idnits 2.17.1 draft-ietf-tcpm-alternativebackoff-ecn-12.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 : ---------------------------------------------------------------------------- No issues found here. Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year == The document seems to lack the recommended RFC 2119 boilerplate, even if it appears to use RFC 2119 keywords -- however, there's a paragraph with a matching beginning. Boilerplate error? (The document does seem to have the reference to RFC 2119 which the ID-Checklist requires). -- The document date (September 14, 2018) is 2048 days in the past. Is this intentional? Checking references for intended status: Experimental ---------------------------------------------------------------------------- == Outdated reference: A later version (-28) exists of draft-ietf-tcpm-accurate-ecn-06 -- Obsolete informational reference (is this intentional?): RFC 8312 (Obsoleted by RFC 9438) Summary: 0 errors (**), 0 flaws (~~), 3 warnings (==), 2 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 Network Working Group N. Khademi 3 Internet-Draft M. Welzl 4 Intended status: Experimental University of Oslo 5 Expires: March 18, 2019 G. Armitage 6 Netflix 7 G. Fairhurst 8 University of Aberdeen 9 September 14, 2018 11 TCP Alternative Backoff with ECN (ABE) 12 draft-ietf-tcpm-alternativebackoff-ecn-12 14 Abstract 16 Active Queue Management (AQM) mechanisms allow for burst tolerance 17 while enforcing short queues to minimise the time that packets spend 18 enqueued at a bottleneck. This can cause noticeable performance 19 degradation for TCP connections traversing such a bottleneck, 20 especially if there are only a few flows or their bandwidth-delay- 21 product is large. The reception of a Congestion Experienced (CE) ECN 22 mark indicates that an AQM mechanism is used at the bottleneck, and 23 therefore the bottleneck network queue is likely to be short. 24 Feedback of this signal allows the TCP sender-side ECN reaction in 25 congestion avoidance to reduce the Congestion Window (cwnd) by a 26 smaller amount than the congestion control algorithm's reaction to 27 inferred packet loss. This specification therefore defines an 28 experimental change to the TCP reaction specified in RFC3168, as 29 permitted by RFC 8311. 31 Status of This Memo 33 This Internet-Draft is submitted in full conformance with the 34 provisions of BCP 78 and BCP 79. 36 Internet-Drafts are working documents of the Internet Engineering 37 Task Force (IETF). Note that other groups may also distribute 38 working documents as Internet-Drafts. The list of current Internet- 39 Drafts is at https://datatracker.ietf.org/drafts/current/. 41 Internet-Drafts are draft documents valid for a maximum of six months 42 and may be updated, replaced, or obsoleted by other documents at any 43 time. It is inappropriate to use Internet-Drafts as reference 44 material or to cite them other than as "work in progress." 46 This Internet-Draft will expire on March 18, 2019. 48 Copyright Notice 50 Copyright (c) 2018 IETF Trust and the persons identified as the 51 document authors. All rights reserved. 53 This document is subject to BCP 78 and the IETF Trust's Legal 54 Provisions Relating to IETF Documents 55 (https://trustee.ietf.org/license-info) in effect on the date of 56 publication of this document. Please review these documents 57 carefully, as they describe your rights and restrictions with respect 58 to this document. Code Components extracted from this document must 59 include Simplified BSD License text as described in Section 4.e of 60 the Trust Legal Provisions and are provided without warranty as 61 described in the Simplified BSD License. 63 Table of Contents 65 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 2 66 2. Definitions . . . . . . . . . . . . . . . . . . . . . . . . . 3 67 3. Specification . . . . . . . . . . . . . . . . . . . . . . . . 4 68 3.1. Choice of ABE Multiplier . . . . . . . . . . . . . . . . 