idnits 2.17.1 draft-ietf-tsvwg-ecn-mpls-01.txt: Checking boilerplate required by RFC 5378 and the IETF Trust (see https://trustee.ietf.org/license-info): ---------------------------------------------------------------------------- ** It looks like you're using RFC 3978 boilerplate. You should update this to the boilerplate described in the IETF Trust License Policy document (see https://trustee.ietf.org/license-info), which is required now. -- Found old boilerplate from RFC 3978, Section 5.1 on line 17. -- Found old boilerplate from RFC 3978, Section 5.5, updated by RFC 4748 on line 974. -- Found old boilerplate from RFC 3979, Section 5, paragraph 1 on line 985. -- Found old boilerplate from RFC 3979, Section 5, paragraph 2 on line 992. -- Found old boilerplate from RFC 3979, Section 5, paragraph 3 on line 998. 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 Copyright Line does not match the current year -- The document seems to lack a disclaimer for pre-RFC5378 work, but may have content which was first submitted before 10 November 2008. If you have contacted all the original authors and they are all willing to grant the BCP78 rights to the IETF Trust, then this is fine, and you can ignore this comment. If not, you may need to add the pre-RFC5378 disclaimer. (See the Legal Provisions document at https://trustee.ietf.org/license-info for more information.) -- The document date (June 19, 2007) is 6149 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(See RFCs 3967 and 4897 for information about using normative references to lower-maturity documents in RFCs) == Outdated reference: A later version (-03) exists of draft-charny-pcn-single-marking-01 == Outdated reference: A later version (-20) exists of draft-ietf-nsis-rmd-09 == Outdated reference: A later version (-07) exists of draft-ietf-tsvwg-diffserv-class-aggr-02 Summary: 1 error (**), 0 flaws (~~), 4 warnings (==), 7 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 Network Working Group B. Davie 3 Internet-Draft Cisco Systems, Inc. 4 Intended status: Standards Track B. Briscoe 5 Expires: December 21, 2007 J. Tay 6 BT Research 7 June 19, 2007 9 Explicit Congestion Marking in MPLS 10 draft-ietf-tsvwg-ecn-mpls-01.txt 12 Status of this Memo 14 By submitting this Internet-Draft, each author represents that any 15 applicable patent or other IPR claims of which he or she is aware 16 have been or will be disclosed, and any of which he or she becomes 17 aware will be disclosed, in accordance with Section 6 of BCP 79. 19 Internet-Drafts are working documents of the Internet Engineering 20 Task Force (IETF), its areas, and its working groups. Note that 21 other groups may also distribute working documents as Internet- 22 Drafts. 24 Internet-Drafts are draft documents valid for a maximum of six months 25 and may be updated, replaced, or obsoleted by other documents at any 26 time. It is inappropriate to use Internet-Drafts as reference 27 material or to cite them other than as "work in progress." 29 The list of current Internet-Drafts can be accessed at 30 http://www.ietf.org/ietf/1id-abstracts.txt. 32 The list of Internet-Draft Shadow Directories can be accessed at 33 http://www.ietf.org/shadow.html. 35 This Internet-Draft will expire on December 21, 2007. 37 Copyright Notice 39 Copyright (C) The IETF Trust (2007). 41 Abstract 43 RFC 3270 defines how to support the Diffserv architecture in MPLS 44 networks, including how to encode Diffserv Code Points (DSCPs) in an 45 MPLS header. DSCPs may be encoded in the EXP field, while other uses 46 of that field are not precluded. RFC3270 makes no statement about 47 how Explicit Congestion Notification (ECN) marking might be encoded 48 in the MPLS header. This draft defines how an operator might define 49 some of the EXP codepoints for explicit congestion notification, 50 without precluding other uses. 52 Requirements Language 54 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 55 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 56 document are to be interpreted as described in RFC 2119 [RFC2119]. 58 Change History 60 [Note to RFC Editor: This section to be removed before publication] 62 Changes in this version (draft-ietf-tsvwg-ecn-mpls-01.txt) relative 63 to the last (draft-ietf-tsvwg-ecn-mpls-00.txt): 65 o Moved the detailed discussion of marking procedures for Pre- 66 Congestion Notification (PCN) to an appendix. 68 o Removed PCN as a motivation for the efficient code-point usage in 69 Section 2. 71 o Clarified the rationale for preferring the ECT-checking approach 72 over the approach of [Floyd] in Section 9.1. 74 o Updated discussion of relationship to RFC3168 in Section 7 76 o Removed discussion of re-ECN from Security Considerations. 78 o Fixed typos and nits. 80 Changes in draft-ietf-tsvwg-ecn-mpls-00.txt relative to 81 draft-davie-ecn-mpls-00: 83 o Corrected the description of ECN-MPLS marking proposed in 84 [Shayman], which closely corresponds to that proposed in this 85 document. 87 o Pre-congestion notification (PCN) marking is now described in a 88 way that does not require normative references to PCN 89 specifications. PCN discussion now serves only to illustrate how 90 the ECN marking concepts can be extended to cover more complex 91 scenarios, with PCN being an example. 93 o Added specification of behavior when MPLS encapsulated packets 94 cross from an ECN-enabled domain to a domain that is not ECN- 95 enabled. 97 o Clarified that copying MPLS ECN or PCN marking into exposed IP 98 header on egress is not mandatory 100 o Fixed typos and nits 102 Table of Contents 104 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 4 105 1.1. Background . . . . . . . . . . . . . . . . . . . . . . . . 4 106 1.2. Intent . . . . . . . . . . . . . . . . . . . . . . . . . . 4 107 1.3. Terminology . . . . . . . . . . . . . . . . . . . . . . . 5 108 2. Use of MPLS EXP Field for ECN . . . . . . . . . . . . . . . . 6 109 3. Per-domain ECT checking . . . . . . . . . . . . . . . . . . . 8 110 4. ECN-enabled MPLS domain . . . . . . . . . . . . . . . . . . . 8 111 4.1. Pushing (adding) one or more labels to an IP packet . . . 9 112 4.2. Pushing one or more labels onto an MPLS labelled packet . 9 113 4.3. Congestion experienced in an interior MPLS node . . . . . 