idnits 2.17.1 draft-ietf-tsvwg-ecn-encap-guidelines-03.txt: Checking boilerplate required by RFC 5378 and the IETF Trust (see https://trustee.ietf.org/license-info): ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/1id-guidelines.txt: ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/checklist : ---------------------------------------------------------------------------- -- The 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 (September 25, 2015) is 3133 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to lower-maturity documents in RFCs) == Missing Reference: 'GF' is mentioned on line 1262, but not defined == Outdated reference: A later version (-07) exists of draft-ietf-trill-rfc7180bis-05 -- Obsolete informational reference (is this intentional?): RFC 1323 (Obsoleted by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 Simula Research Laboratory 4 Updates: 3819 (if approved) J. Kaippallimalil 5 Intended status: Best Current Practice Huawei 6 Expires: March 28, 2016 P. Thaler 7 Broadcom Corporation 8 September 25, 2015 10 Guidelines for Adding Congestion Notification to Protocols that 11 Encapsulate IP 12 draft-ietf-tsvwg-ecn-encap-guidelines-03 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 March 28, 2016. 43 Copyright Notice 45 Copyright (c) 2015 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. Guidelines in All Cases . . . . . . . . . . . . . . . . . . . 7 64 4. Modes of Operation . . . . . . . . . . . . . . . . . . . . . 7 65 4.1. Feed-Forward-and-Up Mode . . . . . . . . . . . . . . . . 8 66 4.2. Feed-Up-and-Forward Mode . . . . . . . . . . . . . . . . 10 67 4.3. Feed-Backward Mode . . . . . . . . . . . . . . . . . . . 10 68 4.4. Null Mode . . . . . . . . . . . . . . . . . . . . . . . . 12 69 5. Feed-Forward-and-Up Mode: Guidelines for Adding Congestion 70 Notification . . . . . . . . . . . . . . . . . . . . . . . . 12 71 5.1. IP-in-IP Tunnels with Tightly Coupled Shim Headers . . . 13 72 5.2. Wire Protocol Design: Indication of ECN Support . . . . . 13 73 5.3. Encapsulation Guidelines . . . . . . . . . . . . . . . . 15 74 5.4. Decapsulation Guidelines . . . . . . . . . . . . . . . . 17 75 5.5. Sequences of Similar Tunnels or Subnets . . . . . . . . . 18 76 5.6. Reframing and Congestion Markings . . . . . . . . . . . . 19 77 6. Feed-Up-and-Forward Mode: Guidelines for Adding Congestion 78 Notification . . . . . . . . . . . . . . . . . . . . . . . . 19 79 7. Feed-Backward Mode: Guidelines for Adding Congestion 80 Notification . . . . . . . . . . . . . . . . . . . . . . . . 21 81 8. IANA Considerations (to be removed by RFC Editor) . . . . . . 22 82 9. Security Considerations . . . . . . . . . . . . . . . . . . . 22 83 10. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . 22 84 11. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 23 85 12. Comments Solicited . . . . . . . . . . . . . . . . . . . . . 23 86 13. References . . . . . . . . . . . . . . . . . . . . . . . . . 23 87 13.1. Normative References . . . . . . . . . . . . . . . . . . 23 88 13.2. Informative References . . . . . . . . . . . . . . . . . 24 89 Appendix A. Outstanding Document Issues . . . . . . . . . . . . 28 90 Appendix B. Changes in This Version (to be removed by RFC 91 Editor) . . . . . . . . . . . . . . . . . . . . . . 28 92 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 30 94 1. Introduction 96 The benefits of Explicit Congestion Notification (ECN) described 97 below can only be fully realised if support for ECN is added to the 98 relevant subnetwork technology, as well as to IP. When a lower layer 99 buffer drops a packet obviously it does not just drop at that layer; 100 the packet disappears from all layers. In contrast, when a lower 101 layer marks a packet with ECN, the marking needs to be explicitly 102 propagated up the layers. The same is true if a buffer marks the 103 outer header of a packet that encapsulates inner tunnelled headers. 104 Forwarding ECN is not as straightforward as other headers because it 105 has to be assumed ECN may be only partially deployed. If an egress 106 at any layer is not ECN-aware, or if the ultimate receiver or sender 107 is not ECN-aware, congestion needs to be indicated by dropping a 108 packet, not marking it. 110 The purpose of this document is to guide the addition of congestion 111 notification to any subnet technology or tunnelling protocol, so that 112 lower layer equipment can signal congestion explicitly and it will 113 propagate consistently into encapsulated (higher layer) headers, 114 otherwise the signals will not reach their ultimate destination. 116 ECN is defined in the IP header (v4 & v6) [RFC3168] to allow a 117 resource to notify the onset of queue build-up without having to drop 118 packets, by explicitly marking a proportion of packets with the 119 congestion experienced (CE) codepoint. 121 Given a suitable marking scheme, ECN removes nearly all congestion 122 loss and it cuts delays for two main reasons: 124 o It avoids the delay when recovering from congestion losses, which 125 particularly benefits small flows or real-time flows, making their 126 delivery time predictably short [RFC2884]; 128 o As ECN is used more widely by end-systems, it will gradually 129 remove the need to configure a degree of delay into buffers before 130 they start to notify congestion (the cause of bufferbloat). This 131 is because drop involves a trade-off between sending a timely 132 signal and trying to avoid impairment, whereas ECN is solely a 133 signal not an impairment, so there is no harm triggering it 134 earlier. 136 Some lower layer technologies (e.g. MPLS, Ethernet) are used to form 137 subnetworks with IP-aware nodes only at the edges. These networks 138 are often sized so that it is rare for interior queues to overflow. 139 However, this has often be more due to the inability of the original 140 TCP protocol to saturate the links. For many years, fixes such as 141 window scaling [RFC1323] proved hard to deploy. But now that modern 142 operating systems are finally capable of saturating interior links, 143 even the buffers of well-provisioned interior switches will need to 144 signal episodes of queuing. 146 Propagation of ECN is defined for MPLS [RFC5129], and is being 147 defined for TRILL [I-D.ietf-trill-rfc7180bis], but it remains to be 148 defined for a number of other subnetwork technologies. 150 Similarly, ECN propagation is yet to be defined for many tunnelling 151 protocols. [RFC6040] defines how ECN should be propagated for IP-in- 152 IP [RFC2003] and IPsec [RFC4301] tunnels. However, as Section 9.3 of 153 RFC3168 pointed out, ECN support will need to be defined for other 154 tunnelling protocols, e.g. L2TP [RFC2661], GRE [RFC1701], [RFC2784], 155 PPTP [RFC2637] and GTP [GTPv1], [GTPv1-U], [GTPv2-C]. 157 Incremental deployment is the most tricky aspect when adding support 158 for ECN. The original ECN protocol in IP [RFC3168] was carefully 159 designed so that a congested buffer would not mark a packet (rather 160 than drop it) unless both source and destination hosts were ECN- 161 capable. Otherwise its congestion markings would never be detected 162 and congestion would just deteriorate further. However, to support 163 congestion marking below the IP layer, it is not sufficient to only 164 check that the two end-points support ECN; correct operation also 165 depends on the decapsulator at each subnet egress faithfully 166 propagating congestion notifications to the higher layer. Otherwise, 167 a legacy decapsulator might silently fail to propagate any ECN 168 signals from the outer to the forwarded header. Then the lost 169 signals would never be detected and again congestion would 170 deteriorate further. The guidelines given later require protocol 171 designers to carefully consider incremental deployment, and suggest 172 various safe approaches for different circumstances. 