idnits 2.17.1 draft-fedyk-ccamp-uni-extensions-03.txt: Checking boilerplate required by RFC 5378 and the IETF Trust (see https://trustee.ietf.org/license-info): ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/1id-guidelines.txt: ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/checklist : ---------------------------------------------------------------------------- ** The abstract seems to contain references ([RFC4208], [RFC4847]), which it shouldn't. Please replace those with straight textual mentions of the documents in question. Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year -- The document date (October 21, 2013) is 3840 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) -- Obsolete informational reference (is this intentional?): RFC 5996 (Obsoleted by RFC 7296) -- Obsolete informational reference (is this intentional?): RFC 4835 (Obsoleted by RFC 7321) Summary: 1 error (**), 0 flaws (~~), 1 warning (==), 3 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 Network Working Group D. Fedyk 3 Internet Draft Hewlett-Packard 4 Intended status: Standards Track D. Beller 5 L. Levrau 6 Alcatel-Lucent 7 D. Ceccarelli 8 Ericsson 9 F. Zhang 10 Huawei Technologies 11 Y. Tochio 12 Fujitsu 13 X. Fu 14 ZTE 16 Expires: April 24, 2014 October 21, 2013 18 UNI Extensions for Diversity and Latency Support 19 draft-fedyk-ccamp-uni-extensions-03.txt 21 Status of this Memo 23 This Internet-Draft is submitted in full conformance with the 24 provisions of BCP 78 and BCP 79. 26 Internet-Drafts are working documents of the Internet Engineering 27 Task Force (IETF), its areas, and its working groups. Note that 28 other groups may also distribute working documents as Internet- 29 Drafts. 31 Internet-Drafts are draft documents valid for a maximum of six months 32 and may be updated, replaced, or obsoleted by other documents at any 33 time. It is inappropriate to use Internet-Drafts as reference 34 material or to cite them other than as "work in progress." 36 The list of current Internet-Drafts can be accessed at 37 http://www.ietf.org/ietf/1id-abstracts.txt 39 The list of Internet-Draft Shadow Directories can be accessed at 40 http://www.ietf.org/shadow.html 42 Copyright Notice 44 Copyright (c) 2013 IETF Trust and the persons identified as the 45 document authors. All rights reserved. 47 This document is subject to BCP 78 and the IETF Trust's Legal 48 Provisions Relating to IETF Documents 49 (http://trustee.ietf.org/license-info) in effect on the date of 50 publication of this document. Please review these documents 51 carefully, as they describe your rights and restrictions with respect 52 to this document. Code Components extracted from this document must 53 include Simplified BSD License text as described in Section 4.e of 54 the Trust Legal Provisions and are provided without warranty as 55 described in the Simplified BSD License. 57 Abstract 59 This document builds on the GMPLS overlay model [RFC4208] and defines 60 extensions to the GMPLS User-Network Interface (UNI) to support route 61 diversity within the core network for sets of LSPs initiated by edge 62 nodes. A particular example where route diversity within the core 63 network is desired, are dual-homed edge nodes. The document also 64 defines GMPLS UNI extensions to deal with latency requirements for 65 edge node initiated LSPs. 67 This document uses a VPN model that is based on the same premise as 68 L1VPN framework [RFC4847] but may also be applied to other 69 technologies. The extensions are applicable both to VPN and non VPN 70 environments. These extensions move the UNI from basic connectivity 71 to enhanced mode connectivity by including additional constraints 72 while minimizing the exchange of CE to PE information. These 73 extensions are applicable to the overlay extension service model. 74 Route Diversity for customer LSPs are a common requirement applicable 75 to L1VPNs. The UNI mechanisms described in this document are L1VPN 76 compatible and can be applied to achieve diversity for sets of 77 customer LSPs. 79 The UNI extensions in support of latency constraints can also be 80 applied to the extended overlay service model in order for the 81 customer LSPs to meet certain latency requirements. 83 Table of Contents 85 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 3 86 2. Conventions used in this document . . . . . . . . . . . . . . . 4 87 3. Contributors . . . . . . . . . . . . . . . . . . . . . . . . . 4 88 4. LSP Diversity in the Overlay Extension Service Model . . . . . 4 89 4.1. LSP diversity for dual-homed customer edge (CE) devices . . 5 90 4.1.1. Exchanging SRLG information between the PEs via the 91 CE device . . . . . . . . . . . . . . . . . . . . . . . 8 92 4.1.1.1. Operational Procedures . . . . . . . . . . . . . . 8 93 4.1.1.2. Error Handling Procedures . . . . . . . . . . . . . 9 94 4.1.2. Using Path Affinity Set Extension . . . . . . . . . . . 10 95 4.1.2.1. Operational Procedures . . . . . . . . . . . . . . 13 96 4.1.2.2. Error Handling Procedures . . . . . . . . . . . . . 13 97 4.1.2.3. Distribution of the Path Affinity Set Information . 14 98 4.2. Multi-domain LSP Diversity Aspects for Dual-homed CE 99 Devices . . . . . . . . . . . . . . . . . . . . . . . . . . 15 100 4.2.1 Subdividing Identifier Spaces into Ranges . . . . . . . 15 101 4.2.2 Scoping Identifier Spaces to Domains . . . . . . . . . . 15 102 5. Latency Signaling Extensions . . . . . . . . . . . . . . . . . 16 103 5.1. RSVP-TE Extensions . . . . . . . . . . . . . . . . . . . . 17 104 5.2. Operational Procedures . . . . . . . . . . . . . . . . . . 18 105 5.3. Error Handling Procedures . . . . . . . . . . . . . . . . . 18 106 6. Security Considerations . . . . . . . . . . . . . . . . . . . . 18 107 7. IANA Considerations . . . . . . . . . . . . . . . . . . . . . . 