idnits 2.17.1 draft-ietf-tsvwg-ecn-encap-guidelines-01.txt: Checking boilerplate required by RFC 5378 and the IETF Trust (see https://trustee.ietf.org/license-info): ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/1id-guidelines.txt: ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/checklist : ---------------------------------------------------------------------------- -- The draft header indicates that this document updates RFC3819, but the abstract doesn't seem to mention this, which it should. Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year (Using the creation date from RFC3819, updated by this document, for RFC5378 checks: 1999-10-14) -- 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 (November 16, 2014) is 3449 days in the past. Is this intentional? Checking references for intended status: Best Current Practice ---------------------------------------------------------------------------- (See RFCs 3967 and 4897 for information about using normative references to lower-maturity documents in RFCs) == Missing Reference: 'GF' is mentioned on line 1225, but not defined == Outdated reference: A later version (-07) exists of draft-ietf-trill-rfc7180bis-00 -- Obsolete informational reference (is this intentional?): RFC 1323 (Obsoleted by RFC 7323) Summary: 0 errors (**), 0 flaws (~~), 3 warnings (==), 4 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 Transport Area Working Group B. Briscoe 3 Internet-Draft BT 4 Updates: 3819 (if approved) J. Kaippallimalil 5 Intended status: Best Current Practice Huawei 6 Expires: May 20, 2015 P. Thaler 7 Broadcom Corporation 8 November 16, 2014 10 Guidelines for Adding Congestion Notification to Protocols that 11 Encapsulate IP 12 draft-ietf-tsvwg-ecn-encap-guidelines-01 14 Abstract 16 The purpose of this document is to guide the design of congestion 17 notification in any lower layer or tunnelling protocol that 18 encapsulates IP. The aim is for explicit congestion signals to 19 propagate consistently from lower layer protocols into IP. Then the 20 IP internetwork layer can act as a portability layer to carry 21 congestion notification from non-IP-aware congested nodes up to the 22 transport layer (L4). Following these guidelines should assure 23 interworking between new lower layer congestion notification 24 mechanisms, whether specified by the IETF or other standards bodies. 26 Status of This Memo 28 This Internet-Draft is submitted in full conformance with the 29 provisions of BCP 78 and BCP 79. 31 Internet-Drafts are working documents of the Internet Engineering 32 Task Force (IETF). Note that other groups may also distribute 33 working documents as Internet-Drafts. The list of current Internet- 34 Drafts is at http://datatracker.ietf.org/drafts/current/. 36 Internet-Drafts are draft documents valid for a maximum of six months 37 and may be updated, replaced, or obsoleted by other documents at any 38 time. It is inappropriate to use Internet-Drafts as reference 39 material or to cite them other than as "work in progress." 41 This Internet-Draft will expire on May 20, 2015. 43 Copyright Notice 45 Copyright (c) 2014 IETF Trust and the persons identified as the 46 document authors. All rights reserved. 48 This document is subject to BCP 78 and the IETF Trust's Legal 49 Provisions Relating to IETF Documents 50 (http://trustee.ietf.org/license-info) in effect on the date of 51 publication of this document. Please review these documents 52 carefully, as they describe your rights and restrictions with respect 53 to this document. Code Components extracted from this document must 54 include Simplified BSD License text as described in Section 4.e of 55 the Trust Legal Provisions and are provided without warranty as 56 described in the Simplified BSD License. 58 Table of Contents 60 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 3 61 1.1. Scope . . . . . . . . . . . . . . . . . . . . . . . . . . 5 62 2. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . 6 63 3. Modes of Operation . . . . . . . . . . . . . . . . . . . . . 7 64 3.1. Feed-Forward-and-Up Mode . . . . . . . . . . . . . . . . 8 65 3.2. Feed-Up-and-Forward Mode . . . . . . . . . . . . . . . . 9 66 3.3. Feed-Backward Mode . . . . . . . . . . . . . . . . . . . 10 67 3.4. Null Mode . . . . . . . . . . . . . . . . . . . . . . . . 12 68 4. Feed-Forward-and-Up Mode: Guidelines for Adding Congestion 69 Notification . . . . . . . . . . . . . . . . . . . . . . . . 12 70 4.1. IP-in-IP Tunnels with Tightly Coupled Shim Headers . . . 13 71 4.2. Wire Protocol Design: Indication of ECN Support . . . . . 13 72 4.3. Encapsulation Guidelines . . . . . . . . . . . . . . . . 15 73 4.4. Decapsulation Guidelines . . . . . . . . . . . . . . . . 17 74 4.5. Sequences of Similar Tunnels or Subnets . . . . . . . . . 18 75 4.6. Reframing and Congestion Markings . . . . . . . . . . . . 19 76 5. Feed-Up-and-Forward Mode: Guidelines for Adding Congestion 77 Notification . . . . . . . . . . . . . . . . . . . . . . . . 19 78 6. Feed-Backward Mode: Guidelines for Adding Congestion 79 Notification . . . . . . . . . . . . . . . . . . . . . . . . 21 80 7. IANA Considerations (to be removed by RFC Editor) . . . . . . 22 81 8. Security Considerations . . . . . . . . . . . . . . . . . . . 22 82 9. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . 22 83 10. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 23 84 11. Comments Solicited . . . . . . . . . . . . . . . . . . . . . 23 85 12. References . . . . . . . . . . . . . . . . . . . . . . . . . 23 86 12.1. Normative References . . . . . . . . . . . . . . . . . . 23 87 12.2. Informative References . . . . . . . . . . . . . . . . . 24 88 Appendix A. Outstanding Document Issues . . . . . . . . . . . . 27 89 Appendix B. Changes in This Version (to be removed by RFC 90 Editor) . . . . . . . . . . . . . . . . . . . . . . 27 92 1. Introduction 94 The benefits of Explicit Congestion Notification (ECN) described 95 below can only be fully realised if support for ECN is added to the 96 relevant subnetwork technology, as well as to IP. When a lower layer 97 buffer drops a packet obviously it does not just drop at that layer; 98 the packet disappears from all layers. In contrast, when a lower 99 layer marks a packet with ECN, the marking needs to be explicitly 100 propagated up the layers. The same is true if a buffer marks the 101 outer header of a packet that encapsulates inner tunnelled headers. 102 Forwarding ECN is not as straightforward as other headers because it 103 has to be assumed ECN may be only partially deployed. If an egress 104 at any layer is not ECN-aware, or if the ultimate receiver or sender 105 is not ECN-aware, congestion needs to be indicated by dropping a 106 packet, not marking it. 108 The purpose of this document is to guide the addition of congestion 109 notification to any subnet technology or tunnelling protocol, so that 110 lower layer equipment can signal congestion explicitly and it will 111 propagate consistently into encapsulated (higher layer) headers, 112 otherwise the signals will not reach their ultimate destination. 114 ECN is defined in the IP header (v4 & v6) [RFC3168] to allow a 115 resource to notify the onset of queue build-up without having to drop 116 packets, by explicitly marking a proportion of packets with the 117 congestion experienced (CE) codepoint. 119 Given a suitable marking scheme, ECN removes nearly all congestion 120 loss and it cuts delays for two main reasons: 122 o It avoids the delay when recovering from congestion losses, which 123 particularly benefits small flows or real-time flows, making their 124 delivery time predictably short [RFC2884]; 126 o As ECN is used more widely by end-systems, it will gradually 127 remove the need to configure a degree of delay into buffers before 128 they start to notify congestion (the cause of bufferbloat). This 129 is because drop involves a trade-off between sending a timely 130 signal and trying to avoid impairment, whereas ECN is solely a 131 signal not an impairment, so there is no harm triggering it 132 earlier. 