9 114 4.4. Crossing a Diffserv Domain Boundary . . . . . . . . . . . 9 115 4.5. Popping an MPLS label (not the end of the stack) . . . . . 10 116 4.6. Popping the last MPLS label in the stack . . . . . . . . . 10 117 4.7. Diffserv Tunneling Models . . . . . . . . . . . . . . . . 10 118 5. ECN-disabled MPLS domain . . . . . . . . . . . . . . . . . . . 11 119 6. The use of more codepoints with E-LSPs and L-LSPs . . . . . . 11 120 7. Relationship to tunnel behavior in RFC 3168 . . . . . . . . . 11 121 8. Example Uses . . . . . . . . . . . . . . . . . . . . . . . . . 12 122 8.1. RFC3168-style ECN . . . . . . . . . . . . . . . . . . . . 12 123 8.2. ECN Co-existence with Diffserv E-LSPs . . . . . . . . . . 12 124 8.3. Congestion-feedback-based Traffic Engineering . . . . . . 13 125 8.4. PCN flow admission control and flow pre-emption . . . . . 13 126 9. Deployment Considerations . . . . . . . . . . . . . . . . . . 14 127 9.1. Marking non-ECN Capable Packets . . . . . . . . . . . . . 14 128 9.2. Non-ECN capable routers in an MPLS Domain . . . . . . . . 15 129 10. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 15 130 11. Security Considerations . . . . . . . . . . . . . . . . . . . 15 131 12. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . 16 132 Appendix A. Extension to Pre-Congestion Notification . . . . . . 16 133 Appendix A.1. Label Push onto IP packet . . . . . . . . . . . . . 17 134 Appendix A.2. Pushing Additional MPLS Labels . . . . . . . . . . . 17 135 Appendix A.3. Admission Control or Pre-emption Marking inside 136 MPLS domain . . . . . . . . . . . . . . . . . . . . 17 137 Appendix A.4. Popping an MPLS Label (not end of stack) . . . . . . 17 138 Appendix A.5. Popping the last MPLS Label to expose IP header . . 18 139 13. References . . . . . . . . . . . . . . . . . . . . . . . . . . 18 140 13.1. Normative References . . . . . . . . . . . . . . . . . . . 18 141 13.2. Informative References . . . . . . . . . . . . . . . . . . 19 142 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . . 20 143 Intellectual Property and Copyright Statements . . . . . . . . . . 22 145 1. Introduction 147 1.1. Background 149 [RFC3168] defines Explicit Congestion Notification for IP. The 150 primary purpose of ECN is to allow congestion to be signalled without 151 dropping packets. 153 [RFC3270] defines how to support the Diffserv architecture in MPLS 154 networks, including how to encode Diffserv Code Points (DSCPs) in an 155 MPLS header. DSCPs may be encoded in the EXP field, while other uses 156 of that field are not precluded. RFC3270 makes no statement about 157 how Explicit Congestion Notification (ECN) marking might be encoded 158 in the MPLS header. 160 This draft defines how an operator might define some of the EXP 161 codepoints for explicit congestion notification, without precluding 162 other uses. In parallel to the activity defining the addition of ECN 163 to IP [RFC3168], two proposals were made to add ECN to MPLS 164 [Floyd][Shayman]. These proposals, however, fell by the wayside. 165 With ECN for IP now being a proposed standard, and developing 166 interest in using pre-congestion notification (PCN) for admission 167 control and flow pre-emption [I-D.briscoe-tsvwg-cl-architecture], 168 there is consequent interest in being able to support ECN across IP 169 networks consisting of MPLS-enabled domains. Therefore it is 170 necessary to specify the protocol for including ECN in the MPLS shim 171 header, and the protocol behavior of edge MPLS nodes. 173 We note that in [RFC3168] there are four codepoints used for ECN 174 marking, which are encoded using two bits of the IP header. The MPLS 175 EXP field is the logical place to encode ECN codepoints, but with 176 only 3 bits (8 codepoints) available, and with the same field being 177 used to convey DSCP information as well, there is a clear incentive 178 to conserve the number of codepoints consumed for ECN purposes. 179 Efficient use of the EXP field has been a focus of prior drafts 180 [Floyd] [Shayman] and we draw on those efforts in this draft as well. 182 We also note that [RFC3168] defines default usage of the ECN field 183 but allows for the possibility that some Diffserv PHBs might include 184 different specifications on how the ECN field is to be used. This 185 draft seeks to preserve that capability. 187 1.2. Intent 189 Our intent is to specify how the MPLS shim header[RFC3032] should 190 denote ECN marking and how MPLS nodes should understand whether the 191 transport for a packet will be ECN capable. We offer this as a 192 building block, from which to build different congestion notification 193 systems. We do not intend to specify how the resulting congestion 194 notification is fed back to an upstream node that can mitigate 195 congestion. For instance, unlike [Shayman], we do not specify edge- 196 to-edge MPLS domain feedback, but we also do not preclude it. 197 Nonetheless, we do specify how the egress node of an MPLS domain 198 should copy congestion notification from the MPLS shim into the 199 encapsulated IP header if the ECN is to be carried onward towards the 200 IP receiver. But we do NOT mandate that MPLS congestion notification 201 must be copied into the IP header for onward transmission. This 202 draft aims to be generic for any use of congestion notification in 203 MPLS. Support of [RFC3168] is our primary motivation; some 204 additional potential applications to illustrate the flexibility of 205 our approach are described in Section 8. In particular, we aim to 206 support possible future schemes that may use more than one level of 207 congestion marking. 209 1.3. Terminology 211 This document draws freely on the terminology of ECN [RFC3168] and 212 MPLS [RFC3031]. For ease of reference, we have included some 213 definitions here, but refer the reader to the references above for 214 complete specifications of the relevant technologies: 216 o CE: Congestion Experienced. One of the states with which a packet 217 may be marked in a network supporting ECN. A packet is marked in 218 this state by an ECN-capable router, to indicate that this router 219 was experiencing congestion at the time the packet arrived. 