19 108 8. References . . . . . . . . . . . . . . . . . . . . . . . . . . 19 109 8.1. Normative References . . . . . . . . . . . . . . . . . . . 19 110 8.2. Informative References . . . . . . . . . . . . . . . . . . 20 111 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . . 21 113 1. Introduction 115 This document builds on the GMPLS overlay model [RFC4208] and defines 116 extensions to the GMPLS User-Network Interface (UNI) to support route 117 diversity within the core network for sets of LSPs initiated by edge 118 nodes. In the following, the term customer edge (CE) device is used 119 synonymously for the term edge node (EN) as in [RFC4208]. 121 Moreover, the VPN terminology (CE and PE) [RFC4026] is used below 122 when the core network is a VPN but is also applicable to UNI 123 interfaces [RFC4208]. 125 This document uses a VPN model that is based on the same premise as 126 L1VPN framework [RFC4847] but may also be applied to other 127 technologies. The extensions are applicable both to VPN and non VPN 128 environments. These extensions move the UNI from basic connectivity 129 to enhanced mode connectivity by including additional constraints 130 while minimizing the exchange of CE to PE information. These 131 extensions are applicable to the overlay extension service model. 133 The overlay model assumes a UNI interface between the edge nodes of 134 the respective transport domains. Route diversity for LSPs from 135 single homed CE and dual-home CEs is a common requirement in optical 136 transport networks. This document describes two signaling variations 137 that may be used for supporting LSP diversity within the overlay 138 extension service model considering dual-homing. Dual-homing is 139 typically used to avoid a single point of failure (UNI link, PE) or 140 if two disjoint connections are forming a protection group in the CE 141 device, e.g., 1+1 protection. While both methods are similar in that 142 they utilize common mechanisms in the PE network to achieve 143 diversity, they are distinguished according to whether the CE is 144 permitted to retrieve provider SRLG diversity information for an LSP 145 from a PE1 and pass it on to a PE2 (SRLG information is shared with 146 the CE), or whether a new attribute is used that allows the PE2 that 147 receives this attribute to derive the SRLG information for an LSP 148 based on the attribute value. Figure 1 below is depicting the 149 scenario. 151 The extended overlay service model can support other extensions for 152 VPN signaling, for example, those related to latency. When requesting 153 diverse LSPs, latency may also be an additional requirement. 155 2. Conventions used in this document 157 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 158 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 159 document are to be interpreted as described in RFC-2119 [RFC2119]. 161 In this document, these words will appear with that interpretation 162 only when in ALL CAPS. Lower case uses of these words are not to be 163 interpreted as carrying RFC-2119 significance. 165 3. Contributors 167 The Authors would like to thank Eve Varma and Sergio Belotti for 168 their review and contributions to this document. 170 4. LSP Diversity in the Overlay Extension Service Model 172 The L1VPN Framework [RFC4847] (Enhanced Mode) describes the overlay 173 extension service model, which builds upon the UNI Overlay [RFC4208] 174 serving as the interface between the CE edge node and the PE edge 175 node. In this service model, a CE receives a list of CE-PE TE link 176 addresses to which it can request a L1VPN connection (i.e., 177 membership information) and may include additional information 178 concerning these TE links. This document further builds on the 179 overlay extension service model by adding shared constraint 180 information for path diversity in the optical transport network. 181 While the L1VPN for optical transport is an example specific VPN 182 technology the term VPN is used generically since the extensions can 183 apply to GMPLS UNIs and VPNs for other technologies. 185 Two signaling variations are outlined here that may be used for 186 supporting LSP diversity within the overlay extension service model 187 considering dual-homing. While both methods utilize common 188 mechanisms in the PE network to achieve diversity, they are 189 distinguished according to whether the CE is permitted to retrieve 190 provider SRLG diversity information for an LSP from a PE1 and pass it 191 on to a PE2 (SRLG information is shared with the CE or whether a new 192 attribute is used that allows the PE2 that receives this attribute to 193 derive the SRLG information for an LSP based on this attribute value. 194 The selection between these methods is governed by both PE-network 195 specific policies and approaches taken (i.e., in terms of how the 196 provider chooses to perform routing internal to their network). 198 The first method (see 4.1.1) assumes that provider Shared Resource 199 Link Group (SRLG) Identifier information is both available and 200 shareable (policy decision) with the CE. Since SRLG IDs can then be 201 used (passed transparently between PEs via the dual-homed CE) as 202 signaled information on a UNI message, a mechanism supporting LSP 203 diversity for the overlay extension service model can be provided via 204 straightforward signaling extensions. 206 The second method (see 3.1.2) assumes that provider SRLG IDs are 207 either not available or not shareable (based on provider network 208 operator policy) with the CE. For this case, a mechanism is provided 209 where information signaled to the PE on UNI messages does not require 210 shared knowledge of provider SRLG IDs to support LSP diversity for 211 the overlay extension model. 