134 Some lower layer technologies (e.g. MPLS, Ethernet) are used to form 135 subnetworks with IP-aware nodes only at the edges. These networks 136 are often sized so that it is rare for interior queues to overflow. 137 However, this has often be more due to the inability of the original 138 TCP protocol to saturate the links. For many years, fixes such as 139 window scaling [RFC1323] proved hard to deploy. But now that modern 140 operating systems are finally capable of saturating interior links, 141 even the buffers of well-provisioned interior switches will need to 142 signal episodes of queuing. 144 Propagation of ECN is defined for MPLS [RFC5129], and is being 145 defined for TRILL [I-D.ietf-trill-rfc7180bis], but it remains to be 146 defined for a number of other subnetwork technologies. 148 Similarly, ECN propagation is yet to be defined for many tunnelling 149 protocols. [RFC6040] defines how ECN should be propagated for IP-in- 150 IP [RFC2003] and IPsec [RFC4301] tunnels. However, as Section 9.3 of 151 RFC3168 pointed out, ECN support will need to be defined for other 152 tunnelling protocols, e.g. L2TP [RFC2661], GRE [RFC1701], [RFC2784], 153 PPTP [RFC2637] and GTP [GTPv1], [GTPv1-U], [GTPv2-C]. 155 Incremental deployment is the most tricky aspect when adding support 156 for ECN. The original ECN protocol in IP [RFC3168] was carefully 157 designed so that a congested buffer would not mark a packet (rather 158 than drop it) unless both source and destination hosts were ECN- 159 capable. Otherwise its congestion markings would never be detected 160 and congestion would just deteriorate further. However, to support 161 congestion marking below the IP layer, it is not sufficient to only 162 check that the two end-points support ECN; correct operation also 163 depends on the decapsulator at each subnet egress faithfully 164 propagating congestion notifications to the higher layer. Otherwise, 165 a legacy decapsulator might silently fail to propagate any ECN 166 signals from the outer to the forwarded header. Then the lost 167 signals would never be detected and again congestion would 168 deteriorate further. The guidelines given later require protocol 169 designers to carefully consider incremental deployment, and suggest 170 various safe approaches for different circumstances. 172 Of course, the IETF does not have standards authority over every link 173 layer protocol. So this document gives guidelines for designing 174 propagation of congestion notification across the interface between 175 IP and protocols that may encapsulate IP (i.e. that can be layered 176 beneath IP). Each lower layer technology will exhibit different 177 issues and compromises, so the IETF or the relevant standards body 178 must be free to define the specifics of each lower layer congestion 179 notification scheme. Nonetheless, if the guidelines are followed, 180 congestion notification should interwork between different 181 technologies, using IP in its role as a 'portability layer'. 183 Therefore, the capitalised term 'SHOULD' or 'SHOULD NOT' are often 184 used in preference to 'MUST' or 'MUST NOT', because it is difficult 185 to know the compromises that will be necessary in each protocol 186 design. If a particular protocol design chooses to contradict a 187 'SHOULD (NOT)' given in the advice below, it MUST include a sound 188 justification. 190 It has not been possible to give common guidelines for all lower 191 layer technologies, because they do not all fit a common pattern. 192 Instead they have been divided into a few distinct modes of 193 operation: feed-forward-and-upward; feed-upward-and-forward; feed- 194 backward; and null mode. These modes are described in Section 3, 195 then in the following sections separate guidelines are given for each 196 mode. 198 This document updates the advice to subnetwork designers about ECN in 199 Section 13 of [RFC3819]. 201 1.1. Scope 203 This document only concerns wire protocol processing of explicit 204 notification of congestion and makes no changes or recommendations 205 concerning algorithms for congestion marking or for congestion 206 response (algorithm issues should be independent of the layer the 207 algorithm operates in). 209 The question of congestion notification signals with different 210 semantics to those of ECN in IP is touched on in a couple of specific 211 cases (e.g. QCN [IEEE802.1Qau]) and with schemes with multiple 212 severity levels such as PCN [RFC6660]). However, no attempt is made 213 to give guidelines about schemes with different semantics that are 214 yet to be invented. 216 The semantics of congestion signals can be relative to the traffic 217 class. Therefore correct propagation of congestion signals could 218 depend on correct propagation of any traffic class field between the 219 layers. In this document, correct propagation of traffic class 220 information is assumed, while what 'correct' means and how it is 221 achieved is covered elsewhere (e.g. [RFC2983]) and is outside the 222 scope of the present document. 224 Note that these guidelines do not require the subnet wire protocol to 225 be changed to accommodate congestion notification. Another way to 226 add congestion notification without consuming header space in the 227 subnet protocol might be to use a parallel control plane protocol. 229 This document focuses on the congestion notification interface 230 between IP and lower layer protocols that can encapsulate IP, where 231 the term 'IP' includes v4 or v6, unicast, multicast or anycast. 232 However, it is likely that the guidelines will also be useful when a 233 lower layer protocol or tunnel encapsulates itself (e.g. Ethernet 234 MAC in MAC [IEEE802.1Qah]) or when it encapsulates other protocols. 236 In the feed-backward mode, propagation of congestion signals for 237 multicast and anycast packets is out-of-scope (because it would be so 238 complicated that it is hoped no-one would attempt such an 239 abomination). 241 2. Terminology 243 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 244 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 245 document are to be interpreted as described in RFC 2119 [RFC2119]. 247 Further terminology used within this document: 249 Protocol data unit (PDU): Information that is delivered as a unit 250 among peer entities of a layered network consisting of protocol 251 control information (typically a header) and possibly user data 252 (payload) of that layer. The scope of this document includes 253 layer 2 and layer 3 networks, where the PDU is respectively termed 254 a frame or a packet (or a cell in ATM). PDU is a general term for 255 any of these. This definition also includes a payload with a shim 256 header lying somewhere between layer 2 & 3. 258 Transport: The end-to-end transmission control function, 259 conventionally considered at layer-4 in the OSI reference model. 260 Given the audience for this document will often use the word 261 transport to mean low level bit carriage, whenever the term is 262 used it will be qualified, e.g. 'L4 transport'. 264 Encapsulator: The link or tunnel endpoint function that adds an 265 outer header to a PDU (also termed the 'link ingress', the 'subnet 266 ingress', the 'ingress tunnel endpoint' or just the 'ingress' 267 where the context is clear). 269 Decapsulator: The link or tunnel endpoint function that removes an 270 outer header from a PDU (also termed the 'link egress', the 271 'subnet egress', the 'egress tunnel endpoint' or just the 'egress' 272 where the context is clear). 274 Incoming header: The header of an arriving PDU before encapsulation. 276 Outer header: The header added to encapsulate a PDU. 278 Inner header: The header encapsulated by the outer header. 280 Outgoing header: The header forwarded by the decapsulator. 282 CE: Congestion Experienced [RFC3168] 283 ECT: ECN-Capable Transport [RFC3168] 285 Not-ECT: Not ECN-Capable Transport [RFC3168] 287 ECN-PDU: A PDU that is part of a feedback loop within which all the 288 nodes that need to propagate explicit congestion notifications 289 back to the Load Regulator are ECN-capable. An IP packet with a 290 non-zero ECN field implies that the endpoints are ECN-capable, so 291 this would be an ECN-PDU. However, ECN-PDU is intended to be a 292 general term for a PDU at any layer, not just IP. 294 Not-ECN-PDU: A PDU that is part of a feedback-loop within which some 295 nodes necessary to propagate explicit congestion notifications 296 back to the load regulator are not ECN-capable. 