174 Of course, the IETF does not have standards authority over every link 175 layer protocol. So this document gives guidelines for designing 176 propagation of congestion notification across the interface between 177 IP and protocols that may encapsulate IP (i.e. that can be layered 178 beneath IP). Each lower layer technology will exhibit different 179 issues and compromises, so the IETF or the relevant standards body 180 must be free to define the specifics of each lower layer congestion 181 notification scheme. Nonetheless, if the guidelines are followed, 182 congestion notification should interwork between different 183 technologies, using IP in its role as a 'portability layer'. 185 Therefore, the capitalised term 'SHOULD' or 'SHOULD NOT' are often 186 used in preference to 'MUST' or 'MUST NOT', because it is difficult 187 to know the compromises that will be necessary in each protocol 188 design. If a particular protocol design chooses to contradict a 189 'SHOULD (NOT)' given in the advice below, it MUST include a sound 190 justification. 192 It has not been possible to give common guidelines for all lower 193 layer technologies, because they do not all fit a common pattern. 194 Instead they have been divided into a few distinct modes of 195 operation: feed-forward-and-upward; feed-upward-and-forward; feed- 196 backward; and null mode. These modes are described in Section 4, 197 then in the following sections separate guidelines are given for each 198 mode. 200 This document updates the advice to subnetwork designers about ECN in 201 Section 13 of [RFC3819]. 203 1.1. Scope 205 This document only concerns wire protocol processing of explicit 206 notification of congestion and makes no changes or recommendations 207 concerning algorithms for congestion marking or for congestion 208 response (algorithm issues should be independent of the layer the 209 algorithm operates in). 211 The question of congestion notification signals with different 212 semantics to those of ECN in IP is touched on in a couple of specific 213 cases (e.g. QCN [IEEE802.1Qau]) and with schemes with multiple 214 severity levels such as PCN [RFC6660]). However, no attempt is made 215 to give guidelines about schemes with different semantics that are 216 yet to be invented. 218 The semantics of congestion signals can be relative to the traffic 219 class. Therefore correct propagation of congestion signals could 220 depend on correct propagation of any traffic class field between the 221 layers. In this document, correct propagation of traffic class 222 information is assumed, while what 'correct' means and how it is 223 achieved is covered elsewhere (e.g. [RFC2983]) and is outside the 224 scope of the present document. 226 Note that these guidelines do not require the subnet wire protocol to 227 be changed to accommodate congestion notification. Another way to 228 add congestion notification without consuming header space in the 229 subnet protocol might be to use a parallel control plane protocol. 231 This document focuses on the congestion notification interface 232 between IP and lower layer protocols that can encapsulate IP, where 233 the term 'IP' includes v4 or v6, unicast, multicast or anycast. 234 However, it is likely that the guidelines will also be useful when a 235 lower layer protocol or tunnel encapsulates itself (e.g. Ethernet 236 MAC in MAC [IEEE802.1Qah]) or when it encapsulates other protocols. 238 In the feed-backward mode, propagation of congestion signals for 239 multicast and anycast packets is out-of-scope (because it would be so 240 complicated that it is hoped no-one would attempt such an 241 abomination). 243 2. Terminology 245 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 246 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 247 document are to be interpreted as described in RFC 2119 [RFC2119]. 249 Further terminology used within this document: 251 Protocol data unit (PDU): Information that is delivered as a unit 252 among peer entities of a layered network consisting of protocol 253 control information (typically a header) and possibly user data 254 (payload) of that layer. The scope of this document includes 255 layer 2 and layer 3 networks, where the PDU is respectively termed 256 a frame or a packet (or a cell in ATM). PDU is a general term for 257 any of these. This definition also includes a payload with a shim 258 header lying somewhere between layer 2 & 3. 260 Transport: The end-to-end transmission control function, 261 conventionally considered at layer-4 in the OSI reference model. 262 Given the audience for this document will often use the word 263 transport to mean low level bit carriage, whenever the term is 264 used it will be qualified, e.g. 'L4 transport'. 266 Encapsulator: The link or tunnel endpoint function that adds an 267 outer header to a PDU (also termed the 'link ingress', the 'subnet 268 ingress', the 'ingress tunnel endpoint' or just the 'ingress' 269 where the context is clear). 271 Decapsulator: The link or tunnel endpoint function that removes an 272 outer header from a PDU (also termed the 'link egress', the 273 'subnet egress', the 'egress tunnel endpoint' or just the 'egress' 274 where the context is clear). 276 Incoming header: The header of an arriving PDU before encapsulation. 278 Outer header: The header added to encapsulate a PDU. 280 Inner header: The header encapsulated by the outer header. 282 Outgoing header: The header forwarded by the decapsulator. 284 CE: Congestion Experienced [RFC3168] 285 ECT: ECN-Capable Transport [RFC3168] 287 Not-ECT: Not ECN-Capable Transport [RFC3168] 289 ECN-PDU: A PDU that is part of a feedback loop within which all the 290 nodes that need to propagate explicit congestion notifications 291 back to the Load Regulator are ECN-capable. An IP packet with a 292 non-zero ECN field implies that the endpoints are ECN-capable, so 293 this would be an ECN-PDU. However, ECN-PDU is intended to be a 294 general term for a PDU at any layer, not just IP. 296 Not-ECN-PDU: A PDU that is part of a feedback-loop within which some 297 nodes necessary to propagate explicit congestion notifications 298 back to the load regulator are not ECN-capable. 300 Load Regulator: For each flow of PDUs, the transport function that 301 is capable of controlling the data rate. Typically located at the 302 data source, but in-path nodes can regulate load in some 303 congestion control arrangements (e.g. admission control or 304 policing nodes). Note the term "a function capable of controlling 305 the load" deliberately includes a transport that doesn't actually 306 control the load but ideally it ought to (e.g. a sending 307 application without congestion control that uses UDP). 309 Congestion Baseline: The location of the function on the path that 310 initialised the values of all congestion notification fields in a 311 sequence of packets, before any are set to the congestion 312 experienced (CE) codepoint if they experience congestion further 313 downstream. Typically the original data source at layer-4. 315 3. Guidelines in All Cases 317 RFC 3168 specifies that the ECN field in the IP header is intended to 318 be marked by active queue management algorithms. Any congestion 319 notification from an algorithm that does not conform to the 320 recommendations in [I-D.ietf-aqm-recommendation] MUST NOT be 321 propagated from a lower layer into the ECN field in IP (see also 322 [RFC4774] on alternate uses of the ECN field). 324 4. Modes of Operation 326 This section sets down the different modes by which congestion 327 information is passed between the lower layer and the higher one. It 328 acts as a reference framework for the following sections, which give 329 normative guidelines for designers of explicit congestion 330 notification protocols, taking each mode in turn: 332 Feed-Forward-and-Up: Nodes feed forward congestion notification 333 towards the egress within the lower layer then up and along the 334 layers towards the end-to-end destination at the transport layer. 