213 While both methods could be implemented in the same PE network, it is 214 likely that a GMPLS VPN CE network would use only one mechanism at a 215 time. 217 4.1. LSP diversity for dual-homed customer edge (CE) devices 219 Single-homed CE devices are connected to a single PE device via a 220 single UNI link (could be a bundle of parallel links which are 221 typically using the same fiber cable). This single UNI link may 222 constitute a single point of failure. Such a single point of failure 223 can be avoided when the CE device is connected to two PE devices via 224 two UNI interfaces as depicted for CE1 in Figure 1 below. 226 For the dual-homing case, it is possible to establish two connections 227 from the source CE device to the same destination CE device where one 228 connection is using one UNI link to, for example, PE1 and the other 229 connection is using the UNI link to PE2. In order to avoid single 230 points of failure within the provider network, it is necessary to 231 also ensure path (LSP) diversity within the provider network in order 232 to achieve end-to-end diversity for the two LSPs between the two CE 233 devices. This document describes how it is possible to enable such 234 path diversity to be achieved within the provider network (which is 235 subject to additional routing constraints). [RFC4202] defines SRLG 236 information that can be used to allow GMPLS to provide path diversity 237 in a GMPLS controlled transport network. As the two connections are 238 entering the provider network at different PE devices, the PE device 239 that receives the connection request for the second connection needs 240 to be capable of determining the additional path computation 241 constraints such that the path of the second LSP is disjoint with 242 respect to the already established first connection entering the 243 network at a different PE device. The methods described in this 244 document allow a PE device to determine the SRLG information for a 245 connection in the provider network that is entering the network on a 246 different PE device. 248 PE SRLG information can be used directly by a CE if the CE 249 understands the context, and the CE view is limited to its VPN 250 context. In this case, there is a dependency on the provider 251 information and there is a need to be able to query the SRLG in the 252 provider network. 254 It may, on the other hand, be preferable to avoid this dependency and 255 to decouple the SRLG identifier space used in the provider network 256 from the SRLG space used in the client network. This is possible with 257 both methods detailed below. Even for the method where provider SRLG 258 information is passing through the CE device (note the CE device does 259 not need to process and decode this information) the two SRLG 260 identifier spaces can remain fully decoupled and the operator of the 261 client network is free to assign SRLG identifiers from the client 262 SRLG identifier space to the CE to CE connection that is passing 263 through the provider network. 265 Referring to Figure 1, the UNI signaling mechanism must support at 266 least one of the two mechanisms described in this document for CE 267 dual homing to achieve LSP diversity in the provider network. 269 The described mechanisms can also be applied to a scenario where two 270 CE devices are connected to two different PE devices. In this case, 271 the additional information that is exchanged across the UNI 272 interfaces also needs to be exchanged between the two CE devices in 273 order to achieve the desired diversity in the provider network. 275 This information may be configured or exchanged by some automated 276 mechanism not described in this document. 278 In the dual-homing example, CE1 can locally correlate the LSP 279 requests. For the slightly more complicated example involving CE2 and 280 CE3, both requiring a path that shall be diverse to a connection 281 initiated by the other CE device, CE2 and CE3 need to have a common 282 view of the SRLG information to be signaled. In this document, we 283 detail the required diversity information and the signaling of this 284 diversity information; however, the means for distributing this 285 information within the PE domain or the CE domain is out of scope. 287 +---+ +---+ 288 | P |....| P | 289 +---+ +---+ 290 / \ 291 +-----+ +-----+ +---+ 292 +---+ | PE1 | | |----| | 293 |CE1|----| | | | |CE2| 294 +---+\ +-----+ | |----| | 295 \ | | PE3 | +---+ 296 \ +-----+ | | 297 \| PE2 | | | +---+ 298 | | | |----|CE3| 299 +-----+ +-----+ +---+ 300 \ / 301 +---+ +---+ 302 | P |....| P | 303 +---+ +---+ 305 Figure 1 Overlay Reference Diagram 307 In an overlay model, the information exchanged between the CE and the 308 PE is kept to a minimum. 310 How diversity is achieved, in terms of configuration, distribution 311 and usage in each part of the transport networks should be kept 312 independent and separate from how diversity is signaled at the UNI 313 between the two transport networks. 315 Signaling parameters discussed in this document are: 317 o SRLG information (see [RFC4202]) 319 o Path Affinity Set 321 4.1.1. Exchanging SRLG information between the PEs via the CE device 323 SRLG information is defined in [RFC4202] and if the SRLG information 324 of an LSP is known, it can be used to calculate a path for another 325 LSP that is SRLG diverse with respect to an existing LSP. SRLG 326 information is an unordered list of SRLGs. SRLG information is 327 normally not shared between the transport network and the client 328 network; i.e., not shared with the CEs of a VPN in the VPN context. 