ---------------------------------------------------------------------------- == The document doesn't use any RFC 2119 keywords, yet seems to have RFC 2119 boilerplate text. -- The document date (7 March 2022) is 68 days in the past. Is this intentional? Checking references for intended status: Proposed Standard ---------------------------------------------------------------------------- (See RFCs 3967 and 4897 for information about using normative references to lower-maturity documents in RFCs) No issues found here. Summary: 1 error (**), 0 flaws (~~), 1 warning (==), 1 comment (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 RTGWG D.H. Daniel 3 Internet-Draft B.T. Bin 4 Intended status: Standards Track ZTE Corporation 5 Expires: 8 September 2022 P.L. Peng 6 China Mobile 7 7 March 2022 9 Computing Delivery in Routing Network 10 draft-huang-computing-delivery-in-routing-network-01 12 Abstract 14 This document drafts a proposal of the architecture of Computing 15 Delivery in Routing Network which incorporates both computing and 16 networking metrics into the routing polices and enables the network 17 sensing and scheduling computing services based upon traditional 18 networking services. A mechanism of two-class computing status 19 granularity and two segment routing is illustrated for end-to-end 20 networking and computing service in the cloud sites, while major 21 networking and computing actors is defined in terms of functionality. 22 An example work flow is demonstrated, and both control plane and data 23 plane solution consideration is proposed.. 25 Status of This Memo 27 This Internet-Draft is submitted in full conformance with the 28 provisions of BCP 78 and BCP 79. 30 Internet-Drafts are working documents of the Internet Engineering 31 Task Force (IETF). Note that other groups may also distribute 32 working documents as Internet-Drafts. The list of current Internet- 33 Drafts is at https://datatracker.ietf.org/drafts/current/. 35 Internet-Drafts are draft documents valid for a maximum of six months 36 and may be updated, replaced, or obsoleted by other documents at any 37 time. It is inappropriate to use Internet-Drafts as reference 38 material or to cite them other than as "work in progress." 40 This Internet-Draft will expire on 8 September 2022. 42 Copyright Notice 44 Copyright (c) 2022 IETF Trust and the persons identified as the 45 document authors. All rights reserved. 47 This document is subject to BCP 78 and the IETF Trust's Legal 48 Provisions Relating to IETF Documents (https://trustee.ietf.org/ 49 license-info) in effect on the date of publication of this document. 50 Please review these documents carefully, as they describe your rights 51 and restrictions with respect to this document. Code Components 52 extracted from this document must include Revised BSD License text as 53 described in Section 4.e of the Trust Legal Provisions and are 54 provided without warranty as described in the Revised BSD License. 56 Table of Contents 58 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 2 59 1.1. Requirements Language . . . . . . . . . . . . . . . . . . 3 60 2. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . 3 61 3. Computing delivery in routing network reference 62 architecture . . . . . . . . . . . . . . . . . . . . . . 4 63 3.1. Hierarchical granularity routing scheme . . . . . . . . . 5 64 3.2. Two-segment routing and forwarding . . . . . . . . . . . 6 65 3.3. Cross-domain computing routing and forwarding . . . . . . 6 66 3.4. CSI routing . . . . . . . . . . . . . . . . . . . . . . . 7 67 3.5. Traffic affinity . . . . . . . . . . . . . . . . . . . . 7 68 4. Computing delivery in routing network architecture work 69 flow . . . . . . . . . . . . . . . . . . . . . . . . . . 8 70 4.1. Computing resource and service update work flow . . . . . 8 71 4.2. Service flow routing and forwarding work flow . . . . . . 8 72 5. Control plane . . . . . . . . . . . . . . . . . . . . . . . . 9 73 5.1. Centralized control plane . . . . . . . . . . . . . . . . 9 74 5.2. Distributed control plane . . . . . . . . . . . . . . . . 9 75 5.3. Hybrid control plane . . . . . . . . . . . . . . . . . . 9 76 6. Data plane . . . . . . . . . . . . . . . . . . . . . . . . . 9 77 6.1. CSI encapsulation . . . . . . . . . . . . . . . . . . . . 10 78 6.2. CSI for GCR, CUR and LCR . . . . . . . . . . . . . . . . 10 79 7. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 80 8. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 10 81 9. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 10 82 10. Security Considerations . . . . . . . . . . . . . . . . . . . 11 83 11. Informative References . . . . . . . . . . . . . . . . . . . 