221 o ECT: ECN-capable Transport. One of the ECN states which a packet 222 may be in when it is sent by an end system. An end system marks a 223 packet with an ECT codepoint to indicate that the end-points of 224 the transport protocol are ECN-capable. A router may not mark a 225 packet as CE unless the packet was marked ECT when it arrived. 227 o Not-ECT: Not ECN capable transport. An end system marks a packet 228 with this codepoint to indicate that the end-points of the 229 transport protocol are not ECN-capable. A congested router cannot 230 mark such packets as CE, and thus can only drop them to indicate 231 congestion. 233 o EXP field. A 3 bit field in the MPLS label header [RFC3032] which 234 may be used to convey Diffserv information (and is also used in 235 this draft to carry ECN information). 237 o PHP. Penultimate Hop Popping. An MPLS operation in which the 238 penultimate Label Switching Router (LSR) on a Label Switched Path 239 (LSP) removes the top label from the packet before forwarding the 240 packet to the final LSR on the LSP. 242 2. Use of MPLS EXP Field for ECN 244 We propose that LSRs configured for explicit congestion notification 245 should use the EXP field in the MPLS shim header. However, [RFC3270] 246 already defines use of codepoints in the EXP field for differentiated 247 services. Although it does not preclude other compatible uses of the 248 EXP field, this clearly seems to limit the space available for ECN, 249 given the field is only 3 bits (8 codepoints). 251 [RFC3270] defines two possible approaches for requesting 252 differentiated service treatment from an LSR. 254 o In the E-LSP approach, different codepoints of the EXP field in 255 the MPLS shim header are used to indicate the packet's per hop 256 behavior (PHB). 258 o In the L-LSP approach, an MPLS label is assigned for each PHB 259 scheduling class (PSC, as defined in [RFC3260], so that an LSR 260 determines both its forwarding and its scheduling behavior from 261 the label. 263 If an MPLS domain uses the L-LSP approach, there is likely to be 264 space in the EXP field for ECN codepoint(s). Where the E-LSP 265 approach is used, then codepoint space in the EXP field is likely to 266 be scarce. This draft focuses on interworking ECN marking with the 267 E-LSP approach as it is the tougher problem. Consequently the same 268 approach can also be applied with L-LSPs. 270 We recommend that explicit congestion notification in MPLS should use 271 codepoints instead of bits in the EXP field. Since not every PHB 272 will necessarily require an associated ECN codepoint it would be 273 wasteful to assign a dedicated bit for ECN. (There may also be cases 274 where a given PHB might need more than one ECN-like codepoint; see 275 Section 8.4 for an example.) 277 For each PHB that uses ECN marking, we assume one EXP codepoint will 278 be defined meaning not congestion marked (Not-CM), and at least one 279 other codepoint will be defined meaning congestion marked (CM). 280 Therefore, each PHB that uses ECN marking will consume at least two 281 EXP codepoints. But PHBs that do not use ECN marking will only 282 consume one. 284 Further, we wish to use minimal space in the MPLS shim header to tell 285 interior LSRs whether each packet will be received by an ECN-capable 286 transport (ECT). Nonetheless, we must ensure that an end-point that 287 would not understand an ECN mark will not receive one, otherwise it 288 will not be able to respond to congestion as it should. In the past, 289 three solutions to this problem have been proposed: 291 o One possible approach is for congested LSRs to mark the ECN field 292 in the underlying IP header at the bottom of the label stack. 293 Although many commercial LSRs routinely access the IP header for 294 other reasons (ECMP), there are numerous drawbacks to attempting 295 to find an IP header beneath an MPLS label stack. Notably, there 296 is the challenge of detecting the absence of an IP header when 297 non-IP packets are carried on an LSP. Therefore we will not 298 consider this approach further. 300 o In the scheme suggested by [Floyd] ECT and CE are overloaded into 301 one bit, so that a 0 means ECT while a 1 might either mean Not-ECT 302 or it might mean CE. A packet that has been marked as having 303 experienced congestion upstream, and then is picked out for 304 marking at a second congested LSR, will be dropped by the second 305 LSR since it cannot determine whether the packet has previously 306 experienced congestion or if ECN is not supported by the 307 transport. 309 While such an approach seemed potentially palatable, we do not 310 recommend it here for the following reasons. In some cases we 311 wish to be able to use ECN marking long before actual congestion 312 (e.g. pre-congestion notification). In these circumstances, 313 marking rates at each LSR might be non-negligible most of the 314 time, so the chances of a previously marked packet encountering an 315 LSR that wants to mark it again will also be non-negligible. In 316 the case where CE and not-ECT are indistinguishable to core 317 routers, such a scenario could lead to unacceptable drop rates. 318 If the typical marking rate at every router or LSR is p, and the 319 typical diameter of the network of LSRs is d, then the probability 320 that a marked packet will be chosen for marking more than once is 321 1-[Pr(never marked) + Pr(marked at exactly one hop)] = 1- [(1-p)^d 322 + dp(1-p)^(d-1)]. For instance, with 6 LSRs in a row, each 323 marking ECN with 1% probability, the chances of a packet that is 324 already marked being chosen for marking a second time is 0.15%. 325 The bit overloading scheme would therefore introduce a drop rate 326 of 0.15% unnecessarily. Given that most modern core networks are 327 sized to introduce near-zero packet drop, it may be unacceptable 328 to drop over one in a thousand packets unnecessarily. 330 o A third possible approach was suggested by [Shayman]. In this 331 scheme, interior LSRs assume that the endpoints are ECN-capable, 332 but this assumption is checked when the final label is popped. If 333 an interior LSR has marked ECN in the EXP field of the shim 334 header, but the IP header says the endpoints are not ECN capable, 335 the edge router (or penultimate router, if using penultimate hop 336 popping) drops the packet. We recommend this scheme, which we 337 call `per-domain ECT checking', and define it more precisely in 338 the following section. Its chief drawback is that it can cause 339 packets to be forwarded after encountering congestion only to be 340 dropped at the egress of the MPLS domain. The rationale for this 341 decision is given in Section 9.1. 