329 However, this becomes more challenging when a CE is dual-homed. For 330 example, CE1 in Figure 1 may have requested an LSP1 from CE1 to CE2 331 via PE1 and PE3. CE1 could subsequently request an LSP2 to CE2 via 332 PE2 and PE3 with the requirement that it should be maximally SRLG 333 disjoint with respect to LSP1. Since PE2 does not have any 334 information about LSP1, PE2 would need to know the SRLG information 335 associated with LSP1. If CE1 could request the SRLG information of 336 LSP1 from PE1, it could then transparently pass this information to 337 PE2 as part of the LSP2 setup request, and PE2 would now be capable 338 of calculating a path for LSP2 that is SRLG disjoint with respect to 339 LSP1. 341 The exchange of SRLG information is achieved on a per VPN LSP basis 342 using the existing RSVP-TE signaling procedures. It can be exchanged 343 in the PATH (exclusion information) or RESV message in the original 344 request or it can be requested by the CE at any time the path is 345 active. 347 It shall be noted that SRLG information is an unordered list of SRLG 348 identifiers and the encoding of SRLG information for RSVP signaling 349 is already defined in [SRLG_info]. Even if SRLG information is known 350 for several LSPs it is not possible for the CEs to derive the 351 provider network topology from this information. 353 4.1.1.1. Operational Procedures 355 Retrieving SRLG information from a PE for an existing LSP: 357 When a dual-homed CE device intends to establish an LSP to the same 358 destination CE device via another PE node, it can request the SRLG 359 information for an already established LSP by setting the SRLG 360 information flag in the LSP attributes sub-object of the RSVP PATH 361 message (IANA to assign the new SRLG flag). As long as the SRLG 362 information flag is set in the PATH message, the PE node inserts the 363 SRLG sub-object as defined in [SRLG_info] into the RSVP RESV message 364 that contains the current SRLG information for the LSP. If the 365 provider network's policy has been configured so as not to share SRLG 366 information with the client network, the SRLG sub-object is not 367 inserted in the RESV message even if the SRLG information flag was 368 set in the received PATH message. Note that the SRLG information is 369 expected to be always up-to-date. 371 Establishment of a new LSP with SRLG diversity constraints: 373 When a dual-homed CE device sends an LSP setup requests to a PE 374 device for a new LSP that is required to be SRLG diverse with respect 375 to an existing LSP that is entering the network via another PE 376 device, the CE device sets the SRLG diversity flag (note: IANA to 377 assign the new SRLG diversity flag) in the LSP attributes sub-object 378 of the PATH message that initiates the setup of this new LSP. When 379 the PE device receives this request it calculates a path to the given 380 destination and uses the received SRLG information as path 381 computation constraints. 383 4.1.1.2. Error Handling Procedures 385 When the CE device receives a RSVP PATH message with the SRLG 386 information flag set and if the provider's network policy does not 387 permit sharing of SRLG information, the PE device shall notify the CE 388 device by sending a RSVP PathErr with a Notify error code (error code 389 to be defined) "Retrieval of SRLG information not permitted". As 390 described above, the PE device must not include the SRLG sub-object 391 with the SRLG information for the LSP in the RSVP RESV message. 393 If the PE device receives a RSVP PATH message for a new LSP with the 394 SRLG diversity flag set and SRLG information in the SRLG sub-object, 395 the PE device tries to calculate a route to the given destination 396 that is SRLG diverse with respect to the provided SRLG information. 397 If no route can be found, a RSVP PathErr message with an error code 398 (error code to be defined) "No SRLG diverse route available toward 399 destination". 401 If the PE device receives a RSVP PATH message for a new LSP with the 402 SRLG diversity flag set and SRLG information in the SRLG sub-object 403 and if the PE device does not support the SRLG sub-object, the PE 404 device shall send a PathErr message to the CE device, indicating an 405 "Unknown object class". 407 Further error handling cases will be added in the next revision of 408 this document. 410 4.1.2. Using Path Affinity Set Extension 412 The Path Affinity Set (PAS) is used to signal diversity in a pure CE 413 context by abstracting SRLG information. There are two types of 414 diversity information in the PAS. The first type of information is a 415 single PAS identifier. The Second part is the optional PATH 416 information, in the form of Source and Destination addresses of an 417 exclude path or set of paths that MAY be specified. The motive behind 418 the PAS information is to have as little exchange of diversity 419 information as possible between the VPN CE and PE elements. 421 Rather than a detailed CE or PE SRLG list, the Path Affinity Set 422 contains an abstract SRLG identifier that associates the given path 423 as diverse. Logically the identifier is in a VPN context and 424 therefore only unique with respect to a particular VPN. 426 How the CE determines the PAS identifier is a local matter for the CE 427 administrator. A CE may signal the PAS identifier as a diversity 428 object in the PATH message. This identifier is a suggested identifier 429 and may be overridden by a PE under some conditions. 431 For example, a PAS identifier can be used with no prior exchange of 432 PAS information between the CE and the PE. Upon reception of the PAS 433 identifier information the PE can infer the CEs requirements. The 434 actual PAS identifier used will be returned in the RESV message. 435 Optionally an empty PAS identifier allows the PE to pick the PAS 436 identifier. 438 Similar to the section 4.1.1 on SRLG information, a PE can return PAS 439 identifier as the response to a Query allowing flexibility. 