11 84 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 11 86 1. Introduction 88 Computing-related services have been provided in such a way that 89 computing resources either are confined within isolated sites (data 90 centers, MECs etc.) without coordination among multiple sites or they 91 are coordinated and managed within specific and closed service 92 systems without fine-grained networking facilitation, while the 93 industry develops into an era in which the computing resources start 94 migrating from centralized data centers to distributed edge nodes. 96 Therefore substantial benefits in light of both cost and efficiency 97 resulting from scale of economy, would be brought into multiple 98 industries by intelligently and dynamically connecting the 99 distributed computing resources and rendering the coordinated 100 computing resources as a unified and virtual resource pool. On top 101 of the cost and efficiency gains, applications as well as services 102 would be served in a more sophisticated way in which computing and 103 networking resources could be aligned more efficiently and agilely 104 than conventional way in which the two are delivered in separate 105 systems. Some impressive drafts such as 106 [I-D.liu-dyncast-ps-usecases] and [I-D.li-dyncast-architecture] 107 analyze the benefits of routing related solution, and give the 108 reference architecture and preliminary test results. End 109 applications could be served not only by fine-grained computing 110 services but also fine-grained networking services rather than the 111 best-effort networking services without routing network involved 112 otherwise. The cost is the burden of maintaining and sensing 113 computing resource status in the networking layer. The architecture 114 proposal is designed to be as much smoothly compatible with the 115 ongoing routing architecture as possible. 117 1.1. Requirements Language 119 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 120 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 121 document are to be interpreted as described in [RFC2119]. 123 2. Terminology 125 * Global Computing-related Routing Node (GCR): routing node 126 maintaining computing resource as well as service status from 127 remote cloud sites, and executing the cross-site routing policies 128 in terms of the aforementioned status as well as the 129 identification of computing resource and service. GCR usually 130 resides at the network edge and works as ingress of the end to end 131 service flow. 133 * Local Computing-related Routing Node (LCR): routing node 134 maintaining computing resource as well as service status from the 135 geographically local cloud sites and being responsible for the 136 last hop of the service flow towards the computing resource and 137 service instance in the specific cloud site. LCAT usually resides 138 at the network edge and works as egress of the end to end service 139 flow. 141 * Computing Unaware Routing Node (CUR): routing node unaware of 142 computing resource and service status and disregarding 143 encapsulation of the identification of computing service. CUR 144 usually resides between GCR and LCR and works as ordinary routing 145 nodes and stays irrelevant of computing delivery. 147 * Global Computing Resource and Service Status (GCRS): General cloud 148 site status of the computing resource and service which consists 149 of overall resource occupation and types of computing service 150 (algorithms, functions etc.) the specific cloud site provides. 151 GCRS is maintained at GCR and expected to remain relatively stable 152 and change in slow frequency. 154 * Local Computing Resource and Service Status (LCRS): fine-grained 155 cloud site status of the computing resource and service which 156 consists of status of each active computing service instance as 157 well as its parameters which impact the way the instance would be 158 selected and visited by LCR. LCRS is maintained at LCR and 159 expected to stay quite active and change in high frequency. 161 * Computing Service Identification (CSI): a globally unique 162 identification of a computing service with optional parameters, 163 and it could be an IPv6-like address or specifically designed 164 identification structure. 166 * Instantiated Computing Service (ICS): an active instance of a 167 computing service identification which resides in a host usually 168 purporting to a server, container or virtual machine. 170 3. Computing delivery in routing network reference architecture 172 Routing network is enabled sensing the computing resource and service 173 from the cloud sites and routing the service flow according to both 174 network and computing metrics by a computing delivery in routing 175 network architecture as illustrated in figure 1. The architecture is 176 a horizontal convergence of cloud and network, while the latter 177 maintains the converged resource status and thus is able to achieve 178 an end to end routing and forwarding policy from a perspective of 179 cloud and network resource. PE1 maintains GCRS with a whole picture 180 of the multiple cloud sites, and executes the routing policy for the 181 network segment between PE1 and PE2 or PE3, namely between ingress 182 and egress, while PE2 maintains LCRS with a focus picture of the 183 cloud site where S1 resides, and establishes a connection towards S1. 