343 3. Per-domain ECT checking 345 For the purposes of this discussion, we define the egress nodes of an 346 MPLS domain as the nodes that pop the last MPLS label from the label 347 stack, exposing the IP (or, potentially non-IP) header. Note that 348 such a node may be the ultimate or penultimate hop of an LSP, 349 depending on whether penultimate hop popping (PHP) is employed. 351 In the per-domain ECT checking approach, the egress nodes take 352 responsibility for checking whether the transport is ECN capable. 353 This draft does not specify how these nodes should pass on congestion 354 notification, because different approaches are likely in different 355 scenarios. However, if congestion notification in the MPLS header is 356 copied into the IP header, the procedure MUST conform to the 357 specification given here. 359 If congestion notification is passed to the transport without first 360 passing it onward in the IP header, the approach used must take 361 similar care to check that the transport is ECN capable before 362 passing it ECN markings. Specifically, if the transport for a 363 particular congestion marked MPLS packet is found not to be ECN- 364 capable, the packet MUST be dropped at this egress node. 366 In the per-domain ECT checking approach, only the egress nodes check 367 whether an IP packet is destined for an ECN-capable transport. 368 Therefore, any single LSR within an MPLS domain MUST NOT be 369 configured to enable ECN marking unless all the egress LSRs 370 surrounding it are already configured to handle ECN marking. 372 We call a domain surrounded by ECN-capable egress LSRs an ECN-enabled 373 MPLS domain. This term only implies that all the egress LSRs are 374 ECN-enabled; some interior LSRs may not be ECN-enabled. For 375 instance, it would be possible to use some legacy LSRs incapable of 376 supporting ECN in the interior of an MPLS domain as long as all the 377 egress LSRs were ECN-capable. Note that if PHP is used, the 378 "penultimate hop" routers which perform the pop operation do need to 379 be ECN-enabled, since they are acting in this context as egress LSRs. 381 4. ECN-enabled MPLS domain 383 In the following subsections we describe various operations affecting 384 the ECN marking of a packet that may be performed at MPLS edge and 385 core LSRs. 387 4.1. Pushing (adding) one or more labels to an IP packet 389 On encapsulating an IP packet with an MPLS label stack, the ECN field 390 must be translated from the IP packet into the MPLS EXP field. The 391 Not-CM (not congestion marked) state is set in the MPLS EXP field if 392 the ECN status of the IP packet is "Not ECT" or ECT(1) or ECT(0). 393 The CM state is set if the ECN status of the IP packet is "CE". If 394 more than one label is pushed at one time, the same value should be 395 placed in the EXP value of all label stack entries. 397 4.2. Pushing one or more labels onto an MPLS labelled packet 399 The EXP field is copied directly from the topmost label before the 400 push to the newly added outer label. If more than one label is being 401 pushed, the same EXP value is copied to all label stack entries. 403 4.3. Congestion experienced in an interior MPLS node 405 If the EXP codepoint of the packet maps to a PHB that uses ECN 406 marking and the marking algorithm requires the packet to be marked, 407 the CM state is set (irrespective of whether it is already in the CM 408 state). 410 If the buffer is full, a packet is dropped. 412 4.4. Crossing a Diffserv Domain Boundary 414 If an MPLS-encapsulated packet crosses a Diffserv domain boundary, it 415 may be the case that the two domains use different encodings of the 416 same PHB in the EXP field. In such cases, the EXP field must be 417 rewritten at the domain boundary. If the PHB is one that supports 418 ECN, then the appropriate ECN marking should also be preserved when 419 the EXP field is mapped at the boundary. 421 If an MPLS-encapsulated packet that is in the CM state crosses from a 422 domain that is ECN-enabled (as defined in Section 3) to a domain that 423 is not ECN-enabled, then it is necessary to perform the egress 424 checking procedures at the egress LSR of the ECN-enabled domain. 425 This means that if the encapsulated packet is not ECN capable, the 426 packet MUST be dropped. Note that this implies the egress LSR must 427 be able to look beneath the MPLS header without popping the label 428 stack. 430 The related issue of Diffserv tunnel models is discussed in 431 Section 4.7. 433 4.5. Popping an MPLS label (not the end of the stack) 435 When a packet has more than one MPLS label in the stack and the top 436 label is popped, another MPLS label is exposed. In this case the ECN 437 information should be transferred from the outer EXP field to the 438 inner MPLS label in the following manner. If the inner EXP field is 439 Not-CM, the inner EXP field is set to the same CM or Not-CM state as 440 the outer EXP field. If the inner EXP field is CM, it remains 441 unchanged whatever the outer EXP field. Note that an inner value of 442 CM and an outer value of not-CM should be considered anomalous, and 443 SHOULD be logged in some way by the LSR. 445 4.6. Popping the last MPLS label in the stack 447 When the last MPLS label is popped from the packet, its payload is 448 exposed. If that packet is not IP, and does not have any capability 449 equivalent to ECT, it is assumed Not-ECT and treated as such. That 450 means that if the EXP value of the MPLS header was CM, the packet 451 MUST be dropped. 453 Assuming an IP packet was exposed, we have to examine whether that 454 packet is ECT or not. A Not-ECT packet MUST be dropped if the EXP 455 field is CM. 457 For the remainder of this section, we describe the behavior that is 458 required if the ECN information is to be transferred from the MPLS 459 header into the exposed IP header for onward transmission. As noted 460 in Section 1.2, such behavior is not mandated by this document, but 461 may be selected by an operator. 