298 Load Regulator: For each flow of PDUs, the transport function that 299 is capable of controlling the data rate. Typically located at the 300 data source, but in-path nodes can regulate load in some 301 congestion control arrangements (e.g. admission control or 302 policing nodes). Note the term "a function capable of controlling 303 the load" deliberately includes a transport that doesn't actually 304 control the load but ideally it ought to (e.g. a sending 305 application without congestion control that uses UDP). 307 Congestion Baseline: The location of the function on the path that 308 initialised the values of all congestion notification fields in a 309 sequence of packets, before any are set to the congestion 310 experienced (CE) codepoint if they experience congestion further 311 downstream. Typically the original data source at layer-4. 313 3. Modes of Operation 315 This section sets down the different modes by which congestion 316 information is passed between the lower layer and the higher one. It 317 acts as a reference framework for the following sections, which give 318 normative guidelines for designers of explicit congestion 319 notification protocols, taking each mode in turn: 321 Feed-Forward-and-Up: Nodes feed forward congestion notification 322 towards the egress within the lower layer then up and along the 323 layers towards the end-to-end destination at the transport layer. 324 The following local optimisation is possible: 326 Feed-Up-and-Forward: A lower layer switch feeds-up congestion 327 notification directly into the ECN field in the higher layer 328 (e.g. IP) header, irrespective of whether the node is at the 329 egress of a subnet. 331 Feed-Backward: Nodes feed back congestion signals towards the 332 ingress of the lower layer and (optionally) attempt to control 333 congestion within their own layer. 335 Null: Nodes cannot experience congestion at the lower layer except 336 at ingress nodes (which are IP-aware or equivalently higher-layer- 337 aware). 339 3.1. Feed-Forward-and-Up Mode 341 Like IP and MPLS, many subnet technologies are based on self- 342 contained protocol data units (PDUs) or frames sent unreliably. They 343 provide no feedback channel at the subnetwork layer, instead relying 344 on higher layers (e.g. TCP) to feed back loss signals. 346 In these cases, ECN may best be supported by standardising explicit 347 notification of congestion into the lower layer protocol that carries 348 the data forwards. It will then also be necessary to define how the 349 egress of the lower layer subnet propagates this explicit signal into 350 the forwarded upper layer (IP) header. It can then continue forwards 351 until it finally reaches the destination transport (at L4). Then 352 typically the destination will feed this congestion notification back 353 to the source transport using an end-to-end protocol (e.g. TCP). 354 This is the arrangement that has already been used to add ECN to IP- 355 in-IP tunnels [RFC6040], IP-in-MPLS and MPLS-in-MPLS [RFC5129]. 357 This mode is illustrated in Figure 1. Along the middle of the 358 figure, layers 2, 3 & 4 of the protocol stack are shown, and one 359 packet is shown along the bottom as it progresses across the network 360 from source to destination, crossing two subnets connected by a 361 router, and crossing two switches on the path across each subnet. 362 Congestion at the output of the first switch (shown as *) leads to a 363 congestion marking in the L2 header (shown as C in the illustration 364 of the packet). The chevrons show the progress of the resulting 365 congestion indication. It is propagated from link to link across the 366 subnet in the L2 header, then when the router removes the marked L2 367 header, it propagates the marking up into the L3 (IP) header. The 368 router forwards the marked L3 header into subnet 2, and when it adds 369 a new L2 header it copies the L3 marking into the L2 header as well, 370 as shown by the 'C's in both layers (assuming the technology of 371 subnet 2 also supports explicit congestion marking). 373 Note that there is no implication that each 'C' marking is encoded 374 the same; a different encoding might be used for the 'C' marking in 375 each protocol. 377 Finally, for completeness, we show the L3 marking arriving at the 378 destination, where the host transport protocol (e.g. TCP) feeds it 379 back to the source in the L4 acknowledgement (the 'C' at L4 in the 380 packet at the top of the diagram). 382 _ _ _ 383 /_______ | | |C| ACK Packet (V) 384 \ |_|_|_| 385 +---+ layer: 2 3 4 header +---+ 386 | <|<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<< Packet V <<<<<<<<<<<<<|<< |L4 387 | | +---+ | ^ | 388 | | . . . . . . Packet U. . | >>|>>> Packet U >>>>>>>>>>>>|>^ |L3 389 | | +---+ +---+ | ^ | +---+ +---+ | | 390 | | | *|>>>>>|>>>|>>>>>|>^ | | | | | | |L2 391 |___|_____|___|_____|___|_____|___|_____|___|_____|___|_____|___| 392 source subnet A router subnet B dest 393 __ _ _ _ __ _ _ _ __ _ _ __ _ _ _ 394 | | | | | | | | |C| | | |C| | | |C|C| Data________\ 395 |__|_|_|_| |__|_|_|_| |__|_|_| |__|_|_|_| Packet (U) / 396 layer: 4 3 2A 4 3 2A 4 3 4 3 2B 397 header 399 Figure 1: Feed-Forward-and-Up Mode 401 Of course, modern networks are rarely as simple as this text-book 402 example, often involving multiple nested layers. For example, a 3GPP 403 mobile network may have two IP-in-IP (GTP) tunnels in series and an 404 MPLS backhaul between the base station and the first router. 405 Nonetheless, the example illustrates the general idea of feeding 406 congestion notification forward then upward whenever a header is 407 removed at the egress of a subnet. 409 Note that the FECN (forward ECN) bit in Frame Relay and the explicit 410 forward congestion indication (EFCI [ITU-T.I.371]) bit in ATM user 411 data cells follow a feed-forward pattern. However, in ATM, this is 412 only as part of a feed-forward-and-backward pattern at the lower 413 layer, not feed-forward-and-up out of the lower layer--the intention 414 was never to interface to IP ECN at the subnet egress. To our 415 knowledge, Frame Relay FECN is solely used to detect where more 416 capacity should be provisioned [Buck00]. 418 3.2. Feed-Up-and-Forward Mode 420 Ethernet is particularly difficult to extend incrementally to support 421 explicit congestion notification. One way to support ECN in such 422 cases has been to use so called 'layer-3 switches'. These are 423 Ethernet switches that bury into the Ethernet payload to find an IP 424 header and manipulate or act on certain IP fields (specifically 425 Diffserv & ECN). For instance, in Data Center TCP [DCTCP], layer-3 426 switches are configured to mark the ECN field of the IP header within 427 the Ethernet payload when their output buffer becomes congested. 428 With respect to switching, a layer-3 switch acts solely on the 429 addresses in the Ethernet header; it doesn't use IP addresses, and it 430 doesn't decrement the TTL field in the IP header. 432 _ _ _ 433 /_______ | | |C| ACK packet (V) 434 \ |_|_|_| 435 +---+ layer: 2 3 4 header +---+ 436 | <|<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<< Packet V <<<<<<<<<<<<<|<< |L4 437 | | +---+ | ^ | 438 | | . . . >>>> Packet U >>>|>>>|>>> Packet U >>>>>>>>>>>>|>^ |L3 439 | | +--^+ +---+ | | +---+ +---+ | | 440 | | | *| | | | | | | | | | |L2 441 |___|_____|___|_____|___|_____|___|_____|___|_____|___|_____|___| 442 source subnet E router subnet F dest 443 __ _ _ _ __ _ _ _ __ _ _ __ _ _ _ 444 | | | | | | | |C| | | | |C| | | |C|C| data________\ 445 |__|_|_|_| |__|_|_|_| |__|_|_| |__|_|_|_| packet (U) / 446 layer: 4 3 2 4 3 2 4 3 4 3 2 447 header 449 Figure 2: Feed-Up-and-Forward Mode 451 By comparing Figure 2 with Figure 1, it can be seen that subnet E 452 (perhaps a subnet of layer-3 Ethernet switches) works in feed-up-and- 453 forward mode by notifying congestion directly into L3 at the point of 454 congestion, even though the congested switch does not otherwise act 455 at L3. In this example, the technology in subnet F (e.g. MPLS) does 456 support ECN natively, so when the router adds the layer-2 header it 457 copies the ECN marking from L3 to L2 as well. 459 3.3. Feed-Backward Mode 461 In some layer 2 technologies, explicit congestion notification has 462 been defined for use internally within the subnet with its own 463 feedback and load regulation, but typically the interface with IP for 464 ECN has not been defined. 