184 S1 is an active instance of a specific computing service type (CSI). 185 On top of the role of LCR which maintains LCRS, PE2 and PE3 also 186 fulfill the role GCR which maintains GCRS from neighboring cloud 187 sites. P provides traditional routing and forwarding functionality 188 for computing service flow, and remains unaware of any computing- 189 related status as well as CSI encapsulations. 191 +--------+ +--------+ 192 +------>|LCR/GCR |------->| ICS | 193 | +--------+ +--------+ 194 +--------+ +--------+ | PE2 S1 195 | GCR |--->| CUR |--+ 196 +--------+ +--------+ | PE3 S2 197 PE1 P | +--------+ +--------+ 198 +------>|LCR/GCR |------->| ICS | 199 +--------+ +--------+ 200 |<------------ Network domain --------------->|<--Computing->| 201 domain 203 Figure 1 205 3.1. Hierarchical granularity routing scheme 207 Status updates of computing resource and service in the cloud sites 208 extend in a quite broad range from relatively stable service types 209 and overall resource occupation to extremely dynamic capacity changes 210 as well as busy and idle cycle of service instance. It would be a 211 disaster to build all of the status updates in the network layer 212 which would bring overburdened and volatile routing tables. 214 It should be reasonable to divide the wide range of computing 215 resource and services into different categories with differentiated 216 characteristics from routing perspective. GCRS and LCRS correspond 217 to cross-site domain and local site domain respectively, and GCRS 218 aggregates the computing resource and service status with low update 219 frequency from multiple cloud sites while LCRS focuses only upon the 220 status with high frequency in the local sites. Under this two- 221 granularity scheme, computing-related routing table of GCRS in the 222 GCR remains in a position roughly as stable as the traditional 223 routing table, and the LCRS in the LCR maintains a near synchronized 224 state table of the highly dynamic updates of computing service 225 instances in the local cloud site. Nonetheless, LCRS focusing upon a 226 single and local cloud site is the normal case while upon multiple 227 sites should be exemption if not impossible. 229 3.2. Two-segment routing and forwarding 231 When it comes to end to end service flow routing and forwarding, 232 there is an status information gap between GCRS and LCRS, therefore a 233 two-segment mechanism has to be in place in line with the two- 234 granularity routing scheme demonstrated in 3.1. As is illustrated in 235 figure 2, R1as ingress determines the specific service flow's egress 236 which turns out to be R2 according to policy calculation from GCRS. 237 In particular, the CSI from both in-band (user plane) and out-band 238 (control plane) is the only index for R1 to calculate and determine 239 the egress, it's highly possible to make this egress calculation in 240 terms of both networking (bandwidth, latency etc) and computing 241 Service Agreement Level. Nevertheless, the two SLA routing 242 optimization could be decoupled to such a degree that the traditional 243 routing algorithms could remain as they are. The convergence of the 244 SLA policies as well as the methods to make GCR aware of the two SLA 245 is out of scope of this proposal. 247 +--------+ +--------+ +--------+ +--------+ 248 | GCRS |--->| |--->| LCRS |--->| ICS | 249 +--------+ +--------+ +--------+ +--------+ 250 R1 R 251 |<---------- GCRS segment ---------->|<- LCRS ->| 252 segment 254 Figure 2 256 When the service flow arrives at R2 which terminates the GCRS segment 257 routing and determines S1 which is the service instance selected 258 according to LCRS maintained at R2. Again CSI is the only index for 259 LCRS segment routing process. 261 3.3. Cross-domain computing routing and forwarding 263 Co-ordinated computing resource scheduling among multiple regions 264 which are usually connected by multiple network domains, as 265 illustrated in section 1, is an important part of intended scenarios 266 with regard to why computing-based scheduling and routing is proposed 267 in the first place. The two-segment routing and forwarding scheme 268 illustrated in 3.2 is a typical use case of cross-domain computing 269 routing and forwarding and a good building block for the full-domain 270 scenario solution. Computing metric information is brought into 271 network domain to enable the latter scheduling routing policies 272 beyond network. However, a particular scheme has to be put in place 273 to ensure mild and acceptable impacts upon the ongoing IP routing 274 scheme. A consistent CSI across terminal, network (multiple domains) 275 and cloud along with hierarchical CSI-associated computing resource 276 and service status which corresponds with different network domains, 277 is the enhanced full-domain routing and forwarding solution. Each 278 domain maintains a corresponding computing resource and service 279 status at its edge node and makes the computing-based routing for the 280 domain-related segment which should be connected by the neighboring 281 segments. 