335 The following local optimisation is possible: 337 Feed-Up-and-Forward: A lower layer switch feeds-up congestion 338 notification directly into the ECN field in the higher layer 339 (e.g. IP) header, irrespective of whether the node is at the 340 egress of a subnet. 342 Feed-Backward: Nodes feed back congestion signals towards the 343 ingress of the lower layer and (optionally) attempt to control 344 congestion within their own layer. 346 Null: Nodes cannot experience congestion at the lower layer except 347 at ingress nodes (which are IP-aware or equivalently higher-layer- 348 aware). 350 4.1. Feed-Forward-and-Up Mode 352 Like IP and MPLS, many subnet technologies are based on self- 353 contained protocol data units (PDUs) or frames sent unreliably. They 354 provide no feedback channel at the subnetwork layer, instead relying 355 on higher layers (e.g. TCP) to feed back loss signals. 357 In these cases, ECN may best be supported by standardising explicit 358 notification of congestion into the lower layer protocol that carries 359 the data forwards. It will then also be necessary to define how the 360 egress of the lower layer subnet propagates this explicit signal into 361 the forwarded upper layer (IP) header. It can then continue forwards 362 until it finally reaches the destination transport (at L4). Then 363 typically the destination will feed this congestion notification back 364 to the source transport using an end-to-end protocol (e.g. TCP). 365 This is the arrangement that has already been used to add ECN to IP- 366 in-IP tunnels [RFC6040], IP-in-MPLS and MPLS-in-MPLS [RFC5129]. 368 This mode is illustrated in Figure 1. Along the middle of the 369 figure, layers 2, 3 & 4 of the protocol stack are shown, and one 370 packet is shown along the bottom as it progresses across the network 371 from source to destination, crossing two subnets connected by a 372 router, and crossing two switches on the path across each subnet. 373 Congestion at the output of the first switch (shown as *) leads to a 374 congestion marking in the L2 header (shown as C in the illustration 375 of the packet). The chevrons show the progress of the resulting 376 congestion indication. It is propagated from link to link across the 377 subnet in the L2 header, then when the router removes the marked L2 378 header, it propagates the marking up into the L3 (IP) header. The 379 router forwards the marked L3 header into subnet 2, and when it adds 380 a new L2 header it copies the L3 marking into the L2 header as well, 381 as shown by the 'C's in both layers (assuming the technology of 382 subnet 2 also supports explicit congestion marking). 384 Note that there is no implication that each 'C' marking is encoded 385 the same; a different encoding might be used for the 'C' marking in 386 each protocol. 388 Finally, for completeness, we show the L3 marking arriving at the 389 destination, where the host transport protocol (e.g. TCP) feeds it 390 back to the source in the L4 acknowledgement (the 'C' at L4 in the 391 packet at the top of the diagram). 393 _ _ _ 394 /_______ | | |C| ACK Packet (V) 395 \ |_|_|_| 396 +---+ layer: 2 3 4 header +---+ 397 | <|<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<< Packet V <<<<<<<<<<<<<|<< |L4 398 | | +---+ | ^ | 399 | | . . . . . . Packet U. . | >>|>>> Packet U >>>>>>>>>>>>|>^ |L3 400 | | +---+ +---+ | ^ | +---+ +---+ | | 401 | | | *|>>>>>|>>>|>>>>>|>^ | | | | | | |L2 402 |___|_____|___|_____|___|_____|___|_____|___|_____|___|_____|___| 403 source subnet A router subnet B dest 404 __ _ _ _ __ _ _ _ __ _ _ __ _ _ _ 405 | | | | | | | | |C| | | |C| | | |C|C| Data________\ 406 |__|_|_|_| |__|_|_|_| |__|_|_| |__|_|_|_| Packet (U) / 407 layer: 4 3 2A 4 3 2A 4 3 4 3 2B 408 header 410 Figure 1: Feed-Forward-and-Up Mode 412 Of course, modern networks are rarely as simple as this text-book 413 example, often involving multiple nested layers. For example, a 3GPP 414 mobile network may have two IP-in-IP (GTP) tunnels in series and an 415 MPLS backhaul between the base station and the first router. 416 Nonetheless, the example illustrates the general idea of feeding 417 congestion notification forward then upward whenever a header is 418 removed at the egress of a subnet. 420 Note that the FECN (forward ECN) bit in Frame Relay and the explicit 421 forward congestion indication (EFCI [ITU-T.I.371]) bit in ATM user 422 data cells follow a feed-forward pattern. However, in ATM, this is 423 only as part of a feed-forward-and-backward pattern at the lower 424 layer, not feed-forward-and-up out of the lower layer--the intention 425 was never to interface to IP ECN at the subnet egress. To our 426 knowledge, Frame Relay FECN is solely used to detect where more 427 capacity should be provisioned [Buck00]. 429 4.2. Feed-Up-and-Forward Mode 431 Ethernet is particularly difficult to extend incrementally to support 432 explicit congestion notification. One way to support ECN in such 433 cases has been to use so called 'layer-3 switches'. These are 434 Ethernet switches that bury into the Ethernet payload to find an IP 435 header and manipulate or act on certain IP fields (specifically 436 Diffserv & ECN). For instance, in Data Center TCP [DCTCP], layer-3 437 switches are configured to mark the ECN field of the IP header within 438 the Ethernet payload when their output buffer becomes congested. 439 With respect to switching, a layer-3 switch acts solely on the 440 addresses in the Ethernet header; it doesn't use IP addresses, and it 441 doesn't decrement the TTL field in the IP header. 443 _ _ _ 444 /_______ | | |C| ACK packet (V) 445 \ |_|_|_| 446 +---+ layer: 2 3 4 header +---+ 447 | <|<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<< Packet V <<<<<<<<<<<<<|<< |L4 448 | | +---+ | ^ | 449 | | . . . >>>> Packet U >>>|>>>|>>> Packet U >>>>>>>>>>>>|>^ |L3 450 | | +--^+ +---+ | | +---+ +---+ | | 451 | | | *| | | | | | | | | | |L2 452 |___|_____|___|_____|___|_____|___|_____|___|_____|___|_____|___| 453 source subnet E router subnet F dest 454 __ _ _ _ __ _ _ _ __ _ _ __ _ _ _ 455 | | | | | | | |C| | | | |C| | | |C|C| data________\ 456 |__|_|_|_| |__|_|_|_| |__|_|_| |__|_|_|_| packet (U) / 457 layer: 4 3 2 4 3 2 4 3 4 3 2 458 header 460 Figure 2: Feed-Up-and-Forward Mode 462 By comparing Figure 2 with Figure 1, it can be seen that subnet E 463 (perhaps a subnet of layer-3 Ethernet switches) works in feed-up-and- 464 forward mode by notifying congestion directly into L3 at the point of 465 congestion, even though the congested switch does not otherwise act 466 at L3. In this example, the technology in subnet F (e.g. MPLS) does 467 support ECN natively, so when the router adds the layer-2 header it 468 copies the ECN marking from L3 to L2 as well. 470 4.3. Feed-Backward Mode 472 In some layer 2 technologies, explicit congestion notification has 473 been defined for use internally within the subnet with its own 474 feedback and load regulation, but typically the interface with IP for 475 ECN has not been defined. 477 For instance, for the available bit-rate (ABR) service in ATM, the 478 relative rate mechanism was one of the more popular mechanisms for 479 managing traffic, tending to supersede earlier designs. In this 480 approach ATM switches send special resource management (RM) cells in 481 both the forward and backward directions to control the ingress rate 482 of user data into a virtual circuit. If a switch buffer is 483 approaching congestion or congested it sends an RM cell back towards 484 the ingress with respectively the No Increase (NI) or Congestion 485 Indication (CI) bit set in its message type field [ATM-TM-ABR]. The 486 ingress then holds or decreases its sending bit-rate accordingly. 488 _ _ _ 489 /_______ | | |C| ACK packet (X) 490 \ |_|_|_| 491 +---+ layer: 2 3 4 header +---+ 492 | <|<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<< Packet X <<<<<<<<<<<<<|<< |L4 493 | | +---+ | ^ | 494 | | | *|>>> Packet W >>>>>>>>>>>>|>^ |L3 495 | | +---+ +---+ | | +---+ +---+ | | 496 | | | | | | | <|<<<<<|<<<|<(V)<|<<<| | |L2 497 | | . . | . |Packet U | . . | . | . . | . | . . | .