466 For instance, for the available bit-rate (ABR) service in ATM, the 467 relative rate mechanism was one of the more popular mechanisms for 468 managing traffic, tending to supersede earlier designs. In this 469 approach ATM switches send special resource management (RM) cells in 470 both the forward and backward directions to control the ingress rate 471 of user data into a virtual circuit. If a switch buffer is 472 approaching congestion or congested it sends an RM cell back towards 473 the ingress with respectively the No Increase (NI) or Congestion 474 Indication (CI) bit set in its message type field [ATM-TM-ABR]. The 475 ingress then holds or decreases its sending bit-rate accordingly. 477 _ _ _ 478 /_______ | | |C| ACK packet (X) 479 \ |_|_|_| 480 +---+ layer: 2 3 4 header +---+ 481 | <|<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<< Packet X <<<<<<<<<<<<<|<< |L4 482 | | +---+ | ^ | 483 | | | *|>>> Packet W >>>>>>>>>>>>|>^ |L3 484 | | +---+ +---+ | | +---+ +---+ | | 485 | | | | | | | <|<<<<<|<<<|<(V)<|<<<| | |L2 486 | | . . | . |Packet U | . . | . | . . | . | . . | .*| . . | |L2 487 |___|_____|___|_____|___|_____|___|_____|___|_____|___|_____|___| 488 source subnet G router subnet H dest 489 __ _ _ _ __ _ _ _ __ _ _ __ _ _ _ later 490 | | | | | | | | | | | | | | | | |C| | data________\ 491 |__|_|_|_| |__|_|_|_| |__|_|_| |__|_|_|_| packet (W) / 492 4 3 2 4 3 2 4 3 4 3 2 493 _ 494 /__ |C| Feedback control 495 \ |_| cell/frame (V) 496 2 497 __ _ _ _ __ _ _ _ __ _ _ __ _ _ _ earlier 498 | | | | | | | | | | | | | | | | | | | data________\ 499 |__|_|_|_| |__|_|_|_| |__|_|_| |__|_|_|_| packet (U) / 500 layer: 4 3 2 4 3 2 4 3 4 3 2 501 header 503 Figure 3: Feed-Backward Mode 505 ATM's feed-backward approach doesn't fit well when layered beneath 506 IP's feed-forward approach--unless the initial data source is the 507 same node as the ATM ingress. Figure 3 shows the feed-backward 508 approach being used in subnet H. If the final switch on the path is 509 congested (*), it doesn't feed-forward any congestion indications on 510 packet (U). Instead it sends a control cell (V) back to the router 511 at the ATM ingress. 513 However, the backward feedback doesn't reach the original data source 514 directly because IP doesn't support backward feedback (and subnet G 515 is independent of subnet H). Instead, the router in the middle 516 throttles down its sending rate but the original data sources don't 517 reduce their rates. The resulting rate mismatch causes the middle 518 router's buffer at layer 3 to back up until it becomes congested, 519 which it signals forwards on later data packets at layer 3 (e.g. 520 packet W). Note that the forward signal from the middle router is 521 not triggered directly by the backward signal. Rather, it is 522 triggered by congestion resulting from the middle router's mismatched 523 rate response to the backward signal. 525 In response to this later forward signalling, end-to-end feedback at 526 layer-4 finally completes the tortuous path of congestion indications 527 back to the origin data source, as before. 529 3.4. Null Mode 531 Often link and physical layer resources are 'non-blocking' by design. 532 In these cases congestion notification may be implemented but it does 533 not need to be deployed at the lower layer; ECN in IP would be 534 sufficient. 536 A degenerate example is a point-to-point Ethernet link. Excess 537 loading of the link merely causes the queue from the higher layer to 538 back up, while the lower layer remains immune to congestion. Even a 539 whole meshed subnetwork can be made immune to interior congestion by 540 limiting ingress capacity and careful sizing of links, particularly 541 if multi-path routing is used to ensure even worst-case patterns of 542 load cannot congest any link. 544 4. Feed-Forward-and-Up Mode: Guidelines for Adding Congestion 545 Notification 547 Feed-forward-and-up is the mode already used for signalling ECN up 548 the layers through MPLS into IP [RFC5129] and through IP-in-IP 549 tunnels [RFC6040]. These RFCs take a consistent approach and the 550 following guidelines are designed to ensure this consistency 551 continues as ECN support is added to other protocols that encapsulate 552 IP. The guidelines are also designed to ensure compliance with the 553 more general best current practice for the design of alternate ECN 554 schemes given in [RFC4774]. 556 The rest of this section is structured as follows: 558 o Section 4.1 addresses the most straightforward cases, where 559 [RFC6040] can be applied directly to add ECN to tunnels that are 560 effectively the same as IP-in-IP tunnels. 562 o The subsequent sections give guidelines for adding ECN to a subnet 563 technology that uses feed-forward-and-up mode like IP, but it is 564 not so similar to IP that [RFC6040] rules can be applied directly. 565 Specifically: 567 * Sections 4.2, 4.3 and 4.4 respectively address how to add ECN 568 support to the wire protocol and to the encapsulators and 569 decapsulators at the ingress and egress of the subnet. 571 * Section 4.5 deals with the special, but common, case of 572 sequences of tunnels or subnets that all use the same 573 technology 575 * Section 4.6 deals with the question of reframing when IP 576 packets do not map 1:1 into lower layer frames. 578 4.1. IP-in-IP Tunnels with Tightly Coupled Shim Headers 580 A common pattern for many tunnelling protocols is to encapsulate an 581 inner IP header with shim header(s) then an outer IP header. In many 582 cases the shim header(s) always have to be tightly coupled to the 583 outer IP header because they are not sufficient as outer headers in 584 their own right. In such cases the shim header(s) and the outer IP 585 header are always added (or removed) in the same operation. 586 Therefore, in all such tightly coupled IP-in-IP tunnelling protocols, 587 the rules in [RFC6040] for propagating the ECN field between the two 588 IP headers SHOULD be applied directly. 590 Examples of tightly coupled IP-in-IP tunnelling protocols where 591 [RFC6040] can be applied directly are: 593 o L2TP [RFC2661] 595 o GRE [RFC1701], [RFC2784] 597 o PPTP [RFC2637] 599 o GTP [GTPv1], [GTPv1-U], [GTPv2-C] 601 o VXLAN [RFC7348]. 603 4.2. Wire Protocol Design: Indication of ECN Support 605 This section is intended to guide the redesign of any lower layer 606 protocol that encapsulate IP to add native ECN support at the lower 607 layer. It reflects the approaches used in [RFC6040] and in 608 [RFC5129]. Therefore IP-in-IP tunnels or IP-in-MPLS or MPLS-in-MPLS 609 encapsulations that already comply with [RFC6040] or [RFC5129] will 610 already satisfy this guidance. 612 A lower layer (or subnet) congestion notification system: 614 1. SHOULD NOT apply explicit congestion notifications to PDUs that 615 are destined for legacy layer-4 transport implementations that 616 will not understand ECN, and 618 2. SHOULD NOT apply explicit congestion notifications to PDUs if the 619 egress of the subnet might not propagate congestion notifications 620 onward into the higher layer. 622 We use the term ECN-PDUs for a PDU on a feedback loop that will 623 propagate congestion notification properly because it meets both 624 the above criteria. And a Not-ECN-PDU is a PDU on a feedback 625 loop that does not meet both criteria, and will therefore not 626 propagate congestion notification properly. A corollary of the 627 above is that a lower layer congestion notification protocol: 629 3. SHOULD be able to distinguish ECN-PDUs from Not-ECN-PDUs. 631 Note that there is no need for all interior nodes within a subnet to 632 be able to mark congestion explicitly. A mix of ECN and drop signals 633 from different nodes is fine. However, if _any_ interior nodes might 634 generate ECN markings, guideline 2 above says that all relevant 635 egress node(s) SHOULD be able to propagate those markings up to the 636 higher layer. 