441 A PE interprets the specific PAS identifier, for example, "123" as 442 meaning to exclude the PE SRLG information (or equivalent) that has 443 been allocated by LSPs associated with this Path Affinity Set 444 identifier "123", for any LSPs associated with the resources assigned 445 to the VPN. For example, if a Path exists for the LSP with the 446 identifier "123", the PE would use local knowledge of the PE SRLGs 447 associated with the "123" LSPs and exclude those SRLGs in the path 448 request. In other words, two LSPs that need to be diverse both 449 signal "123" and the PEs interpret this as meaning not to use shared 450 resources. Alternatively, a PE could use the PAS identifier to 451 select from already established LSPs. Once the path is established it 452 becomes the "123" identifier or optionally another PAS identifier for 453 that VPN that replaces "123". 455 The optional PAS Source and Destination Address tuple represents one 456 or more source addresses and destination addresses associated with 457 the CE Path Affinity Set identifier. These associated address tuples 458 represent paths that use resources that should be excluded for the 459 establishment of the current LSP. The address tuple information 460 gives both finer grain details on the path diversity request and 461 serves as an alternative identifier in the case when the PAS 462 identifier is not known by the PE. The address tuples used in 463 signaling is within a CE context and its interpretation is local to a 464 PE that receives a Path request from a CE. The PE can use the address 465 information to relate to PE Addresses and PE SRLG information. When 466 a PE satisfies a connection setup for a (SRLG) diverse signaled path, 467 the PE may optionally record the PE SRLG information for that 468 connection in terms of PE based parameters and associate that with 469 the CE addresses in the Path message. 471 Specifically for L1VPNs, Port Information table (PIT) [RFC5251] can 472 be leveraged to translate between CE based addresses and PE based 473 addresses. The Path Affinity Set and associated PE addresses with PE 474 SRLG information can be distributed via the IGP in the provider 475 transport network (or by other means such as configuration); they can 476 be utilized by other PEs when other CE Paths are setup that would 477 require path/connection diversity. This information is distributed on 478 a VPN basis and contains a PAS identifier, PE addresses and SRLG 479 information. 481 If diversity is not signaled, the assumption is that no diversity is 482 required and the Provider network is free to route the LSP to 483 optimize traffic. No Path affinity set information needs to be 484 recorded for these LSPs. If a diversity object is included in the 485 connection request, the PE in the Provider Network should be able to 486 look-up the existing Provider SRLG information from the provider 487 network and choose an LSP that is maximally diverse from other LSPs. 489 The mechanisms to achieve this are outside the scope of this 490 document. 492 A new VPN Diverse LSP LABEL object is specified: 494 0 1 2 3 495 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 496 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 497 | Length | Type (TBA) |0| C-type (TBA)| 498 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 500 1 2 3 501 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 503 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 504 | ADDR Length |Number of PAS |D| reserved | 505 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 506 | Path Affinity Set identifier | 507 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 508 | Source Address (variable) | 509 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 510 | Destination Address (variable) | 511 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 513 Figure 2 Diverse LSP information 515 1. The Address Length field (8 bits) is the number of bytes for both 516 the source address and destination address. The address may be in 517 any format from 1 to 32 bytes but the key point is the customers 518 can maintain their existing addresses. A value of zero indicates 519 there are no addresses included. 521 2. The Number of Path Affinity (8 bits) sets is included in the 522 object. This is typically 1. Addition of other sets is for further 523 study. 525 3. The Path affinity Set identifier (4 bytes) is a single number that 526 represents a summarized SRLG for this path. Paths with that same 527 Path Affinity set should be set up with diverse paths and 528 associated with the path affinity set. A value of all zeros 529 allows the PE to pick a PAS identifier to return. A PAS 530 identifier of an established path may be different than the 531 requested path identifier. 533 4. The diversity Bit (D) (one Bit) indicates if the diversity must be 534 satisfied when set as a one. If a PE finds an established path 535 with a Path Affinity set matching the signaled Path Affinity Set 536 or the signaled Address tuple it should attempt find a diverse 537 path. 539 5. The Diverse Path Source address/destination address tuple is that 540 of an established LSP in the PE network that belongs to the same 541 Path Affinity Set identifier. If the path for these addresses is 542 not established or cannot be determined by the PE edge processing 543 the PATH request then the path is established only with the Path 544 Affinity identifier. If the path(s) for these address tuples are 545 known by the PE the PE uses the SRLG information associated with 546 these addresses. If in any case a diverse path cannot be setup 547 then the Diverse bit controls whether a path is established 548 anyway. The PE must use the PIT to translate CE Addresses into 549 provider addresses when correlating with provider SRLG 550 information. How SRLG information and network address tuples are 551 distributed is for future study. 