idnits 2.17.1 draft-ietf-tcpm-ecnsyn-04.txt: Checking boilerplate required by RFC 5378 and the IETF Trust (see https://trustee.ietf.org/license-info): ---------------------------------------------------------------------------- ** It looks like you're using RFC 3978 boilerplate. You should update this to the boilerplate described in the IETF Trust License Policy document (see https://trustee.ietf.org/license-info), which is required now. -- Found old boilerplate from RFC 3978, Section 5.1 on line 19. -- Found old boilerplate from RFC 3978, Section 5.5, updated by RFC 4748 on line 1126. -- Found old boilerplate from RFC 3979, Section 5, paragraph 1 on line 1137. -- Found old boilerplate from RFC 3979, Section 5, paragraph 2 on line 1144. -- Found old boilerplate from RFC 3979, Section 5, paragraph 3 on line 1150. Checking nits according to https://www.ietf.org/id-info/1id-guidelines.txt: ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/checklist : ---------------------------------------------------------------------------- No issues found here. Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust Copyright Line does not match the current year -- The document seems to lack a disclaimer for pre-RFC5378 work, but may have content which was first submitted before 10 November 2008. If you have contacted all the original authors and they are all willing to grant the BCP78 rights to the IETF Trust, then this is fine, and you can ignore this comment. If not, you may need to add the pre-RFC5378 disclaimer. (See the Legal Provisions document at https://trustee.ietf.org/license-info for more information.) -- The document date (8 January 2008) is 5952 days in the past. Is this intentional? Checking references for intended status: Proposed Standard ---------------------------------------------------------------------------- (See RFCs 3967 and 4897 for information about using normative references to lower-maturity documents in RFCs) -- Obsolete informational reference (is this intentional?): RFC 2309 (Obsoleted by RFC 7567) -- Obsolete informational reference (is this intentional?): RFC 2581 (Obsoleted by RFC 5681) -- Obsolete informational reference (is this intentional?): RFC 2988 (Obsoleted by RFC 6298) == Outdated reference: A later version (-05) exists of draft-irtf-tmrg-tools-04 Summary: 1 error (**), 0 flaws (~~), 2 warnings (==), 10 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 1 Internet Engineering Task Force A. Kuzmanovic 2 INTERNET-DRAFT A. Mondal 3 Intended status: Proposed Standard Northwestern University 4 Expires: 8 July 2008 S. Floyd 5 ICIR 6 K.K. Ramakrishnan 7 AT&T 8 8 January 2008 10 Adding Explicit Congestion Notification (ECN) Capability 11 to TCP's SYN/ACK Packets 12 draft-ietf-tcpm-ecnsyn-04.txt 14 Status of this Memo 16 By submitting this Internet-Draft, each author represents that any 17 applicable patent or other IPR claims of which he or she is aware 18 have been or will be disclosed, and any of which he or she becomes 19 aware will be disclosed, in accordance with Section 6 of BCP 79. 21 Internet-Drafts are working documents of the Internet Engineering 22 Task Force (IETF), its areas, and its working groups. Note that 23 other groups may also distribute working documents as Internet- 24 Drafts. 26 Internet-Drafts are draft documents valid for a maximum of six months 27 and may be updated, replaced, or obsoleted by other documents at any 28 time. It is inappropriate to use Internet-Drafts as reference 29 material or to cite them other than as "work in progress." 31 The list of current Internet-Drafts can be accessed at 32 http://www.ietf.org/ietf/1id-abstracts.txt. 34 The list of Internet-Draft Shadow Directories can be accessed at 35 http://www.ietf.org/shadow.html. 37 This Internet-Draft will expire on December 2007. 39 Copyright Notice 41 Copyright (C) The IETF Trust (2007). 43 Abstract 45 This draft specifies a modification to RFC 3168 to allow TCP SYN/ACK 46 packets to be ECN-Capable. For TCP, RFC 3168 only specifies setting 47 an ECN-Capable codepoint on data packets, and not on SYN and SYN/ACK 48 packets. However, because of the high cost to the TCP transfer of 49 having a SYN/ACK packet dropped, with the resulting retransmit 50 timeout, this document specifies the use of ECN for the SYN/ACK 51 packet itself, when sent in response to a SYN packet with the two ECN 52 flags set in the TCP header, indicating a willingness to use ECN. 53 Setting TCP SYN/ACK packets as ECN-Capable can be of great benefit to 54 the TCP connection, avoiding the severe penalty of a retransmit 55 timeout for a connection that has not yet started placing a load on 56 the network. The sender of the SYN/ACK packet must respond to a 57 report of an ECN-marked SYN/ACK packet by reducing its initial 58 congestion window from two, three, or four segments to one segment, 59 thereby reducing the subsequent load from that connection on the 60 network. This document is intended to update RFC 3168. 62 Table of Contents 64 1. Introduction ....................................................4 65 2. Conventions and Terminology .....................................6 66 3. Proposal ........................................................6 67 4. Discussion ......................................................9 68 5. Related Work ...................................................12 69 6. Performance Evaluation .........................................13 70 6.1. The Costs and Benefit of Adding ECN-Capability ............13 71 6.2. An Evaluation of Different Responses to ECN-Marked SYN/ACK 72 Packets ........................................................14 73 7. Security Considerations ........................................15 74 8. Conclusions ....................................................16 75 9. Acknowledgements ...............................................17 76 A. Report on Simulations ..........................................17 77 A.1. Simulations with RED in Packet Mode .......................17 78 A.2. Simulations with RED in Byte Mode .........................19 79 B. Issues of Incremental Deployment ...............................20 80 Normative References ..............................................23 81 Informative References ............................................23 82 IANA Considerations ...............................................24 83 Full Copyright Statement ..........................................25 84 Intellectual Property .............................................25 86 NOTE TO RFC EDITOR: PLEASE DELETE THIS NOTE UPON PUBLICATION. 88 Changes from draft-ietf-tcpm-ecnsyn-03: 90 * General editing. This includes using the terms "initiator" 91 and "responder" for the two ends of the TCP connection. 92 Feedback from Alfred Hoenes. 94 * Added some text to the backwards compatibility discussion, 95 now in Appendix B, about the pros and cons of using a TCP 96 flag for the TCP initiator to signal that it understands 97 ECN-Capable SYN/ACK packets. The consensus at this time is 98 not to use such a flag. Also added a recommendation that 99 TCP implementations include a management interface to turn 100 off the use of ECN for SYN/ACK packets. From email from 101 Bob Briscoe. 