638 In IP, if the ECN field in each PDU is cleared to the Not-ECT (not 639 ECN-capable transport) codepoint, it indicates that the L4 transport 640 will not understand congestion markings. A congested buffer must not 641 mark these Not-ECT PDUs, and therefore drops them instead. 643 The mechanism a lower layer uses to distinguish the ECN-capability of 644 PDUs need not mimic that of IP. All the above guidelines say is that 645 the lower layer system, as a whole, should achieve the same outcome. 646 For instance, ECN-capable feedback loops might use PDUs that are 647 identified by a particular set of labels or tags. Alternatively, 648 logical link protocols that use flow state might determine whether a 649 PDU can be congestion marked by checking for ECN-support in the flow 650 state. Other protocols might depend on out-of-band control signals. 652 The per-domain checking of ECN support in MPLS [RFC5129] is a good 653 example of a way to avoid sending congestion markings to transports 654 that will not understand them, without using any header space in the 655 subnet protocol. 657 In MPLS, header space is extremely limited, therefore RFC5129 does 658 not provide a field in the MPLS header to indicate whether the PDU is 659 an ECN-PDU or a Not-ECN-PDU. Instead, interior nodes in a domain are 660 allowed to set explicit congestion indications without checking 661 whether the PDU is destined for a transport that will understand 662 them. Nonetheless, this is made safe by requiring that the network 663 operator upgrades all decapsulating edges of a whole domain at once, 664 as soon as even one switch within the domain is configured to mark 665 rather than drop during congestion. Therefore, any edge node that 666 might decapsulate a packet will be capable of checking whether the 667 higher layer transport is ECN-capable. When decapsulating a CE- 668 marked packet, if the decapsulator discovers that the higher layer 669 (inner header) indicates the transport is not ECN-capable, it drops 670 the packet--effectively on behalf of the earlier congested node (see 671 Decapsulation Guideline 1 in Section 4.4). 673 It was only appropriate to define such an incremental deployment 674 strategy because MPLS is targeted solely at professional operators, 675 who can be expected to ensure that a whole subnetwork is consistently 676 configured. This strategy might not be appropriate for other link 677 technologies targeted at zero-configuration deployment or deployment 678 by the general public (e.g. Ethernet). For such 'plug-and-play' 679 environments it will be necessary to invent a failsafe approach that 680 ensures congestion markings will never fall into black holes, no 681 matter how inconsistently a system is put together. Alternatively, 682 congestion notification relying on correct system configuration could 683 be confined to flavours of Ethernet intended only for professional 684 network operators, such as IEEE 802.1ah Provider Backbone Bridges 685 (PBB). 687 QCN [IEEE802.1Qau] provides another example of how to indicate to 688 lower layer devices that the end-points will not understand ECN. An 689 operator can define certain 802.1p classes of service to indicate 690 non-QCN frames and an ingress bridge is required to map arriving not- 691 QCN-capable IP packets to one of these non-QCN 802.1p classes. 693 4.3. Encapsulation Guidelines 695 This section is intended to guide the redesign of any node that 696 encapsulates IP with a lower layer header when adding native ECN 697 support to the lower layer protocol. It reflects the approaches used 698 in [RFC6040] and in [RFC5129]. Therefore IP-in-IP tunnels or IP-in- 699 MPLS or MPLS-in-MPLS encapsulations that already comply with 700 [RFC6040] or [RFC5129] will already satisfy this guidance. 702 1. Egress Capability Check: A subnet ingress needs to be sure that 703 the corresponding egress of a subnet will propagate any 704 congestion notification added to the outer header across the 705 subnet. This is necessary in addition to checking that an 706 incoming PDU indicates an ECN-capable (L4) transport. Examples 707 of how this guarantee might be provided include: 709 * by configuration (e.g. if any label switches in a domain 710 support ECN marking, [RFC5129] requires all egress nodes to 711 have been configured to propagate ECN) 713 * by the ingress explicitly checking that the egress propagates 714 ECN (e.g. TRILL uses IS-IS to check path capabilities before 715 using critical options [I-D.ietf-trill-rfc7180bis]) 717 * by inherent design of the protocol (e.g. by encoding ECN 718 marking on the outer header in such a way that a legacy egress 719 that does not understand ECN will consider the PDU corrupt and 720 discard it, thus at least propagating a form of congestion 721 signal). 723 2. Egress Fails Capability Check: If the ingress cannot guarantee 724 that the egress will propagate congestion notification, the 725 ingress SHOULD disable ECN when it forwards the PDU at the lower 726 layer. An example of how the ingress might disable ECN at the 727 lower layer would be by setting the outer header of the PDU to 728 identify it as a Not-ECN-PDU, assuming the subnet technology 729 supports such a concept. 731 3. Standard Congestion Monitoring Baseline: Once the ingress to a 732 subnet has established that the egress will correctly propagate 733 ECN, on encapsulation it SHOULD encode the same level of 734 congestion in outer headers as is arriving in incoming headers. 735 For example it might copy any incoming congestion notification 736 into the outer header of the lower layer protocol. 738 This ensures that all outer headers reflect congestion 739 accumulated along the whole upstream path since the Load 740 Regulator, not just since the ingress of the subnet. A node that 741 is not the Load Regulator SHOULD NOT re-initialise the level of 742 CE markings in the outer to zero. 744 This guideline is intended to ensure that any bulk congestion 745 monitoring of outer headers (e.g. by a network management node 746 monitoring ECN in passing frames) is most meaningful. For 747 instance, if an operator measures CE in 0.4% of passing outer 748 headers, this information is only useful if the operator knows 749 where the proportion of CE markings was last initialised to 0% 750 (the Congestion Baseline). Such monitoring information will not 751 be useful if some subnet ingress nodes reset all outer CE 752 markings while others copy incoming CE markings into the outer. 754 Most information can be extracted if the Congestion Baseline is 755 standardised at the node that is regulating the load (the Load 756 Regulator--typically the data source). Then the operator can 757 measure both congestion since the Load Regulator, and congestion 758 since the subnet ingress. The latter might be measurable by 759 subtracting the level of CE markings on inner headers from that 760 on outer headers (see Appendix C of [RFC6040]). 762 4.4. Decapsulation Guidelines 764 This section is intended to guide the redesign of any node that 765 decapsulates IP from within a lower layer header when adding native 766 ECN support to the lower layer protocol. It reflects the approaches 767 used in [RFC6040] and in [RFC5129]. Therefore IP-in-IP tunnels or 768 IP-in-MPLS or MPLS-in-MPLS encapsulations that already comply with 769 [RFC6040] or [RFC5129] will already satisfy this guidance. 771 A subnet egress SHOULD NOT simply copy congestion notification from 772 outer headers to the forwarded header. It SHOULD calculate the 773 outgoing congestion notification field from the inner and outer 774 headers using the following guidelines. If there is any conflict, 775 rules earlier in the list take precedence over rules later in the 776 list: 778 1. If the arriving inner header is a Not-ECN-PDU it implies the L4 779 transport will not understand explicit congestion markings. 