4 69 4. Discussion . . . . . . . . . . . . . . . . . . . . . . . . . 6 70 4.1. Why Use ECN to Vary the Degree of Backoff? . . . . . . . 6 71 4.2. An RTT-based response to indicated congestion . . . . . . 7 72 5. ABE Deployment Requirements . . . . . . . . . . . . . . . . . 7 73 6. ABE Experiment Goals . . . . . . . . . . . . . . . . . . . . 8 74 7. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 8 75 8. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 9 76 9. Implementation Status . . . . . . . . . . . . . . . . . . . . 9 77 10. Security Considerations . . . . . . . . . . . . . . . . . . . 9 78 11. Revision Information . . . . . . . . . . . . . . . . . . . . 9 79 12. References . . . . . . . . . . . . . . . . . . . . . . . . . 11 80 12.1. Normative References . . . . . . . . . . . . . . . . . . 11 81 12.2. Informative References . . . . . . . . . . . . . . . . . 11 82 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 13 84 1. Introduction 86 Explicit Congestion Notification (ECN) [RFC3168] makes it possible 87 for an Active Queue Management (AQM) mechanism to signal the presence 88 of incipient congestion without necessarily incurring packet loss. 89 This lets the network deliver some packets to an application that 90 would have been dropped if the application or transport did not 91 support ECN. This packet loss reduction is the most obvious benefit 92 of ECN, but it is often relatively modest. Other benefits of 93 deploying ECN have been documented in RFC8087 [RFC8087]. 95 The rules for ECN were originally written to be very conservative, 96 and required the congestion control algorithms of ECN-Capable 97 transport protocols to treat indications of congestion signalled by 98 ECN exactly the same as they would treat an inferred packet loss 99 [RFC3168]. Research has demonstrated the benefits of reducing 100 network delays that are caused by interaction of loss-based TCP 101 congestion control and excessive buffering [BUFFERBLOAT]. This has 102 led to the creation of AQM mechanisms like Proportional Integral 103 Controller Enhanced (PIE) [RFC8033] and Controlling Queue Delay 104 (CoDel) [CODEL2012][RFC8289], which prevent bloated queues that are 105 common with unmanaged and excessively large buffers deployed across 106 the Internet [BUFFERBLOAT]. 108 The AQM mechanisms mentioned above aim to keep a sustained queue 109 short while tolerating transient (short-term) packet bursts. 110 However, currently used loss-based congestion control mechanisms are 111 not always able to effectively utilise a bottleneck link where there 112 are short queues. For example, a TCP sender using the Reno 113 congestion control needs to be able to store at least an end-to-end 114 bandwidth-delay product (BDP) worth of data at the bottleneck buffer 115 if it is to maintain full path utilisation in the face of loss- 116 induced reduction of the congestion window (cwnd) [RFC5681]. This 117 amount of buffering effectively doubles the amount of data that can 118 be in flight and the maximum round-trip time (RTT) experienced by the 119 TCP sender. 121 Modern AQM mechanisms can use ECN to signal the early signs of 122 impending queue buildup long before a tail-drop queue would be forced 123 to resort to dropping packets. It is therefore appropriate for the 124 transport protocol congestion control algorithm to have a more 125 measured response when it receives an indication with an early- 126 warning of congestion after the remote endpoint receives an ECN CE- 127 marked packet. Recognizing these changes in modern AQM practices, 128 the strict requirement that ECN CE signals be treated identically to 129 inferred packet loss has been relaxed [RFC8311]. This document 130 therefore defines a new sender-side-only congestion control response, 131 called "ABE" (Alternative Backoff with ECN). ABE improves TCP's 132 average throughput when routers use AQM controlled buffers that allow 133 only for short queues. 