463 If the inner IP packet is Not-ECT, its ECN field remains unchanged if 464 the EXP field is Not-CM. If the ECN field of the inner packet is set 465 to ECT(0), ECT(1) or CE, the ECN field remains unchanged if the EXP 466 field is set to Not-CM. The ECN field is set to CE if the EXP field 467 is CM. Note that an inner value of CE and an outer value of not-CM 468 should be considered anomalous, and SHOULD be logged in some way by 469 the LSR. 471 4.7. Diffserv Tunneling Models 473 [RFC3270] describes three tunneling models for Diffserv support 474 across MPLS Domains, referred to as the uniform, short pipe, and pipe 475 models. The differences between these models lie in whether the 476 Diffserv treatment that applies to a packet while it travels along a 477 particular LSP is carried to the last hop of the LSP and beyond the 478 last hop. Depending on which mode is preferred by an operator, the 479 EXP value or DSCP value of an exposed header following a label pop 480 may or may not be dependent on the EXP value of the label that is 481 removed by the pop operation. We believe that in the case of ECN 482 marking, the use of these models should only apply to the encoding of 483 the Diffserv PHB in the EXP value, and that the choice of codepoint 484 for ECN should always be made based on the procedures described 485 above, independent of the tunneling model. 487 5. ECN-disabled MPLS domain 489 If ECN is not enabled on all the egress LSRs of a domain, ECN MUST 490 NOT be enabled on any LSRs throughout the domain. If congestion is 491 experienced on any LSR in an ECN-disabled MPLS domain, packets MUST 492 be dropped, NOT marked. The exact algorithm for deciding when to 493 drop packets during congestion (e.g. tail-drop, RED, etc.) is a local 494 matter for the operator of the domain. 496 6. The use of more codepoints with E-LSPs and L-LSPs 498 [RFC3270] gives different options with E-LSPs and L-LSPs and some of 499 those could potentially provide ample EXP codepoints for ECN. 500 However, deploying L-LSPs vs E-LSPs has many implications such as 501 platform support and operational complexity. The above ECN MPLS 502 solution should provide some flexibility. If the operator has 503 deployed one L-LSP per PHB scheduling class, then EXP space will be a 504 non-issue and it could be used to achieve more sophisticated ECN 505 behavior if required. If the operator wants to stick to E-LSPs and 506 uses a handful of EXP codepoints for Diffserv, it may be desirable to 507 operate with a minimum number of extra ECN codepoints, even if this 508 comes with some compromise on ECN optimality. See Section 8 for 509 discussion of some possible deployment scenarios. 511 7. Relationship to tunnel behavior in RFC 3168 513 [RFC3168] defines two modes of encapsulating ECN-marked IP packets 514 inside additional IP headers when tunnels are used. The two modes 515 are the "full functionality" and "limited functionality" modes. In 516 the full functionality mode, the ECT information from the inner 517 header is copied to the outer header at the tunnel ingress, but the 518 CE information is not. In the limited functionality mode, neither 519 ECT nor CE information is copied to the outer header, and thus ECN 520 cannot be applied to the encapsulated packet. 522 The behavior that is specified in Section 4 of this document 523 resembles the "full functionality" mode in the sense that it conveys 524 some information from inner to outer header, and in the sense that it 525 enables full ECN support along the MPLS LSP (which is analogous to an 526 IP tunnel in this context). However it differs in one respect, which 527 is that the CE information is conveyed from the inner header to the 528 outer header. Our original reason for this different design choice 529 was to give interior routers and LSRs more information about upstream 530 marking in multi-bottleneck cases. For instance, the flow pre- 531 emption marking mechanism proposed for PCN works by only considering 532 packets for marking that have not already been marked upstream. 533 Unless existing pre-emption marking is copied from the inner to the 534 outer header at tunnel ingress, the mechanism doesn't pre-empt enough 535 traffic in cases where anomalous events hit multiple domains at once. 536 [RFC3168] does not give any reasons against conveying CE information 537 from the inner header to the outer in the "full functionality" mode. 538 Furthermore, [RFC4301] specifies that the ECN marking should be 539 copied from inner header to outer header in IPSEC tunnels, consistent 540 with the approach defined here. [Briscoe] discusses this issue in 541 more detail. In summary, the approach described in Section 4 appears 542 to be both a sound technical choice and consistent with the current 543 state of thinking in the IETF. 545 8. Example Uses 547 8.1. RFC3168-style ECN 549 [RFC3168] proposes the use of ECN in TCP and introduces the use of 550 ECN-Echo and CWR flags in the TCP header for initialization. The TCP 551 sender responds accordingly (such as not increasing the congestion 552 window) when it receives an ECN-Echo (ECE) ACK packet (that is, an 553 ACK packet with ECN-Echo flag set in the TCP header), then the sender 554 knows that congestion was encountered in the network on the path from 555 the sender to the receiver. 557 It would be possible to enable ECN in an MPLS domain for Diffserv 558 PHBs like AF and best efforts that are expected to be used by TCP and 559 similar transports (e.g. DCCP [RFC4340]). Then end-to-end 560 congestion control in transports capable of understanding ECN would 561 be able to respond to approaching congestion on LSRs without having 562 to rely on packet discard to signal congestion. 564 8.2. ECN Co-existence with Diffserv E-LSPs 566 Many operators today have deployed Diffserv using the E-LSP approach 567 of [RFC3270]. In many cases the number of PHBs used is less than 8, 568 and hence there remain available codepoints in the EXP space. If an 569 operator wished to support ECN for single PHB, this can be 570 accomplished by simply allocated a second codepoint to the PHB for 571 the "CM" state of that PHB and retaining the old codepoint for the 572 "not-CM" state. An operator with only four deployed PHBs could of 573 course enable ECN marking on all those PHBs. It is easy to imagine 574 cases where some PHBs might benefit more from ECN than others - for 575 example, an operator might use ECN on a premium data service but not 576 on a PHB used for best effort internet traffic. 