283 3.4. CSI routing 285 CSI encapsulated in the headers and maintained in LCRS and GCRS 286 indicates an abstract service type rather than a geographically 287 explicit destination label, thus the routing scheme based upon CSI is 288 actually a two-part and two-layer process in which CSI only indicates 289 the routing intention of user's requested computing service type 290 where routing does not actually materialize in forwarding plane and 291 the explicit routing destination would be determined by LCRS and 292 GCRS. Therefore the actual routing falls within the traditional 293 routing scheme which remains intact. 295 Apart from the indication of computing service routing intention, CSI 296 could also indicates a specific network serivice requirements by 297 associating the networking service policy in GCRS which would 298 therefore schedule the network resources such as an SR tunnel, 299 guaranteed bandwidth etc. at egress. 301 Therefore, GCRS and LCRS in control plane along with CSI 302 encapsulation in user plane enables an logical computing routing sub- 303 layer which is able to be aware of the computing from cloud sites and 304 forward the service flow in terms of computing services as well as 305 computing resources. Nevertheless, this logical sub-layer remains 306 only relevant at ingress and egress nodes and is simply about 307 computing nodes selection rather than executing the real forwarding 308 and routing actions. 310 3.5. Traffic affinity 312 CSI holds the only semantics of the service type that could be 313 deployed as multiple instances within specific cloud site or across 314 multiple cloud sites, CSI in the destination field is not explicit 315 enough for all of the service flow packets to be forwarded to a 316 specific destination. Traffic affinity has to be guaranteed at both 317 GCR and LCR. Once the egress is determined at GCR, the binding 318 relationship between the egress and the service flow's unique 319 identification (5-tuple or other specifically designed labels) is 320 maintained and the subsequent flow could be forwarded upon this 321 binding table. Likewise LCR maintains the binding relationship 322 between the service flow identification and the selected service 323 instance. 325 Traffic affinity could be guaranteed by mechanisms beyond routing 326 layer, but they will not be in the scope of this proposal. 328 4. Computing delivery in routing network architecture work flow 330 4.1. Computing resource and service update work flow 332 The full range of computing resource and service status from a 333 specific cloud site is registered at LCR which maintains LCRS in 334 itself and notifies the part of GCRS to remote GCRs where GCRS would 335 be thus maintained and updated. As is illustrated in figure 3,GCR in 336 R1 from site1 and site 2 is updated by R2 and R3, while LCRS of site 337 1 in R2 is updated by S1 and LCRS of site 2 in R3 is updated by S2. 338 GCRS in R2 and R3 is updated by each other. Edge routers associating 339 with local cloud site establish a mesh fabric to update the according 340 GCRS among the whole network domain, the computing resource and 341 services in distributed cloud sites thus are connected and could be 342 utilized as a single pool for the applications rather than the 343 isolated islands. 345 +--------+ +--------+ 346 +---------------------------|LCR/GCR |<-------| ICS | 347 | +--------+ +--------+ 348 +-----V--+ +--------+ A R2 | S1 349 | GCR | | CUR | | | 350 +-----A--+ +--------+ | R3 V S2 351 R1 | R +--------+ +--------+ 352 +---------------------------|LCR/GCR |<-------| ICS | 353 +--------+ +--------+ 354 |<--------- GCRS update domain ----------->|<-----LCRS------>| 355 domain 357 Figure 3 359 4.2. Service flow routing and forwarding work flow 361 From perspective of the service work flow, more details have actually 362 been demonstrated in 3.2 and 3.3. Rather than the traditional 363 destination-oriented routing mechanism and the segment routing in 364 which the ingress router is explicitly aware of a specific 365 destination, CSI as an abstract label without semantics of physical 366 address works as the required destination from viewpoint of the user 367 in computing delivery in routing network architecture. Therefore the 368 service flow has to be routed and forwarded segment by segment in 369 which the two segment destinations are determined by GCRS and LCRS 370 respectively. 