135 2. Definitions 137 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 138 "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and 139 "OPTIONAL" in this document are to be interpreted as described in BCP 140 14 RFC 2119 [RFC2119] [RFC8174] when, and only when, they appear in 141 all capitals, as shown here. 143 3. Specification 145 This specification changes the congestion control algorithm of an 146 ECN-Capable TCP transport protocol by changing the TCP sender 147 response to feedback from the TCP receiver that indicates reception 148 of a CE-marked packet, i.e., receipt of a packet with the ECN-Echo 149 flag (defined in [RFC3168]) set, following the process defined in 150 [RFC8311]. 152 The TCP sender response is currently specified in section 6.1.2 of 153 the ECN specification [RFC3168], updated by [RFC8311]: 155 The indication of congestion should be treated just as a 156 congestion loss in non-ECN-Capable TCP. That is, the TCP source 157 halves the congestion window "cwnd" and reduces the slow start 158 threshold "ssthresh", unless otherwise specified by an 159 Experimental RFC in the IETF document stream. 161 Following publication of RFC 8311, this document specifies a sender- 162 side change to TCP: 164 Receipt of a packet with the ECN-Echo flag SHOULD trigger the TCP 165 source to set the slow start threshold (ssthresh) to 0.8 times the 166 FlightSize, with a lower bound of 2 * SMSS applied to the result. 167 As in [RFC5681], the TCP sender also reduces the cwnd value to no 168 more than the new ssthresh value. RFC 3168 section 6.1.2 provides 169 guidance on setting a cwnd less than 2 * SMSS. 171 3.1. Choice of ABE Multiplier 173 ABE decouples the reaction of a TCP sender to inferred packet loss 174 and indication of ECN-signalled congestion in the congestion 175 avoidance phase. To achieve this, ABE uses a different scaling 176 factor in Equation 4 in Section 3.1 of [RFC5681]. The description 177 respectively uses beta_{loss} and beta_{ecn} to refer to the 178 multiplicative decrease factors applied in response to inferred 179 packet loss, and in response to a receiver indicating ECN-signalled 180 congestion. For non-ECN-enabled TCP connections, only beta_{loss} 181 applies. 183 In other words, in response to inferred packet loss: 185 ssthresh = max (FlightSize * beta_{loss}, 2 * SMSS) 187 and in response to an indication of an ECN-signalled congestion: 189 ssthresh = max (FlightSize * beta_{ecn}, 2 * SMSS) 190 and 192 cwnd = ssthresh 194 (If ssthresh == 2 * SMSS, RFC 3168 section 6.1.2 provides guidance 195 on setting a cwnd lower than 2 * SMSS.) 197 where FlightSize is the amount of outstanding data in the network, 198 upper-bounded by the smaller of the sender's cwnd and the receiver's 199 advertised window (rwnd) [RFC5681]. The higher the values of 200 beta_{loss} and beta_{ecn}, the less aggressive the response of any 201 individual backoff event. 203 The appropriate choice for beta_{loss} and beta_{ecn} values is a 204 balancing act between path utilisation and draining the bottleneck 205 queue. More aggressive backoff (smaller beta_*) risks underutilising 206 the path, while less aggressive backoff (larger beta_*) can result in 207 slower draining of the bottleneck queue. 209 The Internet has already been running with at least two different 210 beta_{loss} values for several years: the standard value is 0.5 211 [RFC5681], and the Linux implementation of CUBIC [RFC8312] has used a 212 multiplier of 0.7 since kernel version 2.6.25 released in 2008. ABE 213 does not change the value of beta_{loss} used by current TCP 214 implementations. 216 The recommendation in this document specifies a value of 217 beta_{ecn}=0.8. This recommended beta_{ecn} value is only applicable 218 for the standard TCP congestion control [RFC5681]. The selection of 219 beta_{ecn} enables tuning the response of a TCP connection to shallow 220 AQM marking thresholds. beta_{loss} characterizes the response of a 221 congestion control algorithm to packet loss, i.e., exhaustion of 222 buffers (of unknown depth). Different values for beta_{loss} have 223 been suggested for TCP congestion control algorithms. Consequently, 224 beta_{ecn} is likely to be an algorithm-specific parameter rather 225 than a constant multiple of the algorithm's existing beta_{loss}. 