103 Changes from draft-ietf-tcpm-ecnsyn-02: 105 * Added to the discussion in the Security section of whether 106 ECN-Capable TCP SYN packets have problems with firewalls, 107 over and above the known problems of TCP data packets 108 (e.g., as in the Microsoft report). From a question raised 109 at the TCPM meeting at the July 2007 IETF. 111 * Added a sentence to the discussion of routers or middleboxes that 112 *might* drop TCP SYN packets on the basis of IP header fields. 113 Feedback from Remi Denis-Courmont. 115 * General editing. Feedback from Alfred Hoenes. 117 Changes from draft-ietf-tcpm-ecnsyn-01: 119 * Changes in response to feedback from Anil Agarwal. 121 * Added a look at the costs of adding ECN-Capability to 122 SYN/ACKs in a highly-congested scenario. 123 From feedback from Mark Allman and Janardhan Iyengar. 125 * Added a comparative evaluation of two possible responses 126 to an ECN-marked SYN/ACK packet. From Mark Allman. 128 Changes from draft-ietf-tcpm-ecnsyn-00: 130 * Only updating the revision number. 132 Changes from draft-ietf-twvsg-ecnsyn-00: 134 * Changed name of draft to draft-ietf-tcpm-ecnsyn. 136 * Added a discussion in Section 3 of "Response to 137 ECN-marking of SYN/ACK packets". Based on 138 suggestions from Mark Allman. 140 * Added a discussion to the Conclusions about adding 141 ECN-capability to relevant set-up packets in other 142 protocols. From a suggestion from Wesley Eddy. 144 * Added a description of SYN exchanges with SYN cookies. 145 From a suggestion from Wesley Eddy. 147 * Added a discussion of one-way data transfers, where the 148 host sending the SYN/ACK packet sends no data packets. 150 * Minor editing, from feedback from Mark Allman and Janardhan 151 Iyengar. 153 * Future work: a look at the costs of adding 154 ECN-Capability in a worst-case scenario. 155 From feedback from Mark Allman and Janardhan Iyengar. 157 * Future work: a comparative evaluation of two 158 possible responses to an ECN-marked SYN/ACK packet. 160 Changes from draft-kuzmanovic-ecn-syn-00.txt: 162 * Changed name of draft to draft-ietf-twvsg-ecnsyn. 164 END OF NOTE TO RFC EDITOR. 166 1. Introduction 168 TCP's congestion control mechanism has primarily used packet loss as 169 the congestion indication, with packets dropped when buffers 170 overflow. With such tail-drop mechanisms, the packet delay can be 171 high, as the queue at bottleneck routers can be fairly large. 172 Dropping packets only when the queue overflows, and having TCP react 173 only to such losses, results in: 174 1) significantly higher packet delay; 175 2) unnecessarily many packet losses; and 176 3) unfairness due to synchronization effects. 178 The adoption of Active Queue Management (AQM) mechanisms allows 179 better control of bottleneck queues [RFC2309]. This use of AQM has 180 the following potential benefits: 181 1) better control of the queue, with reduced queueing delay; 182 2) fewer packet drops; and 183 3) better fairness because of fewer synchronization effects. 185 With the adoption of ECN, performance may be further improved. When 186 the router detects congestion before buffer overflow, the router can 187 provide a congestion indication either by dropping a packet, or by 188 setting the Congestion Experienced (CE) codepoint in the Explicit 189 Congestion Notification (ECN) field in the IP header [RFC3168]. The 190 IETF has standardized the use of the Congestion Experienced (CE) 191 codepoint in the IP header for routers to indicate congestion. For 192 incremental deployment and backwards compatibility, the RFC on ECN 193 [RFC3168] specifies that routers may mark ECN-capable packets that 194 would otherwise have been dropped, using the Congestion Experienced 195 codepoint in the ECN field. The use of ECN allows TCP to react to 196 congestion while avoiding unnecessary retransmissions and, in some 197 cases, unnecessary retransmit timeouts. Thus, using ECN has several 198 benefits: 200 1) For short transfers, a TCP connection's congestion window may be 201 small. For example, if the current window contains only one packet, 202 and that packet is dropped, TCP will have to wait for a retransmit 203 timeout to recover, reducing its overall throughput. Similarly, if 204 the current window contains only a few packets and one of those 205 packets is dropped, there might not be enough duplicate 206 acknowledgements for a fast retransmission, and the sender of the 207 data packet might have to wait for a delay of several round-trip 208 times using Limited Transmit [RFC3042]. With the use of ECN, short 209 flows are less likely to have packets dropped, sometimes avoiding 210 unnecessary delays or costly retransmit timeouts. 212 2) While longer flows may not see substantially improved throughput 213 with the use of ECN, they experience lower loss. This may benefit TCP 214 applications that are latency- and loss-sensitive, because of the 215 avoidance of retransmissions. 217 RFC 3168 only specifies marking the Congestion Experienced codepoint 218 on TCP's data packets, and not on SYN and SYN/ACK packets. RFC 3168 219 specifies the negotiation of the use of ECN between the two TCP end- 220 points in the TCP SYN and SYN-ACK exchange, using flags in the TCP 221 header. Erring on the side of being conservative, RFC 3168 does not 222 specify the use of ECN for the SYN/ACK packet itself. However, 223 because of the high cost to the TCP transfer of having a SYN/ACK 224 packet dropped, with the resulting retransmit timeout, this document 225 specifies the use of ECN for the SYN/ACK packet itself. This can be 226 of great benefit to the TCP connection, avoiding the severe penalty 227 of a retransmit timeout for a connection that has not yet started 228 placing a load on the network. The sender of the SYN/ACK packet must 229 respond to a report of an ECN-marked SYN/ACK packet by reducing its 230 initial congestion window from two, three, or four segments to one 231 segment, reducing the subsequent load from that connection on the 232 network. 234 The use of ECN for SYN/ACK packets has the following potential 235 benefits: 236 1) Avoidance of a retransmit timeout; 237 2) Improvement in the throughput of short connections. 239 This draft specifies ECN+, a modification to RFC 3168 to allow TCP 240 SYN/ACK packets to be ECN-Capable. Section 3 contains the 241 specification of the change, while Section 4 discusses some of the 242 issues, and Section 5 discusses related work. Section 6 contains an 243 evaluation of the proposed change. 245 2. Conventions and Terminology 247 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 248 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 249 document are to be interpreted as described in [RFC 2119]. 251 We use the following terminology from RFC 3168: 253 The ECN field in the IP header: 254 o CE: the Congestion Experienced codepoint; and 255 o ECT: either one of the two ECN-Capable Transport codepoints. 257 The ECN flags in the TCP header: 258 o CWR: the Congestion Window Reduced flag; and 259 o ECE: the ECN-Echo flag. 261 ECN-setup packets: 262 o ECN-setup SYN packet: a SYN packet with the ECE and CWR flags; 263 o ECN-setup SYN-ACK packet: a SYN-ACK packet with ECE but not CWR. 265 In this document we use the terms "initiator" and "responder" to 266 refer to the sender of the SYN packet and of the SYN-ACK packet, 267 respectively. 