553 4.1.2.1. Operational Procedures 555 When a CE constructs a PATH message it may optionally specify and 556 insert a Path Affinity Set in the PATH message. This Path Affinity 557 Set may optionally include the address of an LSP that that could 558 belong to the same Path Affinity Set. The Path Affinity Set 559 identifier is a value (0 through 2**32-255) that is independent of 560 the mechanism the CE or the PE use for diversity. The Path Affinity 561 Set is a single identifier that can be used to request diversity and 562 associate diversity. 564 When processing a CE PATH message in a VPN Overlay, the PE first 565 looks up the PE based addresses in the Provider Index Table (PIT). If 566 the Path Affinity Set is included in the PATH message, the PE must 567 look up the SRLG information (or equivalent) in the PE network that 568 has been allocated by LSPs associated with a Path Affinity Set and 569 exclude those resources from the path computation for this LSP if it 570 is a new path. The PE may alternatively choose from an existing path 571 with a disjoint set of resources. If a path that is disjoint cannot 572 be found, the value of the PAS diversity bit determines whether a 573 path should be setup anyway. If the PAS diversity bit is clear, one 574 can still attempt to setup the LSP. A PE should still attempt to 575 minimize shared resources but that is an implementation issue, and is 576 outside the scope of this document. 578 Optionally the CE may use a value of all zeros in the PAS identifier 579 allowing the PE to select an appropriate PAS identifier. Also the PE 580 may to override the PAS identifier allowing the PE to re-assign the 581 identifier if required. A CE should not assume that the PAS 582 identifier used for setup is the actual PAS identifier. 584 4.1.2.2. Error Handling Procedures 585 The PAS object must be understood by the PE device. Otherwise, the CE 586 should not use the PAS object. Path Message processing of the PAS 587 object SHOULD follow CTYPE 0. An Error code of IANA (TBD) indicates 588 that the PAS object is not understood. 590 When a PAS identifier is not recognized by a PE it must assume this 591 LSP defines that PAS identifier however the PE may override PAS 592 identifier under certain conditions. 594 If the identifier is recognized but the Source Address-Destination 595 address pair(s) are not recognized, this LSP must be set up using the 596 PAS identifier only. 598 If the identifier is recognized and the Source Address-Destination 599 address pair(s) are also recognized, then the PE SHOULD use the PE 600 SRLG information associated with the LSPs identified by the address 601 pairs to select a disjoint path. 603 The Following are the additional error codes: 605 1. Route Blocked by Exclude Route Value IANA (TBA). 607 4.1.2.3. Distribution of the Path Affinity Set Information 609 Information about SRLG is already available in the IGP TE database. A 610 PE network can be designed to have additional opaque records for 611 Provider paths that distribute PE paths and SRLG on a VPN basis. When 612 a PE path is setup, the following information allows a PE to lookup 613 the PE diversity information: 615 o L1 VPN Identifier 8 bytes 617 o Path Affinity Set Identifier 619 o Source PE Address 621 o Destination PE Address 623 o List of PE SRLG (variable) 625 The source PE address and destination PE address are the same 626 addresses in the VPN PIT and correspond to the respective CE address 627 identifiers. 629 Note that all of the information is local to the PE context and is 630 not shared with the CE. The VPN Identifier is associated with a CE. 631 The only value that is signaled from the CE is the Path Affinity Set 632 and optionally the addresses of an existing LSP. The PE stores source 633 and destination PE addresses of the LSP in their native format along 634 with the SRLG information. This information is internal to the PE 635 network and is always known. 637 PE paths may be setup on demand or they may be pre-established. When 638 paths are pre-established, the Path Affinity Set is set to unassigned 639 0x0000 and is ignored. When a CE uses a pre-established path the PE 640 may set the Path SRLG Path Affinity Set value if the CE signals one 641 otherwise the Path Affinity Set remains unassigned 0x0000. 643 4.2. Multi-domain LSP Diversity Aspects for Dual-homed CE Devices 645 The two mechanisms described above to achieve LSP diversity for 646 dual-homed CE devices can be applied to single-domain provider 647 networks as well as multi-domain provider networks. This section 648 addresses multi-domain aspects including both single provider multi- 649 domain networks and multi-provider networks where the subdivision 650 into multiple domains is obvious due to the organizational boundaries 651 between different providers. Specifically, when multiple providers 652 are involved, SRLG identifiers as well as PAS identifiers must be 653 administrable independently for each provider network. 655 For the single provider multi-domain case, there are two 656 possibilities how SRLG or PAS identifiers can be handled: 658 o Subdividing the identifier space into ranges assigned to domains 660 o Scoping the identifiers to domains 662 4.2.1 Subdividing Identifier Spaces into Ranges 664 Subdividing the identifier space into disjoint ranges and assigning 665 the different ranges to the different domain is one possibility to 666 apply the LSP diversity mechanisms defined in this document to a 667 multi-domain environment. This does not require additional protocol 668 extensions. Caution is, however, required when the identifiers are 669 assigned. They must be selected strictly from the identifier range 670 that has been assigned to the specific domain. From a network 671 operations perspective, this can be an option for a single provider 672 multi-domain network while it may be less applicable to multi- 673 provider networks where minimal dependency is desired. 