*| . . | |L2 498 |___|_____|___|_____|___|_____|___|_____|___|_____|___|_____|___| 499 source subnet G router subnet H dest 500 __ _ _ _ __ _ _ _ __ _ _ __ _ _ _ later 501 | | | | | | | | | | | | | | | | |C| | data________\ 502 |__|_|_|_| |__|_|_|_| |__|_|_| |__|_|_|_| packet (W) / 503 4 3 2 4 3 2 4 3 4 3 2 504 _ 505 /__ |C| Feedback control 506 \ |_| cell/frame (V) 507 2 508 __ _ _ _ __ _ _ _ __ _ _ __ _ _ _ earlier 509 | | | | | | | | | | | | | | | | | | | data________\ 510 |__|_|_|_| |__|_|_|_| |__|_|_| |__|_|_|_| packet (U) / 511 layer: 4 3 2 4 3 2 4 3 4 3 2 512 header 514 Figure 3: Feed-Backward Mode 516 ATM's feed-backward approach doesn't fit well when layered beneath 517 IP's feed-forward approach--unless the initial data source is the 518 same node as the ATM ingress. Figure 3 shows the feed-backward 519 approach being used in subnet H. If the final switch on the path is 520 congested (*), it doesn't feed-forward any congestion indications on 521 packet (U). Instead it sends a control cell (V) back to the router 522 at the ATM ingress. 524 However, the backward feedback doesn't reach the original data source 525 directly because IP doesn't support backward feedback (and subnet G 526 is independent of subnet H). Instead, the router in the middle 527 throttles down its sending rate but the original data sources don't 528 reduce their rates. The resulting rate mismatch causes the middle 529 router's buffer at layer 3 to back up until it becomes congested, 530 which it signals forwards on later data packets at layer 3 (e.g. 531 packet W). Note that the forward signal from the middle router is 532 not triggered directly by the backward signal. Rather, it is 533 triggered by congestion resulting from the middle router's mismatched 534 rate response to the backward signal. 536 In response to this later forward signalling, end-to-end feedback at 537 layer-4 finally completes the tortuous path of congestion indications 538 back to the origin data source, as before. 540 4.4. Null Mode 542 Often link and physical layer resources are 'non-blocking' by design. 543 In these cases congestion notification may be implemented but it does 544 not need to be deployed at the lower layer; ECN in IP would be 545 sufficient. 547 A degenerate example is a point-to-point Ethernet link. Excess 548 loading of the link merely causes the queue from the higher layer to 549 back up, while the lower layer remains immune to congestion. Even a 550 whole meshed subnetwork can be made immune to interior congestion by 551 limiting ingress capacity and careful sizing of links, particularly 552 if multi-path routing is used to ensure even worst-case patterns of 553 load cannot congest any link. 555 5. Feed-Forward-and-Up Mode: Guidelines for Adding Congestion 556 Notification 558 Feed-forward-and-up is the mode already used for signalling ECN up 559 the layers through MPLS into IP [RFC5129] and through IP-in-IP 560 tunnels [RFC6040]. These RFCs take a consistent approach and the 561 following guidelines are designed to ensure this consistency 562 continues as ECN support is added to other protocols that encapsulate 563 IP. The guidelines are also designed to ensure compliance with the 564 more general best current practice for the design of alternate ECN 565 schemes given in [RFC4774]. 567 The rest of this section is structured as follows: 569 o Section 5.1 addresses the most straightforward cases, where 570 [RFC6040] can be applied directly to add ECN to tunnels that are 571 effectively the same as IP-in-IP tunnels. 573 o The subsequent sections give guidelines for adding ECN to a subnet 574 technology that uses feed-forward-and-up mode like IP, but it is 575 not so similar to IP that [RFC6040] rules can be applied directly. 576 Specifically: 578 * Sections 5.2, 5.3 and 5.4 respectively address how to add ECN 579 support to the wire protocol and to the encapsulators and 580 decapsulators at the ingress and egress of the subnet. 582 * Section 5.5 deals with the special, but common, case of 583 sequences of tunnels or subnets that all use the same 584 technology 586 * Section 5.6 deals with the question of reframing when IP 587 packets do not map 1:1 into lower layer frames. 589 5.1. IP-in-IP Tunnels with Tightly Coupled Shim Headers 591 A common pattern for many tunnelling protocols is to encapsulate an 592 inner IP header with shim header(s) then an outer IP header. In many 593 cases the shim header(s) always have to be tightly coupled to the 594 outer IP header because they are not sufficient as outer headers in 595 their own right. In such cases the shim header(s) and the outer IP 596 header are always added (or removed) in the same operation. 597 Therefore, in all such tightly coupled IP-in-IP tunnelling protocols, 598 the rules in [RFC6040] for propagating the ECN field between the two 599 IP headers SHOULD be applied directly. 601 Examples of tightly coupled IP-in-IP tunnelling protocols where 602 [RFC6040] can be applied directly are: 604 o L2TP [RFC2661] 606 o GRE [RFC1701], [RFC2784] 608 o PPTP [RFC2637] 610 o GTP [GTPv1], [GTPv1-U], [GTPv2-C] 612 o VXLAN [RFC7348]. 614 5.2. Wire Protocol Design: Indication of ECN Support 616 This section is intended to guide the redesign of any lower layer 617 protocol that encapsulate IP to add native ECN support at the lower 618 layer. It reflects the approaches used in [RFC6040] and in 619 [RFC5129]. Therefore IP-in-IP tunnels or IP-in-MPLS or MPLS-in-MPLS 620 encapsulations that already comply with [RFC6040] or [RFC5129] will 621 already satisfy this guidance. 623 A lower layer (or subnet) congestion notification system: 625 1. SHOULD NOT apply explicit congestion notifications to PDUs that 626 are destined for legacy layer-4 transport implementations that 627 will not understand ECN, and 629 2. SHOULD NOT apply explicit congestion notifications to PDUs if the 630 egress of the subnet might not propagate congestion notifications 631 onward into the higher layer. 633 We use the term ECN-PDUs for a PDU on a feedback loop that will 634 propagate congestion notification properly because it meets both 635 the above criteria. And a Not-ECN-PDU is a PDU on a feedback 636 loop that does not meet both criteria, and will therefore not 637 propagate congestion notification properly. A corollary of the 638 above is that a lower layer congestion notification protocol: 640 3. SHOULD be able to distinguish ECN-PDUs from Not-ECN-PDUs. 642 Note that there is no need for all interior nodes within a subnet to 643 be able to mark congestion explicitly. A mix of ECN and drop signals 644 from different nodes is fine. However, if _any_ interior nodes might 645 generate ECN markings, guideline 2 above says that all relevant 646 egress node(s) SHOULD be able to propagate those markings up to the 647 higher layer. 649 In IP, if the ECN field in each PDU is cleared to the Not-ECT (not 650 ECN-capable transport) codepoint, it indicates that the L4 transport 651 will not understand congestion markings. A congested buffer must not 652 mark these Not-ECT PDUs, and therefore drops them instead. 654 The mechanism a lower layer uses to distinguish the ECN-capability of 655 PDUs need not mimic that of IP. All the above guidelines say is that 656 the lower layer system, as a whole, should achieve the same outcome. 657 For instance, ECN-capable feedback loops might use PDUs that are 658 identified by a particular set of labels or tags. Alternatively, 659 logical link protocols that use flow state might determine whether a 660 PDU can be congestion marked by checking for ECN-support in the flow 661 state. Other protocols might depend on out-of-band control signals. 663 The per-domain checking of ECN support in MPLS [RFC5129] is a good 664 example of a way to avoid sending congestion markings to transports 665 that will not understand them, without using any header space in the 666 subnet protocol. 