780 Then: 782 * If the outer header carries an explicit congestion marking, 783 the packet SHOULD be dropped--the only indication of 784 congestion that the L4 transport will understand. 786 * If the outer is an ECN-PDU that carries no indication of 787 congestion or a Not-ECN-PDU the PDU SHOULD be forwarded, but 788 still as a Not-ECN-PDU. 790 2. If the outer header does not support explicit congestion 791 notification (a Not-ECN-PDU), but the inner header does (an ECN- 792 PDU), the inner header SHOULD be forwarded unchanged. 794 3. In some lower layer protocols congestion may be signalled as a 795 numerical level, such as in the control frames of quantised 796 congestion notification [IEEE802.1Qau]. If such a multi-bit 797 encoding encapsulates an ECN-capable IP data packet, a function 798 will be needed to convert the quantised congestion level into the 799 frequency of congestion markings in outgoing IP packets. 801 4. Congestion indications may be encoded by a severity level. For 802 instance increasing levels of congestion might be encoded by 803 numerically increasing indications, e.g. pre-congestion 804 notification (PCN) can be encoded in each PDU at three severity 805 levels in IP or MPLS [RFC6660]. 807 If the arriving inner header is an ECN-PDU, where the inner and 808 outer headers carry indications of congestion of different 809 severity, the more severe indication SHOULD be forwarded in 810 preference to the less severe. 812 5. The inner and outer headers might carry a combination of 813 congestion notification fields that should not be possible given 814 any currently used protocol transitions. For instance, if 815 Encapsulation Guideline 3 in Section 4.3 had been followed, it 816 should not be possible to have a less severe indication of 817 congestion in the outer than in the inner. It MAY be appropriate 818 to log unexpected combinations of headers and possibly raise an 819 alarm. 821 If a safe outgoing codepoint can be defined for such a PDU, the 822 PDU SHOULD be forwarded rather than dropped. Some implementers 823 discard PDUs with currently unused combinations of headers just 824 in case they represent an attack. However, an approach using 825 alarms and policy-mediated drop is preferable to hard-coded drop, 826 so that operators can keep track of possible attacks but 827 currently unused combinations are not precluded from future use 828 through new standards actions. 830 4.5. Sequences of Similar Tunnels or Subnets 832 In some deployments, particularly in 3GPP networks, an IP packet may 833 traverse two or more IP-in-IP tunnels in sequence that all use 834 identical technology (e.g. GTP). 836 In such cases, it would be sufficient for every encapsulation and 837 decapsulation in the chain to comply with RFC 6040. Alternatively, 838 as an optimisation, a node that decapsulates a packet and immediately 839 re-encapsulates it for the next tunnel MAY copy the incoming outer 840 ECN field directly to the outgoing outer and the incoming inner ECN 841 field directly to the outgoing inner. Then the overall behavior 842 across the sequence of tunnel segments would still be consistent with 843 RFC 6040. 845 Appendix C of RFC6040 describes how a tunnel egress can monitor how 846 much congestion has been introduced within a tunnel. A network 847 operator might want to monitor how much congestion had been 848 introduced within a whole sequence of tunnels. Using the technique 849 in Appendix C of RFC6040 at the final egress, the operator could 850 monitor the whole sequence of tunnels, but only if the above 851 optimisation were used consistently along the sequence of tunnels, in 852 order to make it appear as a single tunnel. Therefore, tunnel 853 endpoint implementations SHOULD allow the operator to configure 854 whether this optimisation is enabled. 856 When ECN support is added to a subnet technology, consideration 857 SHOULD be given to a similar optimisation between subnets in sequence 858 if they all use the same technology. 860 4.6. Reframing and Congestion Markings 862 The guidance in this section is worded in terms of framing 863 boundaries, but it applies equally whether the protocol data units 864 are frames, cells or packets. 866 Where framing boundaries are different between two layers, congestion 867 indications SHOULD be propagated on the basis that a congestion 868 indication on a PDU applies to all the octets in the PDU. On 869 average, an encapsulator or decapsulator SHOULD approximately 870 preserve the number of marked octets arriving and leaving (counting 871 the size of inner headers, but not added encapsulating headers). 873 The next departing frame SHOULD be immediately marked even if only 874 enough incoming marked octets have arrived for part of the departing 875 frame. This ensures that any outstanding congestion marked octets 876 are propagated immediately, rather than held back waiting for a frame 877 no bigger than the outstanding marked octets--which might involve a 878 long wait. 880 For instance, an algorithm for marking departing frames could 881 maintain a counter representing the balance of arriving marked octets 882 minus departing marked octets. It adds the size of every marked 883 frame that arrives and if the counter is positive it marks the next 884 frame to depart and subtracts its size from the counter. This will 885 often leave a negative remainder in the counter, which is deliberate. 887 5. Feed-Up-and-Forward Mode: Guidelines for Adding Congestion 888 Notification 890 The guidance in this section is applicable when IP packets: 892 o are encapsulated in Ethernet headers; 894 o are forwarded by the eNode-B (base station) of a 3GPP radio access 895 network, which is required to apply ECN marking during congestion 896 [LTE-RA]. 898 This guidance also generalises to encapsulation by other subnet 899 technologies with no native support for explicit congestion 900 notification at the lower layer, but with support for finding and 901 processing an IP header. It is unlikely to be applicable or 902 necessary for IP-in-IP encapsulation, where feed-forward-and-up mode 903 based on [RFC6040] would be more appropriate. 905 Marking the IP header while switching at layer-2 (by using a layer-3 906 switch) or while forwarding in a radio access network seems to 907 represent a layering violation. However, it can be considered as a 908 benign optimisation if the guidelines below are followed. Feed-up- 909 and-forward is certainly not a general alternative to implementing 910 feed-forward congestion notification in the lower layer, because: 912 o IPv4 and IPv6 are not the only layer-3 protocols that might be 913 encapsulated by lower layer protocols 915 o Link-layer encryption might be in use, making the layer-2 payload 916 inaccessible 918 o Many Ethernet switches do not have 'layer-3 switch' capabilities 919 so they cannot read or modify an IP payload 921 o It might be costly to find an IP header (v4 or v6) when it may be 922 encapsulated by more than one lower layer header, e.g. Ethernet 923 MAC in MAC [IEEE802.1Qah]. 925 Nonetheless, configuring lower layer equipment to look for an ECN 926 field in an encapsulated IP header is a useful optimisation. If the 927 implementation follows the guidelines below, this optimisation does 928 not have to be confined to a controlled environment such as within a 929 data centre; it could usefully be applied on any network--even if the 930 operator is not sure whether the above issues will never apply: 932 1. If a native lower-layer congestion notification mechanism exists 933 for a subnet technology, it is safe to mix feed-up-and-forward 934 with feed-forward-and-up on other switches in the same subnet. 935 However, it will generally be more efficient to use the native 936 mechanism. 938 2. The depth of the search for an IP header SHOULD be limited. If 939 an IP header is not found soon enough, or an unrecognised or 940 unreadable header is encountered, the switch SHOULD resort to an 941 alternative means of signalling congestion (e.g. drop, or the 942 native lower layer mechanism if available). 