675 4.2.2 Scoping Identifier Spaces to Domains 677 [DRAFT DOMAIN SUBOBJECTS] defines new RSVP-TE domain sub-objects for 678 the purpose of identifying domains. Domain sub-objects can be used to 679 scope SRLG or PAS identifiers to a specific domain. With this 680 extension, the full SRLG or PAS identifier space can be used within 681 each domain. When a new multi-domain LSP shall be established, the 682 diversity constraints can be signaled in the form of a sequence of a 683 scoping domain sub-object followed by the list of SRLGs or the PAS 684 object, e.g.: [domain_sub-object(Dn), SRLG_sub-object(Dn)] for domain 685 Dn. 687 5. Latency Signaling Extensions 689 Some network applications are sensitive to latency (sometimes also 690 called delay) while other applications are sensitive to latency 691 variation (sometimes also called delay variation). Specifically, real 692 time applications typically do have certain latency requirements. It 693 shall be noted that latency variation is typically not an issue for 694 TDM networks including the WDM layer. For these technologies the 695 latency is constant and there is no latency variation added. Latency 696 variation is typically caused in packet networks or when packet based 697 services are encapsulated into a constant bit rate server layer 698 signal, which requires buffering of the arriving packets that may 699 arrive in bursts. An example is an Ethernet VLAN service that is 700 mapped into a constant bit rate server layer such as an ODUk or 701 ODUflex OTN signal. 703 The GMPLS UNI as defined in [RFC4208] does not support latency as a 704 signaling parameter that would allow a CE device to signal to the PE 705 device that latency and/or latency variation constraints need to be 706 met when a path is calculated for the requested LSP. The path 707 computation function does typically calculate a route to the given 708 destination that has the least TE metric (least cost routing). 709 However, if a CE device requests an LSP via the UNI interface for an 710 application that is sensitive to latency/latency variation, it should 711 be possible to signal to the PE device that the objective function 712 should rather take latency into account instead of the TE metric. 714 In order to support latency/latency variation as path computation 715 constraint, the network has to support latency/latency variation as 716 TE metric extension as defined in [DRAFT OSPF TE METRIC EXT] - note 717 that [DRAFT OSPF TE METRIC EXT] is using the terms delay/delay 718 variation instead of latency/latency variation. 720 A latency requirement can be added to signaling in the form of a 721 constraint [DRAFT OBJECTIVE FUNCTION]. The constraint can take the 722 form of: 724 o Minimal latency 726 o Maximum acceptable latency (upper bound) 728 o Minimal latency variation 730 o Maximum acceptable latency variation (upper bound), if applicable 732 While some systems may be able to compute routes based on delay 733 metrics it is usual that minimizing the accumulated TE link metric 734 (link cost) or the number of hops subject to bandwidth reservation 735 are satisfied as the object function and delay is not considered. 736 When considering diversity latency falls after diversity constraints 737 have been satisfied. 739 Recording the latency of existing paths [DRAFT TE METRIC RECORD] to 740 ensure they meet a maximum acceptable latency can be utilized to 741 ensure latency constraint is met. 743 When a low latency path is required, the minimize latency subject to 744 other constraints criteria should be signaled. A CE device can use 745 the recorded latency to ensure that the maximum acceptable latency 746 has been met. 748 5.1. RSVP-TE Extensions 750 At the UNI, the RSVP-TE extensions as defined in [DRAFT OBJECTIVE 751 FUNCTION] SHALL be used for signaling the PE device whether a path 752 with minimal latency is requested or whether certain latency/latency 753 variation upper bound constraints shall be met for the end-to-end 754 connection, i.e., from the source CE device to the destination CE 755 device. The following objective function (OF) code point SHALL be 756 used in the OF sub-object of the ERO to indicate that latency/latency 757 variation constraints SHALL be taken into account when the path 758 computation function that is invoked by the PE node that expands the 759 route from the PE device to the destination CE device: 761 o OF code value 8 (to be assigned by IANA) is for the Minimum 762 Latency Path (MLP) OF 764 o OF code value 9 (to be assigned by IANA) is for Minimum Latency 765 Variation Path (MLVP) OF 767 Additionally, an optional OF metric-bound sub-object MAY be carried 768 within an ERO object of the RSVP-TE Path message. The two metric- 769 bound sub-objects defined in [DRAFT OBJECTIVE FUNCTION] that are 770 corresponding to the two OFs above are: 772 o metric bound sub-object of Type T=4: Cumulative Latency 774 o metric bound sub-object of Type T=5: Cumulative Latency Variation 776 The metric-bound indicates an upper bound for the path metric that 777 MUST NOT be exceeded for the ERO expending node to consider the 778 computed path as acceptable. It shall be noted that the metric bound 779 included in the RSVP-TE Path message at the UNI has end-to-end 780 significance, which means that the upper bound metric constraint MUST 781 be met for the path from the source CE device to the destination CE 782 device. 