578 As an illustrative example of how the EXP field might be used in this 579 case, consider the example of an operator who is using the aggregated 580 service classes proposed in [I-D.ietf-tsvwg-diffserv-class-aggr]. He 581 may choose to support ECN only for the Assured Elastic Treatment 582 Aggregate, using the EXP codepoint 010 for the not-CM state and 011 583 for the CM state. All other codepoints could be the same as in 584 [I-D.ietf-tsvwg-diffserv-class-aggr]. Of course any other 585 combination of EXP values can be used according to the specific set 586 of PHBs and marking conventions used within that operator's network. 588 8.3. Congestion-feedback-based Traffic Engineering 590 Shayman's traffic engineering [Shayman] proposed the use of ECN by an 591 egress LSR feeding back congestion to an ingress LSR to mitigate 592 congestion by employing dynamic traffic engineering techniques such 593 as shifting flows to an alternate path. It proposed a new RSVP 594 TUNNEL CONGESTION message which was sent to the ingress LSR and 595 ignored by transit LSRs. 597 8.4. PCN flow admission control and flow pre-emption 599 [I-D.briscoe-tsvwg-cl-architecture] proposes using pre-congestion 600 notification (PCN) on routers within an edge-to-edge Diffserv region 601 to control admission of new flows to the region and, if necessary, to 602 pre-empt existing flows in response to disasters and other anomalous 603 routing events. In this approach, the current level of PCN marking 604 is picked up by the signalling used to initiate each flow in order to 605 inform the admission control decision for the whole region at once. 606 As an example, a minor extension to RSVP signalling has been proposed 607 [I-D.lefaucheur-rsvp-ecn] to carry this message, but a similar 608 approach has also been proposed that uses NSIS signalling 609 [I-D.ietf-nsis-rmd]. 611 If it is possible for LSRs to signify congestion in MPLS, PCN marking 612 could be used for admission control and flow pre-emption across a 613 Diffserv region, irrespective of whether it contained pure IP 614 routers, MPLS LSRs, or both. Indeed, the solution could be somewhat 615 more efficient to implement if aggregates could identify themselves 616 by their MPLS label. Appendix A describes the mechanisms by which 617 the necessary markings for PCN could be carried in the MPLS header. 619 As an illustrative example of how the EXP field might be used in this 620 case, consider the example of an operator who is using the aggregated 621 service classes proposed in [I-D.ietf-tsvwg-diffserv-class-aggr]. He 622 may choose to support PCN only for the Real Time Treatment Aggregate, 623 using the EXP codepoint 100 for the not-marked (NM) state, 101 for 624 the Admission Marked (AM) state, and 111 for the Pre-emption Marked 625 (PM) state. All other codepoints could be the same as in 626 [I-D.ietf-tsvwg-diffserv-class-aggr]. Of course any other 627 combination of EXP values can be used according to the specific set 628 of PHBs and marking conventions used within that operator's network. 630 It might also be possible to deploy a similar solution using PCN 631 marking over MPLS for just admission control alone, or just flow pre- 632 emption alone, particularly if codepoint space was at a premium in 633 the MPLS EXP field. However, the feasibility of deploying one 634 without the other would require further study. We also note that an 635 approach to deploying PCN using only a single marking codepoint to 636 support both pre-emption and admission control has been 637 proposed[I-D.charny-pcn-single-marking]. 639 9. Deployment Considerations 641 9.1. Marking non-ECN Capable Packets 643 What are the consequences of marking a packet that is not ECN- 644 capable? Even if it will be dropped before leaving the domain, 645 doesn't this consume resources unnecessarily? 647 The problem only arises if there is congestion downstream of an 648 earlier congested queue in the same MPLS domain. Downstream 649 congested LSRs might forward packets already marked, even though they 650 will be dropped later when the inner IP header is found to be Not-ECT 651 on decapsulation. Such packets might cause the downstream LSRs to 652 mark (or drop) other packets that they would otherwise not have had 653 to. 655 We expect congestion will typically be rare in MPLS networks, but it 656 might not be. The extra unnecessary load at downstream LSRs will not 657 be more than the fraction of marked packets from upstream LSRs, even 658 in the worst case where no transports are ECN capable. Therefore the 659 amount of unnecessary marking (or drop) on an LSR will not be more 660 than the product of its local marking rate and the marking rate due 661 to upstream LSRs within the same domain - typically the product of 662 two small (often zero) probabilities. 664 This is why we decided to use the per-domain ECT checking approach - 665 because the most likely effect would be a very slightly increased 666 marking rate, which would result in very slightly higher drop only 667 for non-ECN-capable transports. We chose not to use the [Floyd] 668 alternative which introduced a low but persistent level of 669 unnecessary packet drop for all time, even for ECN-capable 670 transports. Although that scheme did not carry traffic to the edge 671 of the MPLS domain only to be dropped on decapsulation, we felt our 672 minor inefficiency was a small price to pay. And it would get 673 smaller still if ECN deployment widened. 675 A partial solution would be to preferentially drop packets arriving 676 at a congested router that were already marked. There is no solution 677 to the problem of marking a packet when congestion is caused by 678 another packet that should have been dropped. However, the chance of 679 such an occurrence is very low and the consequences are not 680 significant. It merely causes an application to very occasionally 681 slow down its rate when it did not have to. 683 9.2. Non-ECN capable routers in an MPLS Domain 685 What if an MPLS domain wants to use ECN, but not all legacy routers 686 are able to support it? 688 If the legacy router(s) are used in the interior, this is not a 689 problem. They will simply have to drop the packets if they are 690 congested, rather than mark them, which is the standard behavior for 691 IP routers that are not ECN-enabled. 