668 In MPLS, header space is extremely limited, therefore RFC5129 does 669 not provide a field in the MPLS header to indicate whether the PDU is 670 an ECN-PDU or a Not-ECN-PDU. Instead, interior nodes in a domain are 671 allowed to set explicit congestion indications without checking 672 whether the PDU is destined for a transport that will understand 673 them. Nonetheless, this is made safe by requiring that the network 674 operator upgrades all decapsulating edges of a whole domain at once, 675 as soon as even one switch within the domain is configured to mark 676 rather than drop during congestion. Therefore, any edge node that 677 might decapsulate a packet will be capable of checking whether the 678 higher layer transport is ECN-capable. When decapsulating a CE- 679 marked packet, if the decapsulator discovers that the higher layer 680 (inner header) indicates the transport is not ECN-capable, it drops 681 the packet--effectively on behalf of the earlier congested node (see 682 Decapsulation Guideline 1 in Section 5.4). 684 It was only appropriate to define such an incremental deployment 685 strategy because MPLS is targeted solely at professional operators, 686 who can be expected to ensure that a whole subnetwork is consistently 687 configured. This strategy might not be appropriate for other link 688 technologies targeted at zero-configuration deployment or deployment 689 by the general public (e.g. Ethernet). For such 'plug-and-play' 690 environments it will be necessary to invent a failsafe approach that 691 ensures congestion markings will never fall into black holes, no 692 matter how inconsistently a system is put together. Alternatively, 693 congestion notification relying on correct system configuration could 694 be confined to flavours of Ethernet intended only for professional 695 network operators, such as IEEE 802.1ah Provider Backbone Bridges 696 (PBB). 698 QCN [IEEE802.1Qau] provides another example of how to indicate to 699 lower layer devices that the end-points will not understand ECN. An 700 operator can define certain 802.1p classes of service to indicate 701 non-QCN frames and an ingress bridge is required to map arriving not- 702 QCN-capable IP packets to one of these non-QCN 802.1p classes. 704 5.3. Encapsulation Guidelines 706 This section is intended to guide the redesign of any node that 707 encapsulates IP with a lower layer header when adding native ECN 708 support to the lower layer protocol. It reflects the approaches used 709 in [RFC6040] and in [RFC5129]. Therefore IP-in-IP tunnels or IP-in- 710 MPLS or MPLS-in-MPLS encapsulations that already comply with 711 [RFC6040] or [RFC5129] will already satisfy this guidance. 713 1. Egress Capability Check: A subnet ingress needs to be sure that 714 the corresponding egress of a subnet will propagate any 715 congestion notification added to the outer header across the 716 subnet. This is necessary in addition to checking that an 717 incoming PDU indicates an ECN-capable (L4) transport. Examples 718 of how this guarantee might be provided include: 720 * by configuration (e.g. if any label switches in a domain 721 support ECN marking, [RFC5129] requires all egress nodes to 722 have been configured to propagate ECN) 724 * by the ingress explicitly checking that the egress propagates 725 ECN (e.g. TRILL uses IS-IS to check path capabilities before 726 using critical options [I-D.ietf-trill-rfc7180bis]) 728 * by inherent design of the protocol (e.g. by encoding ECN 729 marking on the outer header in such a way that a legacy egress 730 that does not understand ECN will consider the PDU corrupt and 731 discard it, thus at least propagating a form of congestion 732 signal). 734 2. Egress Fails Capability Check: If the ingress cannot guarantee 735 that the egress will propagate congestion notification, the 736 ingress SHOULD disable ECN when it forwards the PDU at the lower 737 layer. An example of how the ingress might disable ECN at the 738 lower layer would be by setting the outer header of the PDU to 739 identify it as a Not-ECN-PDU, assuming the subnet technology 740 supports such a concept. 742 3. Standard Congestion Monitoring Baseline: Once the ingress to a 743 subnet has established that the egress will correctly propagate 744 ECN, on encapsulation it SHOULD encode the same level of 745 congestion in outer headers as is arriving in incoming headers. 746 For example it might copy any incoming congestion notification 747 into the outer header of the lower layer protocol. 749 This ensures that all outer headers reflect congestion 750 accumulated along the whole upstream path since the Load 751 Regulator, not just since the ingress of the subnet. A node that 752 is not the Load Regulator SHOULD NOT re-initialise the level of 753 CE markings in the outer to zero. 755 This guideline is intended to ensure that any bulk congestion 756 monitoring of outer headers (e.g. by a network management node 757 monitoring ECN in passing frames) is most meaningful. For 758 instance, if an operator measures CE in 0.4% of passing outer 759 headers, this information is only useful if the operator knows 760 where the proportion of CE markings was last initialised to 0% 761 (the Congestion Baseline). Such monitoring information will not 762 be useful if some subnet ingress nodes reset all outer CE 763 markings while others copy incoming CE markings into the outer. 765 Most information can be extracted if the Congestion Baseline is 766 standardised at the node that is regulating the load (the Load 767 Regulator--typically the data source). Then the operator can 768 measure both congestion since the Load Regulator, and congestion 769 since the subnet ingress. The latter might be measurable by 770 subtracting the level of CE markings on inner headers from that 771 on outer headers (see Appendix C of [RFC6040]). 773 5.4. Decapsulation Guidelines 775 This section is intended to guide the redesign of any node that 776 decapsulates IP from within a lower layer header when adding native 777 ECN support to the lower layer protocol. It reflects the approaches 778 used in [RFC6040] and in [RFC5129]. Therefore IP-in-IP tunnels or 779 IP-in-MPLS or MPLS-in-MPLS encapsulations that already comply with 780 [RFC6040] or [RFC5129] will already satisfy this guidance. 782 A subnet egress SHOULD NOT simply copy congestion notification from 783 outer headers to the forwarded header. It SHOULD calculate the 784 outgoing congestion notification field from the inner and outer 785 headers using the following guidelines. If there is any conflict, 786 rules earlier in the list take precedence over rules later in the 787 list: 789 1. If the arriving inner header is a Not-ECN-PDU it implies the L4 790 transport will not understand explicit congestion markings. 791 Then: 793 * If the outer header carries an explicit congestion marking, 794 the packet SHOULD be dropped--the only indication of 795 congestion that the L4 transport will understand. 797 * If the outer is an ECN-PDU that carries no indication of 798 congestion or a Not-ECN-PDU the PDU SHOULD be forwarded, but 799 still as a Not-ECN-PDU. 801 2. If the outer header does not support explicit congestion 802 notification (a Not-ECN-PDU), but the inner header does (an ECN- 803 PDU), the inner header SHOULD be forwarded unchanged. 805 3. In some lower layer protocols congestion may be signalled as a 806 numerical level, such as in the control frames of quantised 807 congestion notification [IEEE802.1Qau]. If such a multi-bit 808 encoding encapsulates an ECN-capable IP data packet, a function 809 will be needed to convert the quantised congestion level into the 810 frequency of congestion markings in outgoing IP packets. 812 4. Congestion indications may be encoded by a severity level. For 813 instance increasing levels of congestion might be encoded by 814 numerically increasing indications, e.g. pre-congestion 815 notification (PCN) can be encoded in each PDU at three severity 816 levels in IP or MPLS [RFC6660]. 