784 5.2. Operational Procedures 786 The processing rules as defined in [DRAFT OBJECTIVE FUNCTION] for the 787 OF sub-object and the optional OF metric-bound sub-object SHALL be 788 applied at the ingress PE device when the source CE device requests 789 an LSP (It shall be noted that [DRAFT OBJECTIVE FUNCTION] has a wider 790 scope and may also apply to inter-domain interfaces, i.e., when the 791 provider network is composed of multiple separate domains.). 793 5.3. Error Handling Procedures 795 The error handling rules as defined in [DRAFT OBJECTIVE FUNCTION] for 796 the OF sub-object and the optional OF metric-bound sub-object SHALL 797 be applied. 799 6. Security Considerations 801 Security for L1VPNs is covered in [RFC4847], [RFC5251] and [RFC5253]. 802 In this document, the model follows a generic GMPLS VPN based on the 803 L1VPN control plane model where CE addresses are completely distinct 804 from the PE addresses. 806 The use of a private network assumes that entities outside the 807 network cannot spoof or modify control plane communications between 808 CE and PE. Furthermore, all entities in the private network are 809 assumed to be trusted. Thus, no security mechanisms are required by 810 the protocol exchanges described in this document. 812 However, an operator that is concerned about the security of their 813 private control plane network may use the authentication and 814 integrity functions available in RSVP-TE [RFC3473] or utilize IPsec 815 ([RFC4301], [RFC4302], [RFC4835], [RFC5996], and [RFC6071]) for the 816 point-to-point signaling between PE and CE. See [RFC5920] for a full 817 discussion of the security options available for the GMPLS control 818 plane. 820 7. IANA Considerations 822 TBD 824 8. References 826 8.1. Normative References 828 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 829 Requirement Levels", BCP 14, RFC 2119, March 1997. 831 [RFC4202] Kompella, K., Rekhter, Y., "Routing Extensions in Support 832 of Generalized Multi-Protocol Label Switching (GMPLS)", RFC 833 4202, October 2005. 835 [RFC4208] Swallow, G., Drake, J., Ishimatsu, H., and Y. Rekhter, 836 "Generalized Multiprotocol Label Switching (GMPLS) User- 837 Network Interface (UNI): Resource ReserVation Protocol- 838 Traffic Engineering (RSVP-TE) Support for the Overlay 839 Model", RFC 4208, October 2005. 841 [RFC5251] Fedyk, D., Rekhter, Y., Editors "Layer 1 VPN Basic Mode", 842 RFC 5251, July 2008. 844 [SRLG_info] Zhang, F., Li, D., Gonzalez de Dios, O., Margaria, C., 845 Hartley, M., "RSVP-TE Extensions for Collecting SRLG 846 Information", draft-ietf-ccamp-rsvp-te-srlg-collect-03.txt, 847 October 2013. 849 [DRAFT OBJECTIVE FUNCTION] Ali, Z., Swallow, G., Filsfils, C., Fang, 850 L., Kumaki, K., Kunze, R., Zhang, X., "Resource ReserVation 851 Protocol - Traffic Engineering (RSVP-TE) extension for 852 signaling Objective Function and Metric Bound", draft-ali- 853 ccamp-rc-objective-function-metric-bound-04.txt, October 854 2013. 856 [DRAFT DOMAIN SUBOBJECTS] Dhody, D., Palle, U., Kondreddy, V., 857 Casellas, R., "Domain Subobjects for Resource ReserVation 858 Protocol - Traffic Engineering (RSVP-TE)", draft-ietf- 859 ccamp-rsvp-te-domain-subobjects-00.txt, October 2013. 861 8.2. Informative References 863 [RFC4026] Andersson, L. and T. Madsen, "Provider Provisioned Virtual 864 Private Network (VPN) Terminology", RFC 4026, March 2005. 866 [RFC6071] Frankel, S. and S. Krishnan, "IP Security (IPsec) and 867 Internet Key Exchange (IKE) Document Roadmap", RFC 6071, 868 February 2011. 870 [RFC3473] Berger, L. (editor), "Generalized MPLS Signaling - RSVP-TE 871 Extensions", RFC 3473, January 2003. 873 [RFC4301] Kent, S. and K. Seo, "Security Architecture for the 874 Internet Protocol", RFC 4301, December 2005. 876 [RFC4302] Kent, S., "IP Authentication Header", RFC 4302, December 877 2005. 879 [RFC5996] Kaufman, C., Hoffman, P., Nir, Y., and P. Eronen, "Internet 880 Key Exchange Protocol Version 2 (IKEv2)", RFC 5996, 881 September 2010. 883 [RFC4835] Manral, V., "Cryptographic Algorithm Implementation 884 Requirements for Encapsulating Security Payload (ESP) and 885 Authentication Header (AH)", RFC 4835, April 2007. 887 [RFC4847] Takeda, T., Editor "Framework and Requirements for Layer 888 Virtual Private Networks", RFC 4847, April 2007. 890 [RFC5253] Takeda, T., Ed., "Applicability Statement for Layer 1 891 Virtual Private Network (L1VPN) Basic Mode", RFC 5253, July 892 2008. 894 [RFC5920] Fang, L., Ed., "Security Framework for MPLS and GMPLS 895 Networks", RFC 5920, July 2010. 897 [DRAFT TE METRIC RECORD] Ali, Z., Swallow, G., Filsfils, C., Hartley, 898 M., Kumaki, K., Kunze, R., "Resource ReserVation Protocol- 899 Traffic Engineering (RSVP-TE) extension for recording TE 900 Metric of a Label Switched Path", draft-ietf-ccamp-te- 901 metric-recording-02.txt, July 2013. 903 [DRAFT OSPF TE METRIC EXT] Giacalone, S., Ward, D., Drake, J., Atlas, 904 A., Previdi, S., "OSPF Traffic Engineering (TE) Metric 905 Extensions", draft-ietf-ospf-te-metric-extensions-04.txt, 906 June 2013. 908 Copyright (c) 2013 IETF Trust and the persons identified as authors 909 of the code. All rights reserved. 911 Redistribution and use in source and binary forms, with or without 912 modification, is permitted pursuant to, and subject to the license 913 terms contained in, the Simplified BSD License set forth in Section 914 4.c of the IETF Trust's Legal Provisions Relating to IETF Documents 915 (http://trustee.ietf.org/license-info). 917 Authors' Addresses 919 Don Fedyk 920 Hewlett-Packard 921 153 Tayor Street 922 Littleton, MA, 01460 923 Email: don.fedyk@hp.com 925 Dieter Beller 926 Alcatel-Lucent 927 Email: Dieter.Beller@alcatel-lucent.com 929 Lieven Levrau 930 Alcatel-Lucent 931 Email: Lieven.Levrau@alcatel-lucent.com 933 Daniele Ceccarelli 934 Ericsson 935 Email: Daniele.Ceccarelli@ericsson.com 936 Fatai Zhang 937 Huawei Technologies 938 Email: zhangfatai@huawei.com 940 Yuji Tochio 941 Fujitsu 942 Email: tochio@jp.fujitsu.com 944 Xihua Fu 945 ZTE 946 Email: fu.xihua@zte.com.cn