693 If the legacy router were used as an egress router, it would not be 694 able to check the ECN capability of the transport correctly. An 695 operator in this position would not be able to use this solution and 696 therefore MUST NOT enable ECN unless all egress routers are ECN- 697 capable. 699 10. IANA Considerations 701 This document makes no request of IANA. 703 Note to RFC Editor: this section may be removed on publication as an 704 RFC. 706 11. Security Considerations 708 We believe no new vulnerabilities are introduced by this draft. 710 We have considered whether malicious sources might be able to exploit 711 the fact that interior LSRs will mark packets that are Not-ECT, 712 relying on their egress LSR to drop them. Although this might allow 713 sources to engineer a situation where more traffic is carried across 714 an MPLS domain than should be, we figured that even if we hadn't 715 introduced this feature, these sources would have been able to 716 prevent these LSRs dropping this traffic anyway, simply by setting 717 ECT in the first place. 719 An ECN sender can use the ECN nonce [RFC3540] to detect a misbehaving 720 receiver. The ECN nonce works correctly across an MPLS domain 721 without requiring any specific support from the proposal in this 722 draft. The nonce does not need to be present in the MPLS shim 723 header. As long as the nonce is present in the IP header when the 724 ECN information is copied from the last MPLS shim header, it will be 725 overwritten if congestion has been experienced by an LSR. This is 726 all that is necessary for the sender to detect a misbehaving 727 receiver. 729 12. Acknowledgments 731 Thanks to K.K. Ramakrishnan and Sally Floyd for getting us thinking 732 about this in the first place and for providing advice on tunneling 733 of ECN packets, and to Sally Floyd, Joe Babiarz, Ben Niven-Jenkins, 734 Phil Eardley, Ruediger Geib, and Magnus Westerlund for their comments 735 on the draft. 737 Appendix A. Extension to Pre-Congestion Notification 739 This appendix describes how the mechanisms decribed in the body of 740 the document can be extended to support PCN 741 [I-D.briscoe-tsvwg-cl-architecture]. Our intent here is to show that 742 the mechanisms are readily extended to more complex scenarios than 743 ECN, particulary in the case where more codepoints are needed, but 744 this appendix may be safely ignored if one is interested only in 745 supporting ECN. Note that the PCN standards are still very much 746 under development at the time of writing, hence the precise details 747 contained in this appendix may be subject to change, and we stress 748 that this appendix is for illustrative purposes only. 750 The relevant aspects of PCN for the purposes of this discussion are: 752 o PCN uses 3 states rather than 2 for ECN - these are referred to as 753 admission marked (AM), pre-emption marked (PM) and not marked (NM) 754 states. (See Section 8.4 for further discussion of PCN and the 755 possibility of using fewer codepoints.) 757 o A packet can go from NM to AM, from NM to PM, or from AM to PM, 758 but no other transition is possible. 760 o The determination of whether a packet is subject to PCN is based 761 on the PHB of the packet. 763 Thus, to support PCN fully in an MPLS domain for a particular PHB, a 764 total of 3 codepoints need to be allocated for that PHB. These 3 765 codepoints represent the admission marked (AM), pre-emption marked 766 (PM) and not marked (NM) states. The procedures described in 767 Section 4 above need to be slightly modified to support this 768 scenario. The following procedures are invoked when the topmost DSCP 769 or EXP value indicates a PHB that supports PCN. 771 Appendix A.1. Label Push onto IP packet 773 If the IP packet header indicates AM, set the EXP value of all 774 entries in the label stack to AM. If the IP packet header indicates 775 PM, set the EXP value of all entries in the label stack to PM. For 776 any other marking of the IP header, set the EXP value of all entries 777 in the label stack to NM. 779 Appendix A.2. Pushing Additional MPLS Labels 781 The procedures of Section 4.2 apply. 783 Appendix A.3. Admission Control or Pre-emption Marking inside MPLS 784 domain 786 The EXP value can be set to AM or PM according to the same procedures 787 as described in [I-D.briscoe-tsvwg-cl-phb]. For the purposes of this 788 document, it does not matter exactly what algorithms are used to 789 decide when to set AM or PM; all that matters is that if a router 790 would have marked AM (or PM) in the IP header, it should set the EXP 791 value in the MPLS header to the AM (or PM) codepoint. 793 Appendix A.4. Popping an MPLS Label (not end of stack) 795 When popping an MPLS Label exposes another MPLS label, the AM or PM 796 marking should be transferred to the exposed EXP field in the 797 following manner: 799 o If the inner EXP value is NM, then it should be set to the same 800 marking state as the EXP value of the popped label stack entry. 802 o If the inner EXP value is AM, it should be unchanged if the popped 803 EXP value was AM, and it should be set to PM if the popped EXP 804 value was PM. If the popped EXP value was NM, this should be 805 logged in some way and the inner EXP value should be unchanged. 807 o If the inner EXP value is PM, it should be unchanged whatever the 808 popped EXP value was, but any EXP value other than PM should be 809 logged. 811 Appendix A.5. Popping the last MPLS Label to expose IP header 813 When popping the last MPLS Label exposes the IP header, there are two 814 cases to consider: 816 o the popping LSR is NOT the egress router of the PCN region, in 817 which case AM or PM marking should be transferred to the exposed 818 IP header field; or 820 o the popping LSR IS the egress router of the PCN region. 