269 3. Proposal 271 This section specifies the modification to RFC 3168 to allow TCP 272 SYN/ACK packets to be ECN-Capable. 274 RFC 3168 in Section 6.1.1. states that "A host MUST NOT set ECT on 275 SYN or SYN-ACK packets." In this section, we specify that a TCP node 276 MAY respond to an ECN-setup SYN packet by setting ECT in the 277 responding ECN-setup SYN/ACK packet, indicating to routers that the 278 SYN/ACK packet is ECN-Capable. This allows a congested router along 279 the path to mark the packet instead of dropping the packet as an 280 indication of congestion. 282 Assume that TCP node A transmits to TCP node B an ECN-setup SYN 283 packet, indicating willingness to use ECN for this connection. As 284 specified by RFC 3168, if TCP node B is willing to use ECN, node B 285 responds with an ECN-setup SYN-ACK packet. 287 Figure 1 shows an interchange with the SYN/ACK packet dropped by a 288 congested router. Node B waits for a retransmit timeout, and then 289 retransmits the SYN/ACK packet. 291 --------------------------------------------------------------- 292 TCP Node A Router TCP Node B 293 ---------- ------ ---------- 295 ECN-setup SYN packet ---> 296 ECN-setup SYN packet ---> 298 <--- ECN-setup SYN/ACK, possibly ECT 299 3-second timer set 300 SYN/ACK dropped . 301 . 302 . 303 3-second timer expires 304 <--- ECN-setup SYN/ACK, not ECT 305 <--- ECN-setup SYN/ACK 306 Data/ACK ---> 307 Data/ACK ---> 308 <--- Data (one to four segments) 309 --------------------------------------------------------------- 311 Figure 1: SYN exchange with the SYN/ACK packet dropped. 313 If the SYN/ACK packet is dropped in the network, the responder (node 314 B) responds by waiting three seconds for the retransmit timer to 315 expire [RFC2988]. If a SYN/ACK packet with the ECT codepoint is 316 dropped, the responder SHOULD resend the SYN/ACK packet without the 317 ECN-Capable codepoint. (Although we are not aware of any middleboxes 318 that drop SYN/ACK packets that contain an ECN-Capable codepoint in 319 the IP header, we have learned to design our protocols defensively in 320 this regard [RFC3360].) 322 We note that if syn-cookies were used by the responder (node B) in 323 the exchange in Figure 1, the responder wouldn't set a timer upon 324 transmission of the SYN/ACK packet [SYN-COOK]. In this case, if the 325 SYN/ACK packet was lost, the initiator (Node A) would have to timeout 326 and retransmit the SYN packet in order to trigger another SYN-ACK. 328 Figure 2 shows an interchange with the SYN/ACK packet sent as ECN- 329 Capable, and ECN-marked instead of dropped at the congested router. 331 --------------------------------------------------------------- 332 TCP Node A Router TCP Node B 333 ---------- ------ ---------- 335 ECN-setup SYN packet ---> 336 ECN-setup SYN packet ---> 338 <--- ECN-setup SYN/ACK, ECT 339 <--- Sets CE on SYN/ACK 340 <--- ECN-setup SYN/ACK, CE 342 Data/ACK, ECN-Echo ---> 343 Data/ACK, ECN-Echo ---> 344 Window reduced to one segment. 345 <--- Data, CWR (one segment only) 346 --------------------------------------------------------------- 348 Figure 2: SYN exchange with the SYN/ACK packet marked. 350 If the initiator (node A) receives a SYN/ACK packet that has been 351 marked by the congested router, with the CE codepoint set, the 352 initiator MUST respond by setting the ECN-Echo flag in the TCP header 353 of the responding ACK packet. As specified in RFC 3168, the 354 initiator continues to set the ECN-Echo flag in packets until it 355 receives a packet with the CWR flag set. 357 When the responder (node B) receives the ECN-Echo packet reporting 358 the Congestion Experienced indication in the SYN/ACK packet, the 359 responder MUST set the initial congestion window to one segment, 360 instead of two segments as allowed by [RFC2581], or three or four 361 segments allowed by [RFC3390]. If the responder (node B) was going 362 to use an initial window of one segment, and receives an ECN-Echo 363 packet informing it of a Congestion Experienced indication on its 364 SYN/ACK packet, the responder MAY continue to send with an initial 365 window of one segment, without waiting for a retransmit timeout. We 366 note that this updates RFC 3168, which specifies that "the sending 367 TCP MUST reset the retransmit timer on receiving the ECN-Echo packet 368 when the congestion window is one." As specified by RFC 3168, the 369 responder (node B) also sets the CWR flag in the TCP header of the 370 next data packet sent, to acknowledge its receipt of and reaction to 371 the ECN-Echo flag. 373 If the data transfer in Figure 2 is entirely from Node A to Node B, 374 then data packets from Node A continue to set the ECN-Echo flag in 375 data packets, waiting for the CWR flag from Node B acknowledging a 376 response to the ECN-Echo flag. 378 The TCP implementation using ECN-Capable SYN/ACK packets SHOULD 379 include a management interface to allow the use of ECN to be turned 380 off for SYN/ACK packets. This is to deal with possible backwards 381 compatibility problems such as those discussed in Appendix B. 383 4. Discussion 385 Motivation: 386 The rationale for the proposed change is the following. When node B 387 receives a TCP SYN packet with ECN-Echo bit set in the TCP header, 388 this indicates that node A is ECN-capable. If node B is also ECN- 389 capable, there are no obstacles to immediately setting one of the 390 ECN-Capable codepoints in the IP header in the responding TCP SYN/ACK 391 packet. 393 There can be a great benefit in setting an ECN-capable codepoint in 394 SYN/ACK packets, as is discussed further in [ECN+], and reported 395 briefly in Section 5 below. Congestion is most likely to occur in 396 the server-to-client direction. As a result, setting an ECN-capable 397 codepoint in SYN/ACK packets can reduce the occurrence of three- 398 second retransmit timeouts resulting from the drop of SYN/ACK 399 packets. 401 Flooding attacks: 402 Setting an ECN-Capable codepoint in the responding TCP SYN/ACK 403 packets does not raise any novel security vulnerabilities. For 404 example, provoking servers or hosts to send SYN/ACK packets to a 405 third party in order to perform a "SYN/ACK flood" attack would be 406 highly inefficient. Third parties would immediately drop such 407 packets, since they would know that they didn't generate the TCP SYN 408 packets in the first place. Moreover, such SYN/ACK attacks would 409 have the same signatures as the existing TCP SYN attacks. Provoking 410 servers or hosts to reply with SYN/ACK packets in order to congest a 411 certain link would also be highly inefficient because SYN/ACK packets 412 are small in size. 414 However, the addition of ECN-Capability to SYN/ACK packets could 415 allow SYN/ACK packets to persist for more hops along a network path 416 before being dropped, thus adding somewhat to the ability of a 417 SYN/ACK attack to flood a network link. 419 The TCP SYN packet: 420 There are several reasons why an ECN-Capable codepoint MUST NOT be 421 set in the IP header of the initiating TCP SYN packet. First, when 422 the TCP SYN packet is sent, there are no guarantees that the other 423 TCP endpoint (node B in Figure 2) is ECN-capable, or that it would be 424 able to understand and react if the ECN CE codepoint was set by a 425 congested router. 