227 A range of tests (section IV, [ABE2017]) with NewReno and CUBIC over 228 CoDel and PIE in lightly-multiplexed scenarios have explored this 229 choice of parameter. The results of these tests indicate that CUBIC 230 connections benefit from beta_{ecn} of 0.85 (cf. beta_{loss} = 0.7), 231 and NewReno connections see improvements with beta_{ecn} in the range 232 0.7 to 0.85 (cf. beta_{loss} = 0.5). 234 4. Discussion 236 Much of the technical background to ABE can be found in a research 237 paper [ABE2017]. This paper used a mix of experiments, theory and 238 simulations with NewReno [RFC5681] and CUBIC [RFC8312] to evaluate 239 the technique. The technique was shown to present "...significant 240 performance gains in lightly-multiplexed [few concurrent flows] 241 scenarios, without losing the delay-reduction benefits of deploying 242 CoDel or PIE". The performance improvement is achieved when reacting 243 to ECN-Echo in congestion avoidance (when ssthresh > cwnd) by 244 multiplying cwnd and ssthresh with a value in the range [0.7,0.85]. 245 Applying ABE when cwnd <= ssthresh is not currently recommended, but 246 may benefit from additional attention, experimentation and 247 specification. 249 4.1. Why Use ECN to Vary the Degree of Backoff? 251 AQM mechanisms such as CoDel [RFC8289] and PIE [RFC8033] set a delay 252 target in routers and use congestion notifications to constrain the 253 queuing delays experienced by packets, rather than in response to 254 impending or actual bottleneck buffer exhaustion. With current 255 default delay targets, CoDel and PIE both effectively emulate a 256 bottleneck with a short queue (section II, [ABE2017]) while also 257 allowing short traffic bursts into the queue. This provides 258 acceptable performance for TCP connections over a path with a low 259 BDP, or in highly multiplexed scenarios (many concurrent transport 260 flows). However, in a lightly-multiplexed case over a path with a 261 large BDP, conventional TCP backoff leads to gaps in packet 262 transmission and under-utilisation of the path. 264 Instead of discarding packets, an AQM mechanism is allowed to mark 265 ECN-Capable packets with an ECN CE-mark. The reception of a CE-mark 266 feedback not only indicates congestion on the network path, it also 267 indicates that an AQM mechanism exists at the bottleneck along the 268 path, and hence the CE-mark likely came from a bottleneck with a 269 controlled short queue. Reacting differently to an ECN-signalled 270 congestion than to an inferred packet loss can then yield the benefit 271 of a reduced back-off when queues are short. Using ECN can also be 272 advantageous for several other reasons [RFC8087]. 274 The idea of reacting differently to inferred packet loss and 275 detection of an ECN-signalled congestion pre-dates this 276 specification. For example, previous research proposed using ECN CE- 277 marked feedback to modify TCP congestion control behaviour via a 278 larger multiplicative decrease factor in conjunction with a smaller 279 additive increase factor [ICC2002]. The goal of this former work was 280 to operate across AQM bottlenecks using Random Early Detection (RED) 281 that were not necessarily configured to emulate a short queue (The 282 current usage of RED as an Internet AQM method is limited [RFC7567]). 