818 If the arriving inner header is an ECN-PDU, where the inner and 819 outer headers carry indications of congestion of different 820 severity, the more severe indication SHOULD be forwarded in 821 preference to the less severe. 823 5. The inner and outer headers might carry a combination of 824 congestion notification fields that should not be possible given 825 any currently used protocol transitions. For instance, if 826 Encapsulation Guideline 3 in Section 5.3 had been followed, it 827 should not be possible to have a less severe indication of 828 congestion in the outer than in the inner. It MAY be appropriate 829 to log unexpected combinations of headers and possibly raise an 830 alarm. 832 If a safe outgoing codepoint can be defined for such a PDU, the 833 PDU SHOULD be forwarded rather than dropped. Some implementers 834 discard PDUs with currently unused combinations of headers just 835 in case they represent an attack. However, an approach using 836 alarms and policy-mediated drop is preferable to hard-coded drop, 837 so that operators can keep track of possible attacks but 838 currently unused combinations are not precluded from future use 839 through new standards actions. 841 5.5. Sequences of Similar Tunnels or Subnets 843 In some deployments, particularly in 3GPP networks, an IP packet may 844 traverse two or more IP-in-IP tunnels in sequence that all use 845 identical technology (e.g. GTP). 847 In such cases, it would be sufficient for every encapsulation and 848 decapsulation in the chain to comply with RFC 6040. Alternatively, 849 as an optimisation, a node that decapsulates a packet and immediately 850 re-encapsulates it for the next tunnel MAY copy the incoming outer 851 ECN field directly to the outgoing outer and the incoming inner ECN 852 field directly to the outgoing inner. Then the overall behavior 853 across the sequence of tunnel segments would still be consistent with 854 RFC 6040. 856 Appendix C of RFC6040 describes how a tunnel egress can monitor how 857 much congestion has been introduced within a tunnel. A network 858 operator might want to monitor how much congestion had been 859 introduced within a whole sequence of tunnels. Using the technique 860 in Appendix C of RFC6040 at the final egress, the operator could 861 monitor the whole sequence of tunnels, but only if the above 862 optimisation were used consistently along the sequence of tunnels, in 863 order to make it appear as a single tunnel. Therefore, tunnel 864 endpoint implementations SHOULD allow the operator to configure 865 whether this optimisation is enabled. 867 When ECN support is added to a subnet technology, consideration 868 SHOULD be given to a similar optimisation between subnets in sequence 869 if they all use the same technology. 871 5.6. Reframing and Congestion Markings 873 The guidance in this section is worded in terms of framing 874 boundaries, but it applies equally whether the protocol data units 875 are frames, cells or packets. 877 Where framing boundaries are different between two layers, congestion 878 indications SHOULD be propagated on the basis that a congestion 879 indication on a PDU applies to all the octets in the PDU. On 880 average, an encapsulator or decapsulator SHOULD approximately 881 preserve the number of marked octets arriving and leaving (counting 882 the size of inner headers, but not added encapsulating headers). 884 The next departing frame SHOULD be immediately marked even if only 885 enough incoming marked octets have arrived for part of the departing 886 frame. This ensures that any outstanding congestion marked octets 887 are propagated immediately, rather than held back waiting for a frame 888 no bigger than the outstanding marked octets--which might involve a 889 long wait. 891 For instance, an algorithm for marking departing frames could 892 maintain a counter representing the balance of arriving marked octets 893 minus departing marked octets. It adds the size of every marked 894 frame that arrives and if the counter is positive it marks the next 895 frame to depart and subtracts its size from the counter. This will 896 often leave a negative remainder in the counter, which is deliberate. 898 6. Feed-Up-and-Forward Mode: Guidelines for Adding Congestion 899 Notification 901 The guidance in this section is applicable when IP packets: 903 o are encapsulated in Ethernet headers; 905 o are forwarded by the eNode-B (base station) of a 3GPP radio access 906 network, which is required to apply ECN marking during congestion 907 [LTE-RA]. 909 This guidance also generalises to encapsulation by other subnet 910 technologies with no native support for explicit congestion 911 notification at the lower layer, but with support for finding and 912 processing an IP header. It is unlikely to be applicable or 913 necessary for IP-in-IP encapsulation, where feed-forward-and-up mode 914 based on [RFC6040] would be more appropriate. 916 Marking the IP header while switching at layer-2 (by using a layer-3 917 switch) or while forwarding in a radio access network seems to 918 represent a layering violation. However, it can be considered as a 919 benign optimisation if the guidelines below are followed. Feed-up- 920 and-forward is certainly not a general alternative to implementing 921 feed-forward congestion notification in the lower layer, because: 923 o IPv4 and IPv6 are not the only layer-3 protocols that might be 924 encapsulated by lower layer protocols 926 o Link-layer encryption might be in use, making the layer-2 payload 927 inaccessible 929 o Many Ethernet switches do not have 'layer-3 switch' capabilities 930 so they cannot read or modify an IP payload 932 o It might be costly to find an IP header (v4 or v6) when it may be 933 encapsulated by more than one lower layer header, e.g. Ethernet 934 MAC in MAC [IEEE802.1Qah]. 936 Nonetheless, configuring lower layer equipment to look for an ECN 937 field in an encapsulated IP header is a useful optimisation. If the 938 implementation follows the guidelines below, this optimisation does 939 not have to be confined to a controlled environment such as within a 940 data centre; it could usefully be applied on any network--even if the 941 operator is not sure whether the above issues will never apply: 943 1. If a native lower-layer congestion notification mechanism exists 944 for a subnet technology, it is safe to mix feed-up-and-forward 945 with feed-forward-and-up on other switches in the same subnet. 946 However, it will generally be more efficient to use the native 947 mechanism. 949 2. The depth of the search for an IP header SHOULD be limited. If 950 an IP header is not found soon enough, or an unrecognised or 951 unreadable header is encountered, the switch SHOULD resort to an 952 alternative means of signalling congestion (e.g. drop, or the 953 native lower layer mechanism if available). 955 3. It is sufficient to use the first IP header found in the stack; 956 the egress of the relevant tunnel can propagate congestion 957 notification upwards to any more deeply encapsulated IP headers 958 later. 960 7. Feed-Backward Mode: Guidelines for Adding Congestion Notification 962 It can be seen from Section 4.3 that congestion notification in a 963 subnet using feed-backward mode has generally not been designed to be 964 directly coupled with IP layer congestion notification. The subnet 965 attempts to minimise congestion internally, and if the incoming load 966 at the ingress exceeds the capacity somewhere through the subnet, the 967 layer 3 buffer into the ingress backs up. Thus, a feed-backward mode 968 subnet is in some sense similar to a null mode subnet, in that there 969 is no need for any direct interaction between the subnet and higher 970 layer congestion notification. Therefore no detailed protocol design 971 guidelines are appropriate. Nonetheless, a more general guideline is 972 appropriate: 974 1. A subnetwork technology intended to eventually interface to IP 975 SHOULD NOT be designed using only the feed-backward mode, which 976 is certainly best for a stand-alone subnet, but would need to be 977 modified to work efficiently as part of the wider Internet, 978 because IP uses feed-forward-and-up mode. 980 The feed-backward approach at least works beneath IP, where the term 981 'works' is used only in a narrow functional sense because feed- 982 backward can result in very inefficient and sluggish congestion 983 control--except if it is confined to the subnet directly connected to 984 the original data source, when it is faster than feed-forward. It 985 would be valid to design a protocol that could work in feed-backward 986 mode for paths that only cross one subnet, and in feed-forward-and-up 987 mode for paths that cross subnets. 