427 Second, the ECN-Capable codepoint in TCP SYN packets could be misused 428 by malicious clients to `improve' the well-known TCP SYN attack. By 429 setting an ECN-Capable codepoint in TCP SYN packets, a malicious host 430 might be able to inject a large number of TCP SYN packets through a 431 potentially congested ECN-enabled router, congesting it even further. 433 For both these reasons, we continue the restriction that the TCP SYN 434 packet MUST NOT have the ECN-Capable codepoint in the IP header set. 436 SYN/ACK packets and packet size: 437 There are a number of router buffer architectures that have smaller 438 dropping rates for small (SYN) packets than for large (data) packets. 439 For example, for a Drop Tail queue in units of packets, where each 440 packet takes a single slot in the buffer regardless of packet size, 441 small and large packets are equally likely to be dropped. However, 442 for a Drop Tail queue in units of bytes, small packets are less 443 likely to be dropped than are large ones. Similarly, for RED in 444 packet mode, small and large packets are equally likely to be dropped 445 or marked, while for RED in byte mode, a packet's chance of being 446 dropped or marked is proportional to the packet size in bytes. 448 For a congested router with an AQM mechanism in byte mode, where a 449 packet's chance of being dropped or marked is proportional to the 450 packet size in bytes, the drop or marking rate for TCP SYN/ACK 451 packets should generally be low. In this case, the benefit of making 452 SYN/ACK packets ECN-Capable should be similarly moderate. However, 453 for a congested router with a Drop Tail queue in units of packets or 454 with an AQM mechanism in packet mode, and with no priority queueing 455 for smaller packets, small and large packets should have the same 456 probability of being dropped or marked. In such a case, making 457 SYN/ACK packets ECN-Capable should be of significant benefit. 459 We believe that there are a wide range of behaviors in the real world 460 in terms of the drop or mark behavior at routers as a function of 461 packet size [Tools] (Section 10). We note that all of these 462 alternatives listed above are available in the NS simulator (Drop 463 Tail queues are by default in units of packets, while the default for 464 RED queue management has been changed from packet mode to byte mode). 466 Response to ECN-marking of SYN/ACK packets: 467 One question is why TCP SYN/ACK packets should be treated differently 468 from other packets in terms of the end node's response to an ECN- 469 marked packet. Section 5 of RFC 3168 specifies the following: 471 "Upon the receipt by an ECN-Capable transport of a single CE packet, 472 the congestion control algorithms followed at the end-systems MUST be 473 essentially the same as the congestion control response to a *single* 474 dropped packet. For example, for ECN-Capable TCP the source TCP is 475 required to halve its congestion window for any window of data 476 containing either a packet drop or an ECN indication." 478 In particular, Section 6.1.2 of RFC 3168 specifies that when the TCP 479 congestion window consists of a single packet and that packet is ECN- 480 marked in the network, then the data sender must reduce the sending 481 rate below one packet per round-trip time, by waiting for one RTO 482 before sending another packet. If the RTO was set to the average 483 round-trip time, this would result in halving the sending rate; 484 because the RTO is in fact larger than the average round-trip time, 485 the sending rate is reduced to less than half of its previous value. 487 TCP's congestion control response to the *dropping* of a SYN/ACK 488 packet is to wait a default time before sending another packet. This 489 document argues that ECN gives end-systems a wider range of possible 490 responses to the *marking* of a SYN/ACK packet, and that waiting a 491 default time before sending a data packet is not the desired 492 response. 494 On the conservative end, one could assume an effective congestion 495 window of one packet for the SYN/ACK packet, and respond to an ECN- 496 marked SYN/ACK packet by reducing the sending rate to one packet 497 every two round-trip times. As an approximation, the TCP end-node 498 could measure the round-trip time T between the sending of the 499 SYN/ACK packet and the receipt of the acknowledgement, and reply to 500 the acknowledgement of the ECN-marked SYN/ACK packet by waiting T 501 seconds before sending a data packet. 503 However, we note that for an ECN-marked SYN/ACK packet, halving the 504 *congestion window* is not the same as halving the *sending rate*; 505 there is no `sending rate' associated with an ECN-Capable SYN/ACK 506 packet, as such packets are only sent as the first packet in a 507 connection from that host. Further, a router's marking of a SYN/ACK 508 packet is not affected by any past history of that connection. 510 Adding ECN-Capability to SYN/ACK packets allows the simple response 511 of the responder setting the initial congestion window to one packet, 512 instead of its allowed default value of two, three, or four packets, 513 with the responder proceeding with a cautious sending rate of one 514 packet per round-trip time. If that data packet is ECN-marked or 515 dropped, then the responder will wait an RTO before sending another 516 packet. This document argues that this approach is useful to users, 517 with no dangers of congestion collapse or of starvation of competing 518 traffic. This is discussed in more detail below in Section 6.2. 520 We note that if the data transfer is entirely from Node A to Node B, 521 then there is no effective difference between the two possible 522 responses to an ECN-marked SYN/ACK packet outlined above. In either 523 case, Node B sends no data packets, only sending acknowledgement 524 packets in response to received data packets. 526 5. Related Work 528 The addition of ECN-capability to TCP's SYN/ACK packets was proposed 529 in [ECN+]. The paper includes an extensive set of simulation and 530 testbed experiments to evaluate the effects of the proposal, using 531 several Active Queue Management (AQM) mechanisms, including Random 532 Early Detection (RED) [RED], Random Exponential Marking (REM) [REM], 533 and Proportional Integrator (PI) [PI]. The performance measures were 534 the end-to-end response times for each request/response pair, and the 535 aggregate throughput on the bottleneck link. The end-to-end response 536 time was computed as the time from the moment when the request for 537 the file is sent to the server, until that file is successfully 538 downloaded by the client. 540 The measurements from [ECN+] show that setting an ECN-Capable 541 codepoint in the IP packet header in TCP SYN/ACK packets 542 systematically improves performance with all evaluated AQM schemes. 543 When SYN/ACK packets at a congested router are ECN-marked instead of 544 dropped, this can avoid a long initial retransmit timeout, improving 545 the response time for the affected flow dramatically. 