284 4.2. An RTT-based response to indicated congestion 286 This specification applies to the use of ECN feedback as defined in 287 [RFC3168], which specifies a response to indicated congestion that is 288 no more frequent that once per path round trip time. Since ABE 289 responds to indicated congestion once per RTT, it therefore does not 290 respond to any further loss within the same RTT, because an ABE 291 sender has already reduced the congestion window. If congestion 292 persists after such reduction, ABE continues to reduce the congestion 293 window in each consecutive RTT. This consecutive reduction can 294 protect the network against long-standing unfairness in the case of 295 AQM algorithms that do not keep a small average queue length. The 296 mechanism does not rely on Accurate ECN 297 ([I-D.ietf-tcpm-accurate-ecn]). 299 In contrast, transport protocol mechanisms can also be designed to 300 utilise more frequent and detailed ECN feedback (e.g., Accurate ECN 301 [I-D.ietf-tcpm-accurate-ecn]), which then permit a congestion control 302 response that adjusts the sending rate more frequently. Datacenter 303 TCP (DCTCP) [RFC8257] is an example of this approach. 305 5. ABE Deployment Requirements 307 This update is a sender-side only change. Like other changes to 308 congestion control algorithms, it does not require any change to the 309 TCP receiver or to network devices. It does not require any ABE- 310 specific changes in routers or the use of Accurate ECN feedback 311 [I-D.ietf-tcpm-accurate-ecn] by a receiver. 313 If the method is only deployed by some senders, and not by others, 314 the senders that use this method can gain some advantage, possibly at 315 the expense of other flows that do not use this updated method. 316 Because this advantage applies only to ECN-marked packets and not to 317 packet loss indications, an ECN-Capable bottleneck will still fall 318 back to dropping packets if an TCP sender using ABE is too 319 aggressive, and the result is no different than if the TCP sender was 320 using traditional loss-based congestion control. 322 When used with bottlenecks that do not support ECN-marking the 323 specification does not modify the transport protocol. 325 6. ABE Experiment Goals 327 [RFC3168] states that the congestion control response following an 328 indication of ECN-signalled congestion is the same as the response to 329 a dropped packet. [RFC8311] updates this specification to allow 330 systems to provide a different behaviour when they experience ECN- 331 signalled congestion rather than packet loss. The present 332 specification defines such an experiment and has thus been assigned 333 an Experimental status before being proposed as a Standards-Track 334 update. 336 The purpose of the Internet experiment is to collect experience with 337 deployment of ABE, and confirm acceptable safety in deployed networks 338 that use this update to TCP congestion control. To evaluate ABE, 339 this experiment therefore requires support in AQM routers for ECN- 340 marking of packets carrying the ECN-Capable Transport, ECT(0), 341 codepoint [RFC3168]. 343 The result of this Internet experiment ought to include an 344 investigation of the implications of experiencing an ECN-CE mark 345 followed by loss within the same RTT. At the end of the experiment, 346 this will be reported to the TCPM WG or the IESG. 348 7. Acknowledgements 350 Authors N. Khademi, M. Welzl and G. Fairhurst were part-funded by 351 the European Community under its Seventh Framework Programme through 352 the Reducing Internet Transport Latency (RITE) project (ICT-317700). 353 The views expressed are solely those of the authors. 355 Author G. Armitage performed most of his work on this document while 356 employed by Swinburne University of Technology, Melbourne, Australia. 358 The authors would like to thank Stuart Cheshire for many suggestions 359 when revising the draft, and the following people for their 360 contributions to [ABE2017]: Chamil Kulatunga, David Ros, Stein 361 Gjessing, Sebastian Zander. Thanks also to (in alphabetical order) 362 Roland Bless, Bob Briscoe, David Black, Markku Kojo, John Leslie, 363 Lawrence Stewart, Dave Taht and the TCPM Working Group for providing 364 valuable feedback on this document. 