372 5. Control plane 374 5.1. Centralized control plane 376 LCRS's volatility makes it infeasible to be maintained and controlled 377 in a centralized entity, GCRS is the chief computing resource and 378 service status information to be collected and managed in the 379 controller when it comes to centralized control plane with regard to 380 computing delivery in routing network architecture. Routing and 381 forwarding policies from GCRS calculated in the centralized 382 controller, as is demonstrated in 3.2, apply only to the segment from 383 ingress and egress, while the second segment routing policy from 384 egress to the selected service instance in the cloud site is 385 determined by LCRS at egress. 387 Hierarchically centralized control plane architecture would be 388 strongly recommended under the circumstances of nationwide network 389 and cloud management. 391 5.2. Distributed control plane 393 GCRS is updated among the edge routers which have been connected in a 394 mesh way that each pair of edge routers could exchange GCRS to each 395 other, while LCRS will be unidirectionally updated from cloud site to 396 the associated edge router in which LCRS is maintained and its update 397 process is terminated. 399 Protocol consideration upon which GCRS and LCRS is updated is out of 400 the scope of this proposal and will be illustrated in another draft. 402 5.3. Hybrid control plane 404 It should be more efficient to update the GCRS by a distributed way 405 than a centralized way in terms of routing request and response in a 406 limited network and cloud domain, but be the opposite case in a 407 nationwide circumstance. This is how hybrid control plane could be 408 deployed in such a scheme that overall optimization is achieved. 410 6. Data plane 411 6.1. CSI encapsulation 413 Computing service identification is the predominant index across the 414 entire computing delivery in routing network architecture under which 415 a new virtual routing sub-layer is employed with CSI working as the 416 virtual destination. Data plane indicates the routing and forwarding 417 orientation with CSI by inquiring GCRS and LCRS at GCR and LCR 418 respectively. CSI encapsulation could be achieved by extending the 419 existing packet header and also achieved by designing a dedicated 420 shim layer, which along with the specific structure of CSI are out of 421 the scope of this proposal and will be illustrated in another draft. 423 6.2. CSI for GCR, CUR and LCR 425 GCR encapsulates CSI in a designated header format as a proxy by 426 translating the user-originated CSI format, and makes the first 427 segment routing policy and starts routing and forwarding the service 428 traffic. CUR ignores CSI and simply forwards the traffic as usual. 429 LCR decapsulates CSI and makes the second segment routing policy and 430 completes the last hop routing and forwarding. 432 7. Summary 434 It would significantly benefit the industry by connecting and 435 coordinating the distributed computing resources and services and 436 more so by further converging networking and computing resource. 437 Uncertainty and the potential impacts over the ongoing network 438 architecture is the main reason for the community to think twice. By 439 classifying the end to end routing and forwarding path into two 440 segments, the impacts from computing status and metrics are to be 441 reduced to a degree they would be as acceptable and comfortable 442 enough as they are as networking status and metrics. In particular, 443 employment of CSI in computing delivery in routing network 444 architecture enables a new service routing possibility perfectly 445 compatible with the ongoing routing architecture. 447 8. Acknowledgements 449 To be added upon contributions, comments and suggestions. 451 9. IANA Considerations 453 This memo includes no request to IANA. 455 10. Security Considerations 457 As information originated from the third party (cloud sites), both 458 GCRS and LCRS would be frequently updated in the network domain, both 459 security threats against the routing mechanisms and credibility and 460 security issues of the computing services should be taken into 461 account by architecture designing. The detailed analysis as well as 462 solution consideration will be proposed in the updated version of the 463 draft. 465 11. Informative References 467 [I-D.li-dyncast-architecture] 468 Li, Y., "Dynamic-Anycast Architecture", February 2021, 469 . 472 [I-D.liu-dyncast-ps-usecases] 473 Liu, Peng., "Dynamic-Anycast (Dyncast) Use Cases and 474 Problem Statement", February 2021, 475 . 478 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 479 Requirement Levels", BCP 14, RFC 2119, 480 DOI 10.17487/RFC2119, March 1997, 481 . 483 Authors' Addresses 485 Daniel Huang 486 ZTE Corporation 487 Nanjing 488 Phone: +86 13770311052 489 Email: huang.guangping@zte.com.cn 491 Bin Tan 492 ZTE Corporation 493 Nanjing 494 Phone: +86 13918622159 495 Email: tan.bin@zte.com.cn 497 Peng Liu 498 China Mobile 499 Beijing 500 Phone: +86 13810146105 501 Email: liupengyjy@chinamobile.com