547 [ECN+] shows that the impact on aggregate throughput can also be 548 quite significant, because marking SYN ACK packets can prevent larger 549 flows from suffering long timeouts before being "admitted" into the 550 network. In addition, the testbed measurements from [ECN+] show that 551 web servers setting the ECN-Capable codepoint in TCP SYN/ACK packets 552 could serve more requests. 554 As a final step, [ECN+] explores the co-existence of flows that do 555 and don't set the ECN-capable codepoint in TCP SYN/ACK packets. The 556 results in [ECN+] show that both types of flows can coexist, with 557 some performance degradation for flows that don't use ECN+. Flows 558 that do use ECN+ improve their end-to-end performance. At the same 559 time, the performance degradation for flows that don't use ECN+, as a 560 result of the flows that do use ECN+, increases as a greater fraction 561 of flows use ECN+. 563 6. Performance Evaluation 565 6.1. The Costs and Benefit of Adding ECN-Capability 567 [ECN+] explores the costs and benefits of adding ECN-Capability to 568 SYN/ACK packets with both simulations and experiments. The addition 569 of ECN-capability to SYN/ACK packets could be of significant benefit 570 for those ECN connections that would have had the SYN/ACK packet 571 dropped in the network, and for which the ECN-Capability would allow 572 the SYN/ACK to be marked rather than dropped. 574 The percent of SYN/ACK packets on a link can be quite high. In 575 particular, measurements on links dominated by web traffic indicate 576 that 15-20% of the packets can be SYN/ACK packets [SCJO01]. 578 The benefit of adding ECN-capability to SYN/ACK packets depends in 579 part on the size of the data transfer. The drop of a SYN/ACK packet 580 can increase the download time of a short file by an order of 581 magnitude, by requiring a three-second retransmit timeout. For 582 longer-lived flows, the effect of a dropped SYN/ACK packet on file 583 download time is less dramatic. However, even for longer-lived 584 flows, the addition of ECN-capability to SYN/ACK packets can improve 585 the fairness among long-lived flows, as newly-arriving flows would be 586 less likely to have to wait for retransmit timeouts. 588 One question that arises is what fraction of connections would see 589 the benefit from making SYN/ACK packets ECN-capable, in a particular 590 scenario. Specifically: 592 (1) What fraction of arriving SYN/ACK packets are dropped at the 593 congested router when the SYN/ACK packets are not ECN-capable? 595 (2) Of those SYN/ACK packets that are dropped, what fraction would 596 have been ECN-marked instead of dropped if the SYN/ACK packets had 597 been ECN-capable? 599 To answer (1), it is necessary to consider not only the level of 600 congestion but also the queue architecture at the congested link. As 601 described in Section 4 above, for some queue architectures small 602 packets are less likely to be dropped than large ones. In such an 603 environment, SYN/ACK packets would have lower packet drop rates; 604 question (1) could not necessarily be inferred from the overall 605 packet drop rate, but could be answered by measuring the drop rate 606 for SYN/ACK packets directly. In such an environment, adding ECN- 607 capability to SYN/ACK packets would be of less dramatic benefit than 608 in environments where all packets are equally likely to be dropped 609 regardless of packet size. 611 As question (2) implies, even if all of the SYN/ACK packets were ECN- 612 capable, there could still be some SYN/ACK packets dropped instead of 613 marked at the congested link; the full answer to question (2) depends 614 on the details of the queue management mechanism at the router. If 615 congestion is sufficiently bad, and the queue management mechanism 616 cannot prevent the buffer from overflowing, then SYN/ACK packets will 617 be dropped rather than marked upon buffer overflow whether or not 618 they are ECN-capable. 620 For some AQM mechanisms, ECN-capable packets are marked instead of 621 dropped any time this is possible, that is, any time the buffer is 622 not yet full. For other AQM mechanisms however, such as the RED 623 mechanism as recommended in [RED], packets are dropped rather than 624 marked when the packet drop/mark rate exceeds a certain threshold, 625 e.g., 10%, even if the packets are ECN-capable. For a router with 626 such an AQM mechanism, when congestion is sufficiently severe to 627 cause a high drop/mark rate, some SYN/ACK packets would be dropped 628 instead of marked whether or not they were ECN-capable. 630 Thus, the degree of benefit of adding ECN-Capability to SYN/ACK 631 packets depends not only on the overall packet drop rate in the 632 network, but also on the queue management architecture at the 633 congested link. 635 6.2. An Evaluation of Different Responses to ECN-Marked SYN/ACK Packets 637 This document specifies that the end-node responds to the report of 638 an ECN-marked SYN/ACK packet by setting the initial congestion window 639 to one segment, instead of its possible default value of two to four 640 segments. We call this ECN+ with NoWaiting. However, Section 4 641 discussed another possible response to an ECN-marked SYN/ACK packet, 642 of the end-node waiting an RTT before sending a data packet. We call 643 this approach ECN+ with Waiting. 645 Simulations comparing the performance with Standard ECN (without ECN- 646 marked SYN/ACK packets), ECN+ with NoWaiting, and ECN+ with Waiting 647 show little difference, in terms of aggregate congestion, between 648 ECN+ with NoWaiting and ECN+ with Waiting. The details are given in 649 Appendix A below. Our conclusions are that ECN+ with NoWaiting is 650 perfectly safe, and there are no congestion-related reasons for 651 preferring ECN+ with Waiting over ECN+ with NoWaiting. That is, 652 there is no need for the TCP end-node to wait a round-trip time 653 before sending a data packet after receiving an acknowledgement of an 654 ECN-marked SYN/ACK packet. 656 7. Security Considerations 658 TCP packets carrying the ECT codepoint in IP headers can be marked 659 rather than dropped by ECN-capable routers. This raises several 660 security concerns that we discuss below. 662 "Bad" routers or middleboxes: 663 There are a number of known deployment problems from using ECN with 664 TCP traffic in the Internet. The first reported problem, dating back 665 to 2000, is of a small but decreasing number of routers or 666 middleboxes that reset a TCP connection in response to TCP SYN 667 packets using flags in the TCP header to negotiate ECN-capability 668 [Kelson00] [RFC3360] [MAF05]. Dave Thaler reported at the March 2007 669 IETF of new two problems encountered by TCP connections using ECN; 670 the first of the two problems concerns routers that crash when a TCP 671 data packet arrives with the ECN field in the IP header with the 672 codepoint ECT(0) or ECT(1), indicating that an ECN-Capable connection 673 has been established [SBT07]. 