989 In the early days of TCP/IP, a similar feed-backward approach was 990 tried for explicit congestion signalling, using source-quench (SQ) 991 ICMP control packets. However, SQ fell out of favour and is now 992 formally deprecated [RFC6633]. The main problem was that it is hard 993 for a data source to tell the difference between a spoofed SQ message 994 and a quench request from a genuine buffer on the path. It is also 995 hard for a lower layer buffer to address an SQ message to the 996 original source port number, which may be buried within many layers 997 of headers, and possibly encrypted. 999 Quantised congestion notification (QCN--also known as backward 1000 congestion notification or BCN) [IEEE802.1Qau] uses a feed-backward 1001 mode structurally similar to ATM's relative rate mechanism. However, 1002 QCN confines its applicability to scenarios such as some data centres 1003 where all endpoints are directly attached by the same Ethernet 1004 technology. If a QCN subnet were later connected into a wider IP- 1005 based internetwork (e.g. when attempting to interconnect multiple 1006 data centres) it would suffer the inefficiency shown Figure 3. 1008 8. IANA Considerations (to be removed by RFC Editor) 1010 This memo includes no request to IANA. 1012 9. Security Considerations 1014 If a lower layer wire protocol is redesigned to include explicit 1015 congestion signalling in-band in the protocol header, care SHOULD be 1016 take to ensure that the field used is specified as mutable during 1017 transit. Otherwise interior nodes signalling congestion would 1018 invalidate any authentication protocol applied to the lower layer 1019 header--by altering a header field that had been assumed as 1020 immutable. 1022 The redesign of protocols that encapsulate IP in order to propagate 1023 congestion signals between layers raises potential signal integrity 1024 concerns. Experimental or proposed approaches exist for assuring the 1025 end-to-end integrity of in-band congestion signals, e.g.: 1027 o Congestion exposure (ConEx ) for networks to audit that their 1028 congestion signals are not being suppressed by other networks or 1029 by receivers, and for networks to police that senders are 1030 responding sufficiently to the signals, irrespective of the 1031 transport protocol used [I-D.ietf-conex-abstract-mech]. 1033 o The ECN nonce [RFC3540] for a TCP sender to detect whether a 1034 network or the receiver is suppressing congestion signals. 1036 o A test with the same goals as the ECN nonce, but without the need 1037 for the receiver to co-operate with the protocol 1038 [I-D.moncaster-tcpm-rcv-cheat]. 1040 Given these end-to-end approaches are already being specified, it 1041 would make little sense to attempt to design hop-by-hop congestion 1042 signal integrity into a new lower layer protocol, because end-to-end 1043 integrity inherently achieves hop-by-hop integrity. 1045 10. Conclusions 1047 Following the guidance in the document enables ECN support to be 1048 extended to numerous protocols that encapsulate IP (v4 & v6) in a 1049 consistent way, so that IP continues to fulfil its role as an end-to- 1050 end interoperability layer. This includes: 1052 o A wide range of tunnelling protocols with various forms of shim 1053 header between two IP headers; 1055 o A wide range of subnet technologies, particularly those that work 1056 in the same 'feed-forward-and-up' mode that is used to support ECN 1057 in IP and MPLS. 1059 Guidelines have been defined for supporting propagation of ECN 1060 between Ethernet and IP on so-called Layer-3 Ethernet switches, using 1061 a 'feed-up-an-forward' mode. This approach could enable other subnet 1062 technologies to pass ECN signals into the IP layer, even if they do 1063 not support ECN natively. 1065 Finally, attempting to add ECN to a subnet technology in feed- 1066 backward mode is deprecated except in special cases, due to its 1067 likely sluggish response to congestion. 1069 11. Acknowledgements 1071 Thanks to Gorry Fairhurst for extensive reviews. Thanks also to the 1072 following reviewers: Ingemar Johansson and Piers O'Hanlon and Michael 1073 Welzl, who pointed out that lower layer congestion notification 1074 signals may have different semantics to those in IP. 1076 Bob Briscoe was part-funded by the European Community under its 1077 Seventh Framework Programme through the Trilogy project (ICT-216372) 1078 for initial drafts and through the Reducing Internet Transport 1079 Latency (RITE) project (ICT-317700) subsequently. The views 1080 expressed here are solely those of the authors. 1082 12. Comments Solicited 1084 Comments and questions are encouraged and very welcome. They can be 1085 addressed to the IETF Transport Area working group mailing list 1086 , and/or to the authors. 1088 13. References 1090 13.1. Normative References 1092 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 1093 Requirement Levels", BCP 14, RFC 2119, 1094 DOI 10.17487/RFC2119, March 1997, 1095 . 1097 [RFC3168] Ramakrishnan, K., Floyd, S., and D. Black, "The Addition 1098 of Explicit Congestion Notification (ECN) to IP", 1099 RFC 3168, DOI 10.17487/RFC3168, September 2001, 1100 . 1102 [RFC3819] Karn, P., Ed., Bormann, C., Fairhurst, G., Grossman, D., 1103 Ludwig, R., Mahdavi, J., Montenegro, G., Touch, J., and L. 1104 Wood, "Advice for Internet Subnetwork Designers", BCP 89, 1105 RFC 3819, DOI 10.17487/RFC3819, July 2004, 1106 . 1108 [RFC4774] Floyd, S., "Specifying Alternate Semantics for the 1109 Explicit Congestion Notification (ECN) Field", BCP 124, 1110 RFC 4774, DOI 10.17487/RFC4774, November 2006, 1111 . 1113 [RFC5129] Davie, B., Briscoe, B., and J. Tay, "Explicit Congestion 1114 Marking in MPLS", RFC 5129, DOI 10.17487/RFC5129, January 1115 2008, . 1117 [RFC6040] Briscoe, B., "Tunnelling of Explicit Congestion 1118 Notification", RFC 6040, DOI 10.17487/RFC6040, November 1119 2010, . 1121 13.2. Informative References 1123 [ATM-TM-ABR] 1124 Cisco, "Understanding the Available Bit Rate (ABR) Service 1125 Category for ATM VCs", Design Technote 10415, June 2005. 1127 [Buck00] Buckwalter, J., "Frame Relay: Technology and Practice", 1128 Pub. Addison Wesley ISBN-13: 978-0201485240, 2000. 1130 [DCTCP] Alizadeh, M., Greenberg, A., Maltz, D., Padhye, J., Patel, 1131 P., Prabhakar, B., Sengupta, S., and M. Sridharan, "Data 1132 Center TCP (DCTCP)", ACM SIGCOMM CCR 40(4)63--74, October 1133 2010, . 1135 [GTPv1] 3GPP, "GPRS Tunnelling Protocol (GTP) across the Gn and Gp 1136 interface", Technical Specification TS 29.060. 1138 [GTPv1-U] 3GPP, "General Packet Radio System (GPRS) Tunnelling 1139 Protocol User Plane (GTPv1-U)", Technical Specification TS 1140 29.281. 1142 [GTPv2-C] 3GPP, "Evolved General Packet Radio Service (GPRS) 1143 Tunnelling Protocol for Control plane (GTPv2-C)", 1144 Technical Specification TS 29.274. 1146 [I-D.ietf-aqm-recommendation] 1147 Baker, F. and G. Fairhurst, "IETF Recommendations 1148 Regarding Active Queue Management", draft-ietf-aqm- 1149 recommendation-11 (work in progress), February 2015. 1151 [I-D.ietf-conex-abstract-mech] 1152 Mathis, M. and B. Briscoe, "Congestion Exposure (ConEx) 1153 Concepts, Abstract Mechanism and Requirements", draft- 1154 ietf-conex-abstract-mech-13 (work in progress), October 1155 2014. 