366 The authors would finally like to thank everyone who provided 367 feedback on the congestion control behaviour specified in this update 368 received from the IRTF Internet Congestion Control Research Group 369 (ICCRG). 371 8. IANA Considerations 373 XX RFC ED - PLEASE REMOVE THIS SECTION XXX 375 This document includes no request to IANA. 377 9. Implementation Status 379 ABE is implemented as a patch for Linux and FreeBSD. This is meant 380 for research and available for download from 381 http://heim.ifi.uio.no/michawe/research/abe/. This code was used to 382 produce the test results that are reported in [ABE2017]. The FreeBSD 383 code has been committed to the mainline kernel on March 19, 2018 384 [ABE-FreeBSD]. 386 10. Security Considerations 388 The described method is a sender-side only transport change, and does 389 not change the protocol messages exchanged. The security 390 considerations for ECN [RFC3168] therefore still apply. 392 This is a change to TCP congestion control with ECN that will 393 typically lead to a change in the capacity achieved when flows share 394 a network bottleneck. This could result in some flows receiving more 395 than their fair share of capacity. Similar unfairness in the way 396 that capacity is shared is also exhibited by other congestion control 397 mechanisms that have been in use in the Internet for many years 398 (e.g., CUBIC [RFC8312]). Unfairness may also be a result of other 399 factors, including the round trip time experienced by a flow. ABE 400 applies only when ECN-marked packets are received, not when packets 401 are lost, hence use of ABE cannot lead to congestion collapse. 403 11. Revision Information 405 XX RFC ED - PLEASE REMOVE THIS SECTION XXX 407 -12. Corrections from Adam Roach; Benjamin Kaduk; & Ben Campbell 409 -10. Incorported changes following the Gen-ART review by Russ 410 Housley. Correction to URL. 412 -09. Changed to "Following publication of RFC 8311, this document 413 specifies a sender-side change to TCP:" 415 -08. Addressed comments from AD review on the document structure, 416 and relationship to existing RFCs. 418 -07. Addressed comments following WGLC. 420 o Updated Reference citations. 422 o Removed paragraph containing a wrong statement related to timeout 423 in section 4.1. 425 o Discuss what happens when cwnd <= ssthresh. 427 o Added text on Concern about lower bound of 2*SMSS. 429 -06. Addressed Michael Scharf's comments. 431 -05. Refined the description of the experiment based on feedback at 432 IETF-100. Incorporated comments from David Black. 434 -04. Incorporates review comments from Lawrence Stewart and the 435 remaining comments from Roland Bless. References are updated. 437 -03. Several review comments from Roland Bless are addressed. 438 Consistent terminology and equations. Clarification on the scope of 439 recommended beta_{ecn} value. 441 -02. Corrected the equations in Section 3.1. Updated the 442 affiliations. Lower bound for cwnd is defined. A recommendation for 443 window-based transport protocols is changed to cover all transport 444 protocols that implement a congestion control reduction to an ECN 445 congestion signal. Added text about ABE's FreeBSD mainline kernel 446 status including a reference to the FreeBSD code review page. 447 References are updated. 449 -01. Text improved, mainly incorporating comments from Stuart 450 Cheshire. The reference to a technical report has been updated to a 451 published version of the tests [ABE2017]. Used "AQM Mechanism" 452 throughout in place of other alternatives, and more consistent use of 453 technical language and clarification on the intended purpose of the 454 experiments required by EXP status. There was no change to the 455 technical content. 457 -00. draft-ietf-tcpm-alternativebackoff-ecn-00 replaces draft- 458 khademi-tcpm-alternativebackoff-ecn-01. Text describing the nature 459 of the experiment was added. 