675 While there is no evidence that any routers or middleboxes drop 676 SYN/ACK packets that contain an ECN-Capable or CE codepoint in the IP 677 header, such behavior cannot be excluded. (There seems to be a 678 number of routers or middleboxes that drop TCP SYN packets that 679 contain known or unknown IP options [MAF05] (Figure 1).) Thus, as 680 specified in Section 3, if a SYN/ACK packet with the ECT or CE 681 codepoint is dropped, the TCP node SHOULD resend the SYN/ACK packet 682 without the ECN-Capable codepoint. There is also no evidence that 683 any routers or middleboxes crash when a SYN/ACK arrives with an ECN- 684 Capable or CE codepoint in the IP header (over and above the routers 685 already known to crash when a data packet arrives with either ECT(0) 686 or ECT(1)), but we have not conducted any measurement studies of this 687 [F07]. 689 Congestion collapse: 690 Because TCP SYN/ACK packets carrying an ECT codepoint could be ECN- 691 marked instead of dropped at an ECN-capable router, the concern is 692 whether this can either invoke congestion, or worsen performance in 693 highly congested scenarios. However, after learning that a SYN/ACK 694 packet was ECN-marked, the responder will only send one data packet; 695 if this data packet is ECN-marked, the responder will then wait for a 696 retransmission timeout. In addition, routers are free to drop rather 697 than mark arriving packets in times of high congestion, regardless of 698 whether the packets are ECN-capable. When congestion is very high 699 and a router's buffer is full, the router has no choice but to drop 700 rather than to mark an arriving packet. 702 The simulations reported in Appendix A show that even with demanding 703 traffic mixes dominated by short flows and high levels of congestion, 704 the aggregate packet dropping rates are not significantly different 705 with Standard ECN, ECN+ with NoWaiting, or ECN+ with Waiting. In 706 particular, the simulations show that in periods of very high 707 congestion the packet-marking rate is low with or without ECN+, and 708 the use of ECN+ does not significantly increase the number of dropped 709 or marked packets. 711 The simulations show that ECN+ is most effective in times of moderate 712 congestion. In these moderate-congested scenarios, the use of ECN+ 713 increases the number of ECN-marked packets, because ECN+ allows 714 SYN/ACK packets to be ECN-marked. At the same time, in these times 715 of moderate congestion, the use of ECN+ instead of Standard ECN does 716 not significantly affect the overall levels of congestion. 718 The simulations show that the use of ECN+ is less effective in times 719 of high congestion; the simulations show that in times of high 720 congestion more packets are dropped instead of marked, both with 721 Standard ECN and with ECN+. In times of high congestion, the buffer 722 can overflow, even with Active Queue Management and ECN; when the 723 buffer is full arriving packets are dropped rather than marked, 724 whether the packets are ECN-capable or not. Thus while ECN+ is less 725 effective in times of high congestion, it still doesn't result in a 726 significant increase in the level of congestion. More details are 727 given in the appendix. 729 8. Conclusions 731 This draft specifies a modification to RFC 3168 to allow TCP nodes to 732 send SYN/ACK packets as being ECN-Capable. Making the SYN/ACK packet 733 ECN-Capable avoids the high cost to a TCP transfer when a SYN/ACK 734 packet is dropped by a congested router, by avoiding the resulting 735 retransmit timeout. This improves the throughput of short 736 connections. The sender of the SYN/ACK packet responds to an ECN 737 mark by reducing its initial congestion window from two, three, or 738 four segments to one segment, reducing the subsequent load from that 739 connection on the network. The addition of ECN-capability to SYN/ACK 740 packets is particularly beneficial in the server-to-client direction, 741 where congestion is more likely to occur. In this case, the initial 742 information provided by the ECN marking in the SYN/ACK packet enables 743 the server to more appropriately adjust the initial load it places on 744 the network. 746 Future work will address the more general question of adding ECN- 747 Capability to relevant handshake packets in other protocols that use 748 retransmission-based reliability in their setup phase (e.g., SCTP, 749 DCCP, HIP, and the like). 751 9. Acknowledgements 753 We thank Anil Agarwal, Mark Allman, Remi Denis-Courmont, Wesley Eddy, 754 Alfred Hoenes, Janardhan Iyengar, and Pasi Sarolahti for feedback on 755 earlier versions of this draft. 757 A. Report on Simulations 759 This section reports on simulations showing the costs of adding ECN+ 760 in highly-congested scenarios. This section also reports on 761 simulations for a comparative evaluation between ECN+ with NoWaiting 762 and ECN+ with Waiting. 764 The simulations are run with a range of file-size distributions. As 765 a baseline, they use the empirical heavy-tailed distribution reported 766 in [SCJO01], with a mean file size of around 7 KBytes. This flow- 767 size distribution is manipulated by skewing the flow sizes towards 768 lower and higher values to get distributions with mean file sizes of 769 3 KBytes, 5 KBytes, 14 KBytes and 17 KBytes. The congested link is 770 100 Mbps. RED is run in gentle mode, and arriving ECN-Capable 771 packets are only dropped instead of marked if the buffer is full (and 772 the router has no choice). 774 We explore two alternatives for a TCP node's response to a report of 775 an ECN-marked SYN/ACK packet. With ECN+ with NoWaiting, the TCP node 776 sends a data packet immediately (with an initial congestion window of 777 one segment). With the alternative ECN+ with Waiting, the TCP node 778 waits a round-trip time before sending a data packet; the responder 779 already has one measurement of the round-trip time when the 780 acknowledgement for the SYN/ACK packet is received. 782 In the tables below, ECN+ refers to ECN+ with NoWaiting, where the 783 responder starts transmitting immediately, and ECN+/wait refers to 784 ECN+ with Waiting, where the responder waits a round-trip time before 785 sending a data packet into the network. 787 The simulation scripts are available on [ECN-SYN], along with graphs 788 showing the distribution of response times for the TCP connections. 790 A.1. Simulations with RED in Packet Mode 792 The simulations with RED in packet mode and with the queue in packets 793 show that ECN+ is useful in times of moderate congestion, though it 794 adds little benefit in times of high congestion. The simulations 795 show a minimal increase in levels of congestion with either ECN+ with 796 Waiting or ECN+ with NoWaiting, either in terms of packet dropping or 797 marking rates or in terms of the distribution of responses times. 799 Thus, the simulations show no problems with ECN+ in times of high 800 congestion, and no reason to use ECN+ with Waiting instead of ECN+ 801 with NoWaiting. 803 Table 1 shows the congestion levels for simulations with RED in 804 packet mode, with a queue in packets. To explore a worst-case 805 scenario, these simulations use a traffic mix with an unrealistically 806 small flow size distribution, with a mean flow size of 3 Kbytes. For 807 each table showing a particular traffic load, the three rows show the 808 number of packets dropped, the number of packets ECN-marked, and the 809 aggregate packet drop rate, and the three columns show the 810 simulations with Standard ECN, ECN+ (NoWaiting) and ECN+/wait. 812 The usefulness of ECN+: The first thing to observe is that for the 813 simulations with the somewhat moderate load of 95%, with packet drop 814 rates of 5-6%, the use of ECN+ or ECN+/wait more than doubled the 815 number of packets marked. This indicates that with ECN+ or 816 ECN+/wait, many SYN/ACK packets are marked instead of dropped. 