1157 [I-D.ietf-trill-rfc7180bis] 1158 Eastlake, D., Zhang, M., Perlman, R., Banerjee, A., 1159 Ghanwani, A., and S. Gupta, "TRILL: Clarifications, 1160 Corrections, and Updates", draft-ietf-trill-rfc7180bis-05 1161 (work in progress), June 2015. 1163 [I-D.moncaster-tcpm-rcv-cheat] 1164 Moncaster, T., Briscoe, B., and A. Jacquet, "A TCP Test to 1165 Allow Senders to Identify Receiver Non-Compliance", draft- 1166 moncaster-tcpm-rcv-cheat-03 (work in progress), July 2014. 1168 [IEEE802.1Qah] 1169 IEEE, "IEEE Standard for Local and Metropolitan Area 1170 Networks--Virtual Bridged Local Area Networks--Amendment 1171 6: Provider Backbone Bridges", IEEE Std 802.1Qah-2008, 1172 August 2008, 1173 . 1175 (Access Controlled link within page) 1177 [IEEE802.1Qau] 1178 Finn, N., Ed., "IEEE Standard for Local and Metropolitan 1179 Area Networks--Virtual Bridged Local Area Networks - 1180 Amendment 13: Congestion Notification", IEEE Std 802.1Qau- 1181 2010, March 2010, . 1184 (Access Controlled link within page) 1186 [ITU-T.I.371] 1187 ITU-T, "Traffic Control and Congestion Control in B-ISDN", 1188 ITU-T Rec. I.371 (03/04), March 2004, 1189 . 1192 [LTE-RA] 3GPP, "Evolved Universal Terrestrial Radio Access (E-UTRA) 1193 and Evolved Universal Terrestrial Radio Access Network 1194 (E-UTRAN); Overall description; Stage 2", Technical 1195 Specification TS 36.300. 1197 [RFC1323] Jacobson, V., Braden, R., and D. Borman, "TCP Extensions 1198 for High Performance", RFC 1323, DOI 10.17487/RFC1323, May 1199 1992, . 1201 [RFC1701] Hanks, S., Li, T., Farinacci, D., and P. Traina, "Generic 1202 Routing Encapsulation (GRE)", RFC 1701, 1203 DOI 10.17487/RFC1701, October 1994, 1204 . 1206 [RFC2003] Perkins, C., "IP Encapsulation within IP", RFC 2003, 1207 DOI 10.17487/RFC2003, October 1996, 1208 . 1210 [RFC2637] Hamzeh, K., Pall, G., Verthein, W., Taarud, J., Little, 1211 W., and G. Zorn, "Point-to-Point Tunneling Protocol 1212 (PPTP)", RFC 2637, DOI 10.17487/RFC2637, July 1999, 1213 . 1215 [RFC2661] Townsley, W., Valencia, A., Rubens, A., Pall, G., Zorn, 1216 G., and B. Palter, "Layer Two Tunneling Protocol "L2TP"", 1217 RFC 2661, DOI 10.17487/RFC2661, August 1999, 1218 . 1220 [RFC2784] Farinacci, D., Li, T., Hanks, S., Meyer, D., and P. 1221 Traina, "Generic Routing Encapsulation (GRE)", RFC 2784, 1222 DOI 10.17487/RFC2784, March 2000, 1223 . 1225 [RFC2884] Hadi Salim, J. and U. Ahmed, "Performance Evaluation of 1226 Explicit Congestion Notification (ECN) in IP Networks", 1227 RFC 2884, DOI 10.17487/RFC2884, July 2000, 1228 . 1230 [RFC2983] Black, D., "Differentiated Services and Tunnels", 1231 RFC 2983, DOI 10.17487/RFC2983, October 2000, 1232 . 1234 [RFC3540] Spring, N., Wetherall, D., and D. Ely, "Robust Explicit 1235 Congestion Notification (ECN) Signaling with Nonces", 1236 RFC 3540, DOI 10.17487/RFC3540, June 2003, 1237 . 1239 [RFC4301] Kent, S. and K. Seo, "Security Architecture for the 1240 Internet Protocol", RFC 4301, DOI 10.17487/RFC4301, 1241 December 2005, . 1243 [RFC6633] Gont, F., "Deprecation of ICMP Source Quench Messages", 1244 RFC 6633, DOI 10.17487/RFC6633, May 2012, 1245 . 1247 [RFC6660] Briscoe, B., Moncaster, T., and M. Menth, "Encoding Three 1248 Pre-Congestion Notification (PCN) States in the IP Header 1249 Using a Single Diffserv Codepoint (DSCP)", RFC 6660, 1250 DOI 10.17487/RFC6660, July 2012, 1251 . 1253 [RFC7348] Mahalingam, M., Dutt, D., Duda, K., Agarwal, P., Kreeger, 1254 L., Sridhar, T., Bursell, M., and C. Wright, "Virtual 1255 eXtensible Local Area Network (VXLAN): A Framework for 1256 Overlaying Virtualized Layer 2 Networks over Layer 3 1257 Networks", RFC 7348, DOI 10.17487/RFC7348, August 2014, 1258 . 1260 Appendix A. Outstanding Document Issues 1262 1. [GF] Concern that certain guidelines warrant a MUST (NOT) rather 1263 than a SHOULD (NOT). Given the guidelines say that if any SHOULD 1264 (NOT)s are not followed, a strong justification will be needed, 1265 they have been left as SHOULD (NOT) pending further list 1266 discussion. In particular: 1268 * If inner is a Not-ECN-PDU and Outer is CE (or highest severity 1269 congestion level), MUST (not SHOULD) drop? 1271 2. Consider whether an IETF Standard Track doc will be needed to 1272 Update the IP-in-IP protocols listed in Section 5.1--at least 1273 those that the IET 1275 Appendix B. Changes in This Version (to be removed by RFC Editor) 1277 From ietf-02 to ietf-03: 1279 * Updated references, ad cited RFC4774. 1281 From ietf-01 to ietf-02: 1283 * Added Section for guidelines that are applicable in all cases. 1285 * Updated references. 1287 From ietf-00 to ietf-01: Updated references. 1289 From briscoe-04 to ietf-00: Changed filename following tsvwg 1290 adoption. 1292 From briscoe-03 to 04: 1294 * Re-arranged the introduction to describe the purpose of the 1295 document first before introducing ECN in more depth. And 1296 clarified the introduction throughout. 1298 * Added applicability to 3GPP TS 36.300. 1300 From briscoe-02 to 03: 1302 * Scope section: 1304 + Added dependence on correct propagation of traffic class 1305 information 1307 + For the feed-backward mode, deemed multicast and anycast out 1308 of scope 1310 * Ensured all guidelines referring to subnet technologies also 1311 refer to tunnels and vice versa by adding applicability 1312 sentences at the start of sections 4.1, 4.2, 4.3, 4.4, 4.6 and 1313 5. 1315 * Added Security Considerations on ensuring congestion signal 1316 fields are classed as immutable and on using end-to-end 1317 congestion signal integrity technologies rather than hop-by- 1318 hop. 1320 From briscoe-01 to 02: 1322 * Added authors: JK & PT 1324 * Added 1326 + Section 4.1 "IP-in-IP Tunnels with Tightly Coupled Shim 1327 Headers" 1329 + Section 4.5 "Sequences of Similar Tunnels or Subnets" 1331 + roadmap at the start of Section 4, given the subsections 1332 have become quite fragmented. 1334 + Section 9 "Conclusions" 1336 * Clarified why transports are starting to be able to saturate 1337 interior links 1339 * Under Section 1.1, addressed the question of alternative signal 1340 semantics and included multicast & anycast. 1342 * Under Section 3.1, included a 3GPP example. 1344 * Section 4.2. "Wire Protocol Design": 1346 + Altered guideline 2. to make it clear that it only applies 1347 to the immediate subnet egress, not later ones 1349 + Added a reminder that it is only necessary to check that ECN 1350 propagates at the egress, not whether interior nodes mark 1351 ECN 1353 + Added example of how QCN uses 802.1p to indicate support for 1354 QCN. 1356 * Added references to Appendix C of RFC6040, about monitoring the 1357 amount of congestion signals introduced within a tunnel 1359 * Appendix A: Added more issues to be addressed, including plan 1360 to produce a standards track update to IP-in-IP tunnel 1361 protocols. 1363 * Updated acks and references 1365 From briscoe-00 to 01: 1367 * Intended status: BCP (was Informational) & updates 3819 added. 1369 * Briefer Introduction: Introductory para justifying benefits of 1370 ECN. Moved all but a brief enumeration of modes of operation 1371 to their own new section (from both Intro & Scope). Introduced 1372 incr. deployment as most tricky part. 1374 * Tightened & added to terminology section 1376 * Structured with Modes of Operation, then Guidelines section for 1377 each mode. 1379 * Tightened up guideline text to remove vagueness / passive voice 1380 / ambiguity and highlight main guidelines as numbered items. 1382 * Added Outstanding Document Issues Appendix 1384 * Updated references 1386 Authors' Addresses 1388 Bob Briscoe 1389 Simula Research Laboratory 1390 UK 1392 EMail: ietf@bobbriscoe.net 1393 URI: http://bobbriscoe.net/ 1395 John Kaippallimalil 1396 Huawei 1397 5340 Legacy Drive, Suite 175 1398 Plano, Texas 75024 1399 USA 1401 EMail: john.kaippallimalil@huawei.com 1402 Pat Thaler 1403 Broadcom Corporation 1404 5025 Keane Drive 1405 Carmichael, CA 95608 1406 USA 1408 EMail: pthaler@broadcom.com