944 3. It is sufficient to use the first IP header found in the stack; 945 the egress of the relevant tunnel can propagate congestion 946 notification upwards to any more deeply encapsulated IP headers 947 later. 949 6. Feed-Backward Mode: Guidelines for Adding Congestion Notification 951 It can be seen from Section 3.3 that congestion notification in a 952 subnet using feed-backward mode has generally not been designed to be 953 directly coupled with IP layer congestion notification. The subnet 954 attempts to minimise congestion internally, and if the incoming load 955 at the ingress exceeds the capacity somewhere through the subnet, the 956 layer 3 buffer into the ingress backs up. Thus, a feed-backward mode 957 subnet is in some sense similar to a null mode subnet, in that there 958 is no need for any direct interaction between the subnet and higher 959 layer congestion notification. Therefore no detailed protocol design 960 guidelines are appropriate. Nonetheless, a more general guideline is 961 appropriate: 963 1. A subnetwork technology intended to eventually interface to IP 964 SHOULD NOT be designed using only the feed-backward mode, which 965 is certainly best for a stand-alone subnet, but would need to be 966 modified to work efficiently as part of the wider Internet, 967 because IP uses feed-forward-and-up mode. 969 The feed-backward approach at least works beneath IP, where the term 970 'works' is used only in a narrow functional sense because feed- 971 backward can result in very inefficient and sluggish congestion 972 control--except if it is confined to the subnet directly connected to 973 the original data source, when it is faster than feed-forward. It 974 would be valid to design a protocol that could work in feed-backward 975 mode for paths that only cross one subnet, and in feed-forward-and-up 976 mode for paths that cross subnets. 978 In the early days of TCP/IP, a similar feed-backward approach was 979 tried for explicit congestion signalling, using source-quench (SQ) 980 ICMP control packets. However, SQ fell out of favour and is now 981 formally deprecated [RFC6633]. The main problem was that it is hard 982 for a data source to tell the difference between a spoofed SQ message 983 and a quench request from a genuine buffer on the path. It is also 984 hard for a lower layer buffer to address an SQ message to the 985 original source port number, which may be buried within many layers 986 of headers, and possibly encrypted. 988 Quantised congestion notification (QCN--also known as backward 989 congestion notification or BCN) [IEEE802.1Qau] uses a feed-backward 990 mode structurally similar to ATM's relative rate mechanism. However, 991 QCN confines its applicability to scenarios such as some data centres 992 where all endpoints are directly attached by the same Ethernet 993 technology. If a QCN subnet were later connected into a wider IP- 994 based internetwork (e.g. when attempting to interconnect multiple 995 data centres) it would suffer the inefficiency shown Figure 3. 997 7. IANA Considerations (to be removed by RFC Editor) 999 This memo includes no request to IANA. 1001 8. Security Considerations 1003 If a lower layer wire protocol is redesigned to include explicit 1004 congestion signalling in-band in the protocol header, care SHOULD be 1005 take to ensure that the field used is specified as mutable during 1006 transit. Otherwise interior nodes signalling congestion would 1007 invalidate any authentication protocol applied to the lower layer 1008 header--by altering a header field that had been assumed as 1009 immutable. 1011 The redesign of protocols that encapsulate IP in order to propagate 1012 congestion signals between layers raises potential signal integrity 1013 concerns. Experimental or proposed approaches exist for assuring the 1014 end-to-end integrity of in-band congestion signals, e.g.: 1016 o Congestion exposure (ConEx ) for networks to audit that their 1017 congestion signals are not being suppressed by other networks or 1018 by receivers, and for networks to police that senders are 1019 responding sufficiently to the signals, irrespective of the 1020 transport protocol used [I-D.ietf-conex-abstract-mech]. 1022 o The ECN nonce [RFC3540] for a TCP sender to detect whether a 1023 network or the receiver is suppressing congestion signals. 1025 o A test with the same goals as the ECN nonce, but without the need 1026 for the receiver to co-operate with the protocol 1027 [I-D.moncaster-tcpm-rcv-cheat]. 1029 Given these end-to-end approaches are already being specified, it 1030 would make little sense to attempt to design hop-by-hop congestion 1031 signal integrity into a new lower layer protocol, because end-to-end 1032 integrity inherently achieves hop-by-hop integrity. 1034 9. Conclusions 1036 Following the guidance in the document enables ECN support to be 1037 extended to numerous protocols that encapsulate IP (v4 & v6) in a 1038 consistent way, so that IP continues to fulfil its role as an end-to- 1039 end interoperability layer. This includes: 1041 o A wide range of tunnelling protocols with various forms of shim 1042 header between two IP headers; 1044 o A wide range of subnet technologies, particularly those that work 1045 in the same 'feed-forward-and-up' mode that is used to support ECN 1046 in IP and MPLS. 1048 Guidelines have been defined for supporting propagation of ECN 1049 between Ethernet and IP on so-called Layer-3 Ethernet switches, using 1050 a 'feed-up-an-forward' mode. This approach could enable other subnet 1051 technologies to pass ECN signals into the IP layer, even if they do 1052 not support ECN natively. 1054 Finally, attempting to add ECN to a subnet technology in feed- 1055 backward mode is deprecated except in special cases, due to its 1056 likely sluggish response to congestion. 1058 10. Acknowledgements 1060 Thanks to Gorry Fairhurst for extensive reviews. Thanks also to the 1061 following reviewers: Ingemar Johansson and Piers O'Hanlon and Michael 1062 Welzl, who pointed out that lower layer congestion notification 1063 signals may have different semantics to those in IP. 1065 Bob Briscoe was part-funded by the European Community under its 1066 Seventh Framework Programme through the Trilogy project (ICT-216372) 1067 for initial drafts and through the Reducing Internet Transport 1068 Latency (RITE) project (ICT-317700) subsequently. The views 1069 expressed here are solely those of the authors. 1071 11. Comments Solicited 1073 Comments and questions are encouraged and very welcome. They can be 1074 addressed to the IETF Transport Area working group mailing list 1075 , and/or to the authors. 1077 12. References 1079 12.1. Normative References 1081 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 1082 Requirement Levels", BCP 14, RFC 2119, March 1997. 1084 [RFC3168] Ramakrishnan, K., Floyd, S., and D. Black, "The Addition 1085 of Explicit Congestion Notification (ECN) to IP", RFC 1086 3168, September 2001. 1088 [RFC3819] Karn, P., Bormann, C., Fairhurst, G., Grossman, D., 1089 Ludwig, R., Mahdavi, J., Montenegro, G., Touch, J., and L. 1090 Wood, "Advice for Internet Subnetwork Designers", BCP 89, 1091 RFC 3819, July 2004. 1093 [RFC4774] Floyd, S., "Specifying Alternate Semantics for the 1094 Explicit Congestion Notification (ECN) Field", BCP 124, 1095 RFC 4774, November 2006. 1097 [RFC5129] Davie, B., Briscoe, B., and J. Tay, "Explicit Congestion 1098 Marking in MPLS", RFC 5129, January 2008. 1100 [RFC6040] Briscoe, B., "Tunnelling of Explicit Congestion 1101 Notification", RFC 6040, November 2010. 1103 12.2. Informative References 1105 [ATM-TM-ABR] 1106 Cisco, "Understanding the Available Bit Rate (ABR) Service 1107 Category for ATM VCs", Design Technote 10415, June 2005. 1109 [Buck00] Buckwalter, J., "Frame Relay: Technology and Practice", 1110 Pub. Addison Wesley ISBN-13: 978-0201485240, 2000. 1112 [DCTCP] Alizadeh, M., Greenberg, A., Maltz, D., Padhye, J., Patel, 1113 P., Prabhakar, B., Sengupta, S., and M. Sridharan, "Data 1114 Center TCP (DCTCP)", ACM SIGCOMM CCR 40(4)63--74, October 1115 2010, . 