818 No increase in congestion: The second thing to observe is that in all 819 of the simulations, the use of ECN+ or ECN+/wait does not 820 significantly increase the aggregate packet drop rate. 822 Comparing ECN+ and ECN+/wait: The third thing to observe is that 823 there is little difference between ECN+ and ECN+/wait in terms of the 824 aggregate packet drop rate. Thus, there is no congestion-related 825 reason to prefer ECN+/wait over ECN+. 827 Traffic Load = 95%: 828 ECN ECN+ ECN+/wait 829 ------- ------- ------- 830 Dropped 74,645 64,034 64,983 831 Marked 7,639 17,681 16,914 832 Loss rate 6.05% 5.26% 5.33% 834 Traffic Load = 110%: 835 ECN ECN+ ECN+/wait 836 ------- ------- ------- 837 Dropped 161,644 163,620 165,196 838 Marked 4,375 6,653 6,144 839 Loss rate 10.38% 10.45% 10.53% 841 Traffic Load = 125%: 842 ECN ECN+ ECN+/wait 843 ------- ------- ------- 844 Dropped 257,671 268,161 264,437 845 Marked 2,885 3,712 3,359 846 Loss rate 14.52% 15.00% 14.83% 848 Traffic Load = 150%: 849 ECN ECN+ ECN+/wait 850 ------- ------- ------- 851 Loss rate 24.36% 24.61% 24.46% 853 Traffic Load = 200%: 854 ECN ECN+ ECN+/wait 855 ------- ------- ------- 856 Loss rate 29.99% 30.22% 30.23% 858 Table 1: Simulations with an average flow size of 3 Kbytes, RED in 859 packet mode, queue in packets. 861 A.2. Simulations with RED in Byte Mode 863 Table 2 below shows simulations with RED in byte mode and the queue 864 in bytes. Like the simulations with RED in packet mode, there is no 865 significant increase in aggregate congestion with the use of ECN+ or 866 ECN+/wait, and no congestion-related reason to prefer ECN+/wait over 867 ECN+. 869 However, unlike the simulations with RED in packet mode, the 870 simulations with RED in byte mode show little benefit from the use of 871 ECN+ or ECN+/wait, in that the packet marking rate with ECN+ or 872 ECN+/wait is not much different than the packet marking rate with 873 Standard ECN. This is because with RED in byte mode, small packets 874 like SYN/ACK packets are rarely dropped or marked - that is, there is 875 no drawback from the use of ECN+ in these scenarios, but not much 876 need for ECN+ either, in a scenario where small packets are unlikely 877 to be dropped or marked. 879 Traffic Load = 95%: 880 ECN ECN+ ECN+/wait 881 ------- ------- ------- 882 Dropped 13,044 13,323 14,855 883 Marked 18,880 19,175 19,049 884 Loss rate 1.13% 1.16% 1.29% 886 Traffic Load = 110%: 887 ECN ECN+ ECN+/wait 888 ------- ------- ------- 889 Dropped 84,809 83,013 83,564 890 Marked 4,086 4,644 4,826 891 Loss rate 5.90% 5.78% 5.81% 893 Traffic Load = 125%: 894 ECN ECN+ ECN+/wait 895 ------- ------- ------- 896 Dropped 157,305 157,435 158,368 897 Marked 2,183 2,363 2,663 898 Loss rate 9.89% 9.87% 9.93% 900 Table 2: Simulations with an average flow size of 3 Kbytes, RED in 901 byte mode, queue in bytes. 903 B. Issues of Incremental Deployment 905 In order for TCP node B to send a SYN/ACK packet as ECN-Capable, node 906 B must have received an ECN-setup SYN packet from node A. However, 907 it is possible that node A supports ECN, but either ignores the CE 908 codepoint on received SYN/ACK packets, or ignores SYN/ACK packets 909 with the ECT or CE codepoint set. If the TCP initiator ignores the 910 CE codepoint on received SYN/ACK packets, this would mean that the 911 TCP responder would not respond to this congestion indication. 912 However, this seems to us an acceptable cost to pay in the 913 incremental deployment of ECN-Capability for TCP's SYN/ACK packets. 914 It would mean that the responder would not reduce the initial 915 congestion window from two, three, or four segments down to one 916 segment, as it should. However, the TCP end nodes would still 917 respond correctly to any subsequent CE indications on data packets 918 later on in the connection. 920 Figure 3 shows an interchange with the SYN/ACK packet ECN-marked, but 921 with the ECN mark ignored by the TCP originator. 923 --------------------------------------------------------------- 924 TCP Node A Router TCP Node B 925 ---------- ------ ---------- 927 ECN-setup SYN packet ---> 928 ECN-setup SYN packet ---> 930 <--- ECN-setup SYN/ACK, ECT 931 <--- Sets CE on SYN/ACK 932 <--- ECN-setup SYN/ACK, CE 934 Data/ACK, No ECN-Echo ---> 935 Data/ACK ---> 936 <--- Data (up to four packets) 937 --------------------------------------------------------------- 939 Figure 3: SYN exchange with the SYN/ACK packet marked, 940 but with the ECN mark ignored by the TCP initiator. 942 Thus, to be explicit, when a TCP connection includes an initiator 943 that supports ECN but *does not* support ECN-Capability for SYN/ACK 944 packets, in combination with a responder that *does* support ECN- 945 Capabililty for SYN/ACK packets, it is possible that the ECN-Capable 946 SYN/ACK packets will be marked rather than dropped in the network, 947 and that the responder will not learn about the ECN mark on the 948 SYN/ACK packet. This would not be a problem if most packets from the 949 responder supporting ECN for SYN/ACK packets were in long-lived TCP 950 connections, but it would be more problematic if most of the packets 951 were from TCP connections consisting of four data packets, and the 952 TCP responder for these connections was ready to send its data 953 packets immediately after the SYN/ACK exchange. Of course, with 954 *severe* congestion, the SYN/ACK packets would likely be dropped 955 rather than ECN-marked at the congested router, preventing the TCP 956 responder from adding to the congestion by sending its initial window 957 of four data packets. 959 It is also possible that in some older TCP implementation, the 960 initiator would ignore arriving SYN/ACK packets that had the ECT or 961 CE codepoint set. This would result in a delay in connection set-up 962 for that TCP connection, with the initiator re-sending the SYN packet 963 after a retransmit timeout. We are not aware of any TCP 964 implementations with this behavior. 966 One possibility for coping with problems of backwards compatibility 967 would be for TCP initiators to use a TCP flag that means "I 968 understand ECN-Capable SYN/ACK packets". If this document were to 969 standardize the use of such an "ECN-SYN" flag, then the TCP responder 970 would only send a SYN/ACK packet as ECN-capable if the incoming SYN 971 packet had the "ECN-SYN" flag set. An ECN-SYN flag would prevent the 972 backwards compatibility problems described in the paragraphs above. 974 One drawback to the use of an ECN-SYN flag is that it would use one 975 of the four remaining reserved bits in the TCP header, for a 976 transient backwards compatibility problem. This drawback is limited 977 by the fact that the "ECN-SYN" flag would be defined only for use 978 with ECN-setup SYN packets; that bit in the TCP header could be 979 defined to have other uses for other kinds of TCP packets. 