461 Individual draft -01. This I-D now refers to draft-black-tsvwg-ecn- 462 experimentation-02, which replaces draft-khademi-tsvwg-ecn- 463 response-00 to make a broader update to RFC 3168 for the sake of 464 allowing experiments. As a result, some of the motivating and 465 discussing text that was moved from draft-khademi-alternativebackoff- 466 ecn-03 to draft-khademi-tsvwg-ecn-response-00 has now been re- 467 inserted here. 469 Individual draft -00. draft-khademi-tsvwg-ecn-response-00 and draft- 470 khademi-tcpm-alternativebackoff-ecn-00 replace draft-khademi- 471 alternativebackoff-ecn-03, following discussion in the TSVWG and TCPM 472 working groups. 474 12. References 476 12.1. Normative References 478 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 479 Requirement Levels", BCP 14, RFC 2119, 480 DOI 10.17487/RFC2119, March 1997, 481 . 483 [RFC3168] Ramakrishnan, K., Floyd, S., and D. Black, "The Addition 484 of Explicit Congestion Notification (ECN) to IP", 485 RFC 3168, DOI 10.17487/RFC3168, September 2001, 486 . 488 [RFC5681] Allman, M., Paxson, V., and E. Blanton, "TCP Congestion 489 Control", RFC 5681, DOI 10.17487/RFC5681, September 2009, 490 . 492 [RFC7567] Baker, F., Ed. and G. Fairhurst, Ed., "IETF 493 Recommendations Regarding Active Queue Management", 494 BCP 197, RFC 7567, DOI 10.17487/RFC7567, July 2015, 495 . 497 [RFC8174] Leiba, B., "Ambiguity of Uppercase vs Lowercase in RFC 498 2119 Key Words", BCP 14, RFC 8174, DOI 10.17487/RFC8174, 499 May 2017, . 501 [RFC8257] Bensley, S., Thaler, D., Balasubramanian, P., Eggert, L., 502 and G. Judd, "Data Center TCP (DCTCP): TCP Congestion 503 Control for Data Centers", RFC 8257, DOI 10.17487/RFC8257, 504 October 2017, . 506 [RFC8311] Black, D., "Relaxing Restrictions on Explicit Congestion 507 Notification (ECN) Experimentation", RFC 8311, 508 DOI 10.17487/RFC8311, January 2018, 509 . 511 12.2. Informative References 513 [ABE-FreeBSD] 514 "ABE patch review in FreeBSD", 515 . 518 [ABE2017] Khademi, N., Armitage, G., Welzl, M., Fairhurst, G., 519 Zander, S., and D. Ros, "Alternative Backoff: Achieving 520 Low Latency and High Throughput with ECN and AQM", IFIP 521 NETWORKING 2017, Stockholm, Sweden, June 2017. 523 [BUFFERBLOAT] 524 Gettys, J. and K. Nichols, "Bufferbloat: Dark Buffers in 525 the Internet", ACM Queue 9, 11, DOI 526 10.1145/2063166.2071893; 527 https://queue.acm.org/detail.cfm?id=2071893", November 528 2011. 530 [CODEL2012] 531 Nichols, K. and V. Jacobson, "Controlling Queue Delay", 532 July 2012, . 534 [I-D.ietf-tcpm-accurate-ecn] 535 Briscoe, B., Kuehlewind, M., and R. Scheffenegger, "More 536 Accurate ECN Feedback in TCP", draft-ietf-tcpm-accurate- 537 ecn-06 (work in progress), March 2018. 539 [ICC2002] Kwon, M. and S. Fahmy, "TCP Increase/Decrease Behavior 540 with Explicit Congestion Notification (ECN)", IEEE 541 ICC 2002, New York, New York, USA, May 2002, 542 . 544 [RFC8033] Pan, R., Natarajan, P., Baker, F., and G. White, 545 "Proportional Integral Controller Enhanced (PIE): A 546 Lightweight Control Scheme to Address the Bufferbloat 547 Problem", RFC 8033, DOI 10.17487/RFC8033, February 2017, 548 . 550 [RFC8087] Fairhurst, G. and M. Welzl, "The Benefits of Using 551 Explicit Congestion Notification (ECN)", RFC 8087, 552 DOI 10.17487/RFC8087, March 2017, 553 . 555 [RFC8289] Nichols, K., Jacobson, V., McGregor, A., Ed., and J. 556 Iyengar, Ed., "Controlled Delay Active Queue Management", 557 RFC 8289, DOI 10.17487/RFC8289, January 2018, 558 . 560 [RFC8312] Rhee, I., Xu, L., Ha, S., Zimmermann, A., Eggert, L., and 561 R. Scheffenegger, "CUBIC for Fast Long-Distance Networks", 562 RFC 8312, DOI 10.17487/RFC8312, February 2018, 563 . 565 Authors' Addresses 567 Naeem Khademi 568 University of Oslo 569 PO Box 1080 Blindern 570 Oslo N-0316 571 Norway 573 Email: naeemk@ifi.uio.no 575 Michael Welzl 576 University of Oslo 577 PO Box 1080 Blindern 578 Oslo N-0316 579 Norway 581 Email: michawe@ifi.uio.no 583 Grenville Armitage 584 Netflix Inc. 586 Email: garmitage@netflix.com 588 Godred Fairhurst 589 University of Aberdeen 590 School of Engineering, Fraser Noble Building 591 Aberdeen AB24 3UE 592 UK 594 Email: gorry@erg.abdn.ac.uk