1117 [GTPv1] 3GPP, "GPRS Tunnelling Protocol (GTP) across the Gn and Gp 1118 interface", Technical Specification TS 29.060, . 1120 [GTPv1-U] 3GPP, "General Packet Radio System (GPRS) Tunnelling 1121 Protocol User Plane (GTPv1-U)", Technical Specification TS 1122 29.281, . 1124 [GTPv2-C] 3GPP, "Evolved General Packet Radio Service (GPRS) 1125 Tunnelling Protocol for Control plane (GTPv2-C)", 1126 Technical Specification TS 29.274, . 1128 [I-D.ietf-conex-abstract-mech] 1129 Mathis, M. and B. Briscoe, "Congestion Exposure (ConEx) 1130 Concepts, Abstract Mechanism and Requirements", draft- 1131 ietf-conex-abstract-mech-13 (work in progress), October 1132 2014. 1134 [I-D.ietf-trill-rfc7180bis] 1135 Eastlake, D., Zhang, M., Perlman, R., Banerjee, A., 1136 Ghanwani, A., and S. Gupta, "TRILL: Clarifications, 1137 Corrections, and Updates", draft-ietf-trill-rfc7180bis-00 1138 (work in progress), November 2014. 1140 [I-D.moncaster-tcpm-rcv-cheat] 1141 Moncaster, T., Briscoe, B., and A. Jacquet, "A TCP Test to 1142 Allow Senders to Identify Receiver Non-Compliance", draft- 1143 moncaster-tcpm-rcv-cheat-03 (work in progress), July 2014. 1145 [IEEE802.1Qah] 1146 IEEE, "IEEE Standard for Local and Metropolitan Area 1147 Networks--Virtual Bridged Local Area Networks--Amendment 1148 6: Provider Backbone Bridges", IEEE Std 802.1Qah-2008, 1149 August 2008, 1150 . 1152 (Access Controlled link within page) 1154 [IEEE802.1Qau] 1155 Finn, N., Ed., "IEEE Standard for Local and Metropolitan 1156 Area Networks--Virtual Bridged Local Area Networks - 1157 Amendment 13: Congestion Notification", IEEE Std 802.1Qau- 1158 2010, March 2010, . 1161 (Access Controlled link within page) 1163 [ITU-T.I.371] 1164 ITU-T, "Traffic Control and Congestion Control in B-ISDN", 1165 ITU-T Rec. I.371 (03/04), March 2004, 1166 . 1169 [LTE-RA] 3GPP, "Evolved Universal Terrestrial Radio Access (E-UTRA) 1170 and Evolved Universal Terrestrial Radio Access Network 1171 (E-UTRAN); Overall description; Stage 2", Technical 1172 Specification TS 36.300, . 1174 [RFC1323] Jacobson, V., Braden, B., and D. Borman, "TCP Extensions 1175 for High Performance", RFC 1323, May 1992. 1177 [RFC1701] Hanks, S., Li, T., Farinacci, D., and P. Traina, "Generic 1178 Routing Encapsulation (GRE)", RFC 1701, October 1994. 1180 [RFC2003] Perkins, C., "IP Encapsulation within IP", RFC 2003, 1181 October 1996. 1183 [RFC2637] Hamzeh, K., Pall, G., Verthein, W., Taarud, J., Little, 1184 W., and G. Zorn, "Point-to-Point Tunneling Protocol", RFC 1185 2637, July 1999. 1187 [RFC2661] Townsley, W., Valencia, A., Rubens, A., Pall, G., Zorn, 1188 G., and B. Palter, "Layer Two Tunneling Protocol "L2TP"", 1189 RFC 2661, August 1999. 1191 [RFC2784] Farinacci, D., Li, T., Hanks, S., Meyer, D., and P. 1192 Traina, "Generic Routing Encapsulation (GRE)", RFC 2784, 1193 March 2000. 1195 [RFC2884] Hadi Salim, J. and U. Ahmed, "Performance Evaluation of 1196 Explicit Congestion Notification (ECN) in IP Networks", 1197 RFC 2884, July 2000. 1199 [RFC2983] Black, D., "Differentiated Services and Tunnels", RFC 1200 2983, October 2000. 1202 [RFC3540] Spring, N., Wetherall, D., and D. Ely, "Robust Explicit 1203 Congestion Notification (ECN) Signaling with Nonces", RFC 1204 3540, June 2003. 1206 [RFC4301] Kent, S. and K. Seo, "Security Architecture for the 1207 Internet Protocol", RFC 4301, December 2005. 1209 [RFC6633] Gont, F., "Deprecation of ICMP Source Quench Messages", 1210 RFC 6633, May 2012. 1212 [RFC6660] Briscoe, B., Moncaster, T., and M. Menth, "Encoding Three 1213 Pre-Congestion Notification (PCN) States in the IP Header 1214 Using a Single Diffserv Codepoint (DSCP)", RFC 6660, July 1215 2012. 1217 [RFC7348] Mahalingam, M., Dutt, D., Duda, K., Agarwal, P., Kreeger, 1218 L., Sridhar, T., Bursell, M., and C. Wright, "Virtual 1219 eXtensible Local Area Network (VXLAN): A Framework for 1220 Overlaying Virtualized Layer 2 Networks over Layer 3 1221 Networks", RFC 7348, August 2014. 1223 Appendix A. Outstanding Document Issues 1225 1. [GF] Concern that certain guidelines warrant a MUST (NOT) rather 1226 than a SHOULD (NOT). Given the guidelines say that if any SHOULD 1227 (NOT)s are not followed, a strong justification will be needed, 1228 they have been left as SHOULD (NOT) pending further list 1229 discussion. In particular: 1231 * If inner is a Not-ECN-PDU and Outer is CE (or highest severity 1232 congestion level), MUST (not SHOULD) drop? 1234 2. Consider whether an IETF Standard Track doc will be needed to 1235 Update the IP-in-IP protocols listed in Section 4.1--at least 1236 those that the IET 1238 Appendix B. Changes in This Version (to be removed by RFC Editor) 1240 From ietf-00 to ietf-01: Updated references. 1242 From briscoe-04 to ietf-00: Changed filename following tsvwg 1243 adoption. 1245 From briscoe-03 to 04: 1247 * Re-arranged the introduction to describe the purpose of the 1248 document first before introducing ECN in more depth. And 1249 clarified the introduction throughout. 1251 * Added applicability to 3GPP TS 36.300. 1253 From briscoe-02 to 03: 1255 * Scope section: 1257 + Added dependence on correct propagation of traffic class 1258 information 1260 + For the feed-backward mode, deemed multicast and anycast out 1261 of scope 1263 * Ensured all guidelines referring to subnet technologies also 1264 refer to tunnels and vice versa by adding applicability 1265 sentences at the start of sections 4.1, 4.2, 4.3, 4.4, 4.6 and 1266 5. 1268 * Added Security Considerations on ensuring congestion signal 1269 fields are classed as immutable and on using end-to-end 1270 congestion signal integrity technologies rather than hop-by- 1271 hop. 1273 From briscoe-01 to 02: 1275 * Added authors: JK & PT 1277 * Added 1279 + Section 4.1 "IP-in-IP Tunnels with Tightly Coupled Shim 1280 Headers" 1282 + Section 4.5 "Sequences of Similar Tunnels or Subnets" 1284 + roadmap at the start of Section 4, given the subsections 1285 have become quite fragmented. 1287 + Section 9 "Conclusions" 1289 * Clarified why transports are starting to be able to saturate 1290 interior links 1292 * Under Section 1.1, addressed the question of alternative signal 1293 semantics and included multicast & anycast. 1295 * Under Section 3.1, included a 3GPP example. 1297 * Section 4.2. "Wire Protocol Design": 1299 + Altered guideline 2. to make it clear that it only applies 1300 to the immediate subnet egress, not later ones 1302 + Added a reminder that it is only necessary to check that ECN 1303 propagates at the egress, not whether interior nodes mark 1304 ECN 1306 + Added example of how QCN uses 802.1p to indicate support for 1307 QCN. 1309 * Added references to Appendix C of RFC6040, about monitoring the 1310 amount of congestion signals introduced within a tunnel 1312 * Appendix A: Added more issues to be addressed, including plan 1313 to produce a standards track update to IP-in-IP tunnel 1314 protocols. 1316 * Updated acks and references 1318 From briscoe-00 to 01: 1320 * Intended status: BCP (was Informational) & updates 3819 added. 1322 * Briefer Introduction: Introductory para justifying benefits of 1323 ECN. Moved all but a brief enumeration of modes of operation 1324 to their own new section (from both Intro & Scope). Introduced 1325 incr. deployment as most tricky part. 1327 * Tightened & added to terminology section 1329 * Structured with Modes of Operation, then Guidelines section for 1330 each mode. 1332 * Tightened up guideline text to remove vagueness / passive voice 1333 / ambiguity and highlight main guidelines as numbered items. 1335 * Added Outstanding Document Issues Appendix 1337 * Updated references 1339 Authors' Addresses 1341 Bob Briscoe 1342 BT 1343 B54/77, Adastral Park 1344 Martlesham Heath 1345 Ipswich IP5 3RE 1346 UK 1348 Phone: +44 1473 645196 1349 EMail: bob.briscoe@bt.com 1350 URI: http://bobbriscoe.net/ 1352 John Kaippallimalil 1353 Huawei 1354 5340 Legacy Drive, Suite 175 1355 Plano, Texas 75024 1356 USA 1358 EMail: john.kaippallimalil@huawei.com 1359 Pat Thaler 1360 Broadcom Corporation 1361 5025 Keane Drive 1362 Carmichael, CA 95608 1363 USA 1365 EMail: pthaler@broadcom.com