981 Factors in deciding not to use an ECN-SYN flag include the following: 983 (1) The limited installed base: At the time that this document was 984 written, the TCP implementations in Microsoft Vista and Mac OS X 985 included ECN, but ECN was not enabled by default [SBT07]. Thus, 986 there was not a large deployed base of ECN-Capable TCP 987 implementations. This limits the scope of any backwards 988 compatibility problems. 990 (2) Limits to the scope of the problem: The backwards compatibility 991 problem would not be serious enough to cause congestion collapse; 992 with severe congestion, the buffer at the congested router will 993 overflow, and the congested router will drop rather than ECN-mark 994 arriving SYN packets. Some active queue management mechanisms might 995 switch from packet-marking to packet-dropping in times of high 996 congestion before buffer overflow, as recommended in Section 19.1 of 997 RFC 3168. This helps to prevent congestion collapse problems with 998 the use of ECN. 1000 (3) Detection of and response to backwards-compatibility problems: A 1001 TCP responder such as a web server can't differentiate between a 1002 SYN/ACK packet that is not ECN-marked in the network, and a SYN/ACK 1003 packet that is ECN-marked, but where the ECN mark is ignored by the 1004 TCP initiator. However, a TCP responder *can* detect if a SYN/ACK 1005 packet is sent as ECN-capable and not reported as ECN-marked, but 1006 data packets are dropped or marked from the initial window of data. 1007 We will call this scenario "initial-window-congestion". If a web 1008 server frequently experienced initial-window congestion (without 1009 SYN/ACK congestion), then the web server *might* be experiencing 1010 backwards compatibility problems with ECN-Capable SYN/ACK packets, 1011 and could respond by not sending SYN/ACK packets as ECN-Capable. 1013 Normative References 1015 [RFC 2119] S. Bradner, Key words for use in RFCs to Indicate 1016 Requirement Levels, RFC 2119, March 1997. 1018 [RFC3168] K.K. Ramakrishnan, S. Floyd, and D. Black, The Addition of 1019 Explicit Congestion Notification (ECN) to IP, RFC 3168, Proposed 1020 Standard, September 2001. 1022 Informative References 1024 [ECN+] A. Kuzmanovic, The Power of Explicit Congestion Notification, 1025 SIGCOMM 2005. 1027 [ECN-SYN] ECN-SYN web page with simulation scripts, URL to be added. 1029 [F07] S. Floyd, "[BEHAVE] Response of firewalls and middleboxes to 1030 TCP SYN packets that are ECN-Capable?", August 2, 2007, email sent to 1031 the BEHAVE mailing list, URL "http://www1.ietf.org/mail- 1032 archive/web/behave/current/msg02644.html". 1034 [Kelson00] Dax Kelson, note sent to the Linux kernel mailing list, 1035 September 10, 2000. 1037 [MAF05] A. Medina, M. Allman, and S. Floyd. Measuring the Evolution 1038 of Transport Protocols in the Internet, ACM CCR, April 2005. 1040 [PI] C. Hollot, V. Misra, W. Gong, and D. Towsley, On Designing 1041 Improved Controllers for AQM Routers Supporting TCP Flows, April 1042 1998. 1044 [RED] Floyd, S., and Jacobson, V. Random Early Detection gateways 1045 for Congestion Avoidance . IEEE/ACM Transactions on Networking, V.1 1046 N.4, August 1993. 1048 [REM] S. Athuraliya, V. H. Li, S. H. Low and Q. Yin, REM: Active 1049 Queue Management, IEEE Network, May 2001. 1051 [RFC2309] B. Braden et al., Recommendations on Queue Management and 1052 Congestion Avoidance in the Internet, RFC 2309, April 1998. 1054 [RFC2581] M. Allman, V. Paxson, and W. Stevens, TCP Congestion 1055 Control, RFC 2581, April 1999. 1057 [RFC2988] V. Paxson and M. Allman, Computing TCP's Retransmission 1058 Timer, RFC 2988, November 2000. 1060 [RFC3042] M. Allman, H. Balakrishnan, and S. Floyd, Enhancing TCP's 1061 Loss Recovery Using Limited Transmit, RFC 3042, Proposed Standard, 1062 January 2001. 1064 [RFC3360] S. Floyd, Inappropriate TCP Resets Considered Harmful, RFC 1065 3360, August 2002. 1067 [RFC3390] M. Allman, S. Floyd, and C. Partridge, Increasing TCP's 1068 Initial Window, RFC 3390, October 2002. 1070 [SCJO01] F. Smith, F. Campos, K. Jeffay, D. Ott, What {TCP/IP} 1071 Protocol Headers Can Tell us about the Web, SIGMETRICS, June 2001. 1073 [SYN-COOK] Dan J. Bernstein, SYN cookies, 1997, see also 1074 1076 [SBT07] M. Sridharan, D. Bansal, and D. Thaler, Implementation Report 1077 on Experiences with Various TCP RFCs, Presentation in the TSVAREA, 1078 IETF 68, March 2007. URL 1079 "http://www3.ietf.org/proceedings/07mar/slides/tsvarea-3/sld6.htm". 1081 [Tools] S. Floyd and E. Kohler, Tools for the Evaluation of 1082 Simulation and Testbed Scenarios, Internet-draft draft-irtf-tmrg- 1083 tools-04, work in progress, July 2007. 1085 IANA Considerations 1087 There are no IANA considerations regarding this document. 1089 Authors' Addresses 1090 Aleksandar Kuzmanovic 1091 Phone: +1 (847) 467-5519 1092 Northwestern University 1093 Email: akuzma at northwestern.edu 1094 URL: http://cs.northwestern.edu/~a 1096 Amit Mondal 1097 Northwestern University 1098 Email: a-mondal at northwestern.edu 1100 Sally Floyd 1101 Phone: +1 (510) 666-2989 1102 ICIR (ICSI Center for Internet Research) 1103 Email: floyd@icir.org 1104 URL: http://www.icir.org/floyd/ 1106 K. K. Ramakrishnan 1107 Phone: +1 (973) 360-8764 1108 AT&T Labs Research 1109 Email: kkrama at research.att.com 1110 URL: http://www.research.att.com/info/kkrama 1112 Full Copyright Statement 1114 Copyright (C) The IETF Trust (2007). 1116 This document is subject to the rights, licenses and restrictions 1117 contained in BCP 78, and except as set forth therein, the authors 1118 retain all their rights. 1120 This document and the information contained herein are provided on an 1121 "AS IS" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS 1122 OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY, THE IETF TRUST AND 1123 THE INTERNET ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS 1124 OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF 1125 THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED 1126 WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. 1128 Intellectual Property 1130 The IETF takes no position regarding the validity or scope of any 1131 Intellectual Property Rights or other rights that might be claimed to 1132 pertain to the implementation or use of the technology described in 1133 this document or the extent to which any license under such rights 1134 might or might not be available; nor does it represent that it has 1135 made any independent effort to identify any such rights. Information 1136 on the procedures with respect to rights in RFC documents can be 1137 found in BCP 78 and BCP 79. 1139 Copies of IPR disclosures made to the IETF Secretariat and any 1140 assurances of licenses to be made available, or the result of an 1141 attempt made to obtain a general license or permission for the use of 1142 such proprietary rights by implementers or users of this 1143 specification can be obtained from the IETF on-line IPR repository at 1144 http://www.ietf.org/ipr. 1146 The IETF invites any interested party to bring to its attention any 1147 copyrights, patents or patent applications, or other proprietary 1148 rights that may cover technology that may be required to implement 1149 this standard. Please address the information to the IETF at ietf- 1150 ipr@ietf.org.