822 In the latter case, the behavior of the egress LSR is defined in 823 [I-D.briscoe-tsvwg-cl-architecture] and is beyond the scope of this 824 document. In the former case, the marking should be transferred from 825 the popped MPLS header to the exposed IP header as follows: 827 o If the inner IP header value is neither AM nor PM, and the EXP 828 value was NM, then the IP header should be unchanged. For any 829 other EXP value, the IP header should be set to the same marking 830 state as the EXP value of the popped label stack entry. 832 o If the inner IP header value is AM, it should be unchanged if the 833 popped EXP value was AM, and it should be set to PM if the popped 834 EXP value was PM. If the popped EXP value was NM, this should be 835 logged in some way and the inner IP header value should be 836 unchanged. 838 o If the IP header value is PM, it should be unchanged whatever the 839 popped EXP value was, but any EXP value other than PM should be 840 logged. 842 13. References 844 13.1. Normative References 846 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 847 Requirement Levels", BCP 14, RFC 2119, March 1997. 849 [RFC3031] Rosen, E., Viswanathan, A., and R. Callon, "Multiprotocol 850 Label Switching Architecture", RFC 3031, January 2001. 852 [RFC3032] Rosen, E., Tappan, D., Fedorkow, G., Rekhter, Y., 853 Farinacci, D., Li, T., and A. Conta, "MPLS Label Stack 854 Encoding", RFC 3032, January 2001. 856 [RFC3168] Ramakrishnan, K., Floyd, S., and D. Black, "The Addition 857 of Explicit Congestion Notification (ECN) to IP", 858 RFC 3168, September 2001. 860 [RFC3270] Le Faucheur, F., Wu, L., Davie, B., Davari, S., Vaananen, 861 P., Krishnan, R., Cheval, P., and J. Heinanen, "Multi- 862 Protocol Label Switching (MPLS) Support of Differentiated 863 Services", RFC 3270, May 2002. 865 [RFC4301] Kent, S. and K. Seo, "Security Architecture for the 866 Internet Protocol", RFC 4301, December 2005. 868 13.2. Informative References 870 [Briscoe] "Layered Encapsulation of Congestion Notification", 871 June 2007. 873 Work in progress. 875 [Floyd] "A Proposal to Incorporate ECN in MPLS", 1999. 877 Work in progress. http://www.icir.org/floyd/papers/ 878 draft-ietf-mpls-ecn-00.txt 880 [I-D.briscoe-tsvwg-cl-architecture] 881 Briscoe, B., "An edge-to-edge Deployment Model for Pre- 882 Congestion Notification: Admission Control over a 883 DiffServ Region", draft-briscoe-tsvwg-cl-architecture-04 884 (work in progress), October 2006. 886 [I-D.briscoe-tsvwg-cl-phb] 887 Briscoe, B., "Pre-Congestion Notification marking", 888 draft-briscoe-tsvwg-cl-phb-03 (work in progress), 889 October 2006. 891 [I-D.charny-pcn-single-marking] 892 Charny, A., "Pre-Congestion Notification Using Single 893 Marking for Admission and Pre-emption", 894 draft-charny-pcn-single-marking-01 (work in progress), 895 March 2007. 897 [I-D.ietf-nsis-rmd] 898 Bader, A., "RMD-QOSM - The Resource Management in Diffserv 899 QOS Model", draft-ietf-nsis-rmd-09 (work in progress), 900 March 2007. 902 [I-D.ietf-tsvwg-diffserv-class-aggr] 903 Chan, K., "Aggregation of DiffServ Service Classes", 904 draft-ietf-tsvwg-diffserv-class-aggr-02 (work in 905 progress), March 2007. 907 [I-D.lefaucheur-rsvp-ecn] 908 Faucheur, F., "RSVP Extensions for Admission Control over 909 Diffserv using Pre-congestion Notification (PCN)", 910 draft-lefaucheur-rsvp-ecn-01 (work in progress), 911 June 2006. 913 [RFC3260] Grossman, D., "New Terminology and Clarifications for 914 Diffserv", RFC 3260, April 2002. 916 [RFC3540] Spring, N., Wetherall, D., and D. Ely, "Robust Explicit 917 Congestion Notification (ECN) Signaling with Nonces", 918 RFC 3540, June 2003. 920 [RFC4340] Kohler, E., Handley, M., and S. Floyd, "Datagram 921 Congestion Control Protocol (DCCP)", RFC 4340, March 2006. 923 [Shayman] "Using ECN to Signal Congestion Within an MPLS Domain", 924 2000. 926 Work in progress. http://www.ee.umd.edu/~shayman/papers.d/ 927 draft-shayman-mpls-ecn-00.txt 929 Authors' Addresses 931 Bruce Davie 932 Cisco Systems, Inc. 933 1414 Mass. Ave. 934 Boxborough, MA 01719 935 USA 937 Email: bsd@cisco.com 938 Bob Briscoe 939 BT Research 940 B54/77, Sirius House 941 Adastral Park 942 Martlesham Heath 943 Ipswich 944 Suffolk IP5 3RE 945 United Kingdom 947 Email: bob.briscoe@bt.com 949 June Tay 950 BT Research 951 B54/77, Sirius House 952 Adastral Park 953 Martlesham Heath 954 Ipswich 955 Suffolk IP5 3RE 956 United Kingdom 958 Email: june.tay@bt.com 960 Full Copyright Statement 962 Copyright (C) The IETF Trust (2007). 964 This document is subject to the rights, licenses and restrictions 965 contained in BCP 78, and except as set forth therein, the authors 966 retain all their rights. 968 This document and the information contained herein are provided on an 969 "AS IS" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS 970 OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY, THE IETF TRUST AND 971 THE INTERNET ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS 972 OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF 973 THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED 974 WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. 976 Intellectual Property 978 The IETF takes no position regarding the validity or scope of any 979 Intellectual Property Rights or other rights that might be claimed to 980 pertain to the implementation or use of the technology described in 981 this document or the extent to which any license under such rights 982 might or might not be available; nor does it represent that it has 983 made any independent effort to identify any such rights. Information 984 on the procedures with respect to rights in RFC documents can be 985 found in BCP 78 and BCP 79. 987 Copies of IPR disclosures made to the IETF Secretariat and any 988 assurances of licenses to be made available, or the result of an 989 attempt made to obtain a general license or permission for the use of 990 such proprietary rights by implementers or users of this 991 specification can be obtained from the IETF on-line IPR repository at 992 http://www.ietf.org/ipr. 994 The IETF invites any interested party to bring to its attention any 995 copyrights, patents or patent applications, or other proprietary 996 rights that may cover technology that may be required to implement 997 this standard. Please address the information to the IETF at 998 ietf-ipr@ietf.org. 1000 Acknowledgment 1002 Funding for the RFC Editor function is provided by the IETF 1003 Administrative Support Activity (IASA).