idnits 2.17.1 draft-ietf-tcpm-newcwv-00.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 obsoletes RFC2861, but the abstract doesn't seem to directly say this. It does mention RFC2861 though, so this could be OK. -- The draft header indicates that this document updates RFC5681, but the abstract doesn't seem to directly say this. It does mention RFC5681 though, so this could be OK. Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year (Using the creation date from RFC5681, updated by this document, for RFC5378 checks: 2006-01-26) -- 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 (February 14, 2013) is 4089 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 normative reference: RFC 793 (Obsoleted by RFC 9293) ** Obsolete normative reference: RFC 2861 (Obsoleted by RFC 7661) ** Obsolete normative reference: RFC 3517 (Obsoleted by RFC 6675) Summary: 3 errors (**), 0 flaws (~~), 1 warning (==), 4 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 TCPM Working Group G. Fairhurst 3 Internet-Draft A. Sathiaseelan 4 Obsoletes: 2861 (if approved) University of Aberdeen 5 Updates: 5681 (if approved) February 14, 2013 6 Intended status: Standards Track 7 Expires: August 18, 2013 9 Updating TCP to support Rate-Limited Traffic 10 draft-ietf-tcpm-newcwv-00 12 Abstract 14 This document proposes an update to RFC 5681 to address issues that 15 arise when TCP is used to support traffic that exhibits periods where 16 the sending rate is limited by the application rather than the 17 congestion window. It updates TCP to allow a TCP sender to restart 18 quickly following either an idle or rate-limited interval. This 19 method is expected to benefit applications that send rate-limited 20 traffic using TCP, while also providing an appropriate response if 21 congestion is experienced. 23 It also evaluates TCP Congestion Window Validation, CWV, an IETF 24 experimental specification defined in RFC 2861, and concludes that 25 CWV sought to address important issues, but failed to deliver a 26 widely used solution. This document therefore proposes an update to 27 the status of RFC 2861 by recommending it is moved from Experimental 28 to Historic status, and that it is replaced by the current 29 specification. 31 NOTE: The standards status of this WG document is under review for 32 consideration as either Experimental (EXP) or Proposed Standard (PS). 33 This decision will be made later as the document is finalised. 35 Status of this Memo 37 This Internet-Draft is submitted in full conformance with the 38 provisions of BCP 78 and BCP 79. 40 Internet-Drafts are working documents of the Internet Engineering 41 Task Force (IETF). Note that other groups may also distribute 42 working documents as Internet-Drafts. The list of current Internet- 43 Drafts is at http://datatracker.ietf.org/drafts/current/. 45 Internet-Drafts are draft documents valid for a maximum of six months 46 and may be updated, replaced, or obsoleted by other documents at any 47 time. It is inappropriate to use Internet-Drafts as reference 48 material or to cite them other than as "work in progress." 49 This Internet-Draft will expire on August 18, 2013. 51 Copyright Notice 53 Copyright (c) 2013 IETF Trust and the persons identified as the 54 document authors. All rights reserved. 56 This document is subject to BCP 78 and the IETF Trust's Legal 57 Provisions Relating to IETF Documents 58 (http://trustee.ietf.org/license-info) in effect on the date of 59 publication of this document. Please review these documents 60 carefully, as they describe your rights and restrictions with respect 61 to this document. Code Components extracted from this document must 62 include Simplified BSD License text as described in Section 4.e of 63 the Trust Legal Provisions and are provided without warranty as 64 described in the Simplified BSD License. 66 Table of Contents 68 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 4 69 2. Reviewing experience with TCP-CWV . . . . . . . . . . . . . . 4 70 3. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . 5 71 4. An updated TCP response to idle and application-limited 72 periods . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 73 4.1. A method for preserving cwnd during the idle and 74 application-limited periods. . . . . . . . . . . . . . . . 7 75 4.2. The nonvalidated phase . . . . . . . . . . . . . . . . . . 7 76 4.3. TCP congestion control during the nonvalidated phase . . . 8 77 4.3.1. Response to congestion in the nonvalidated phase . . . 9 78 4.3.2. Adjustment at the end of the nonvalidated phase . . . 9 79 5. Determining a safe period to preserve cwnd . . . . . . . . . . 10 80 6. Security Considerations . . . . . . . . . . . . . . . . . . . 11 81 7. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 11 82 8. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . 11 83 9. Author Notes . . . . . . . . . . . . . . . . . . . . . . . . . 12 84 9.1. Other related work . . . . . . . . . . . . . . . . . . . . 12 85 9.2. Revision notes . . . . . . . . . . . . . . . . . . . . . . 14 86 10. References . . . . . . . . . . . . . . . . . . . . . . . . . . 15 87 10.1. Normative References . . . . . . . . . . . . . . . . . . . 15 88 10.2. Informative References . . . . . . . . . . . . . . . . . . 15 89 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . . 16 91 1. Introduction 93 TCP is used to support a range of application behaviours. The TCP 94 congestion window (cwnd) controls the number of unacknoeledged 95 packets/bytes that a TCP flow may have in the network at any time, a 96 value known as the FlightSize [RFC5681]. A bulk application will 97 always have data available to transmit. The rate at which it sends 98 is therefore limited by the maximum permitted by the receiver and 99 congestion windows. In contrast, a rate-limited application will 100 experience periods when the sender is either idle or is unable to 101 send at the maximum rate permitted by the cwnd. This latter case is 102 called rate-limited. The focus of this document is on the operation 103 of TCP in such an idle or rate-limited case. 105 Standard TCP [RFC5681] requires the cwnd to be reset to the restart 106 window (RW) when an application becomes idle. [RFC2861] noted that 107 this TCP behaviour was not always observed in current 108 implementations. Recent experiments [Bis08] confirm this to still be 109 the case. 111 Standard TCP does not impose additional restrictions on the growth of 112 the cwnd when a TCP sender is rate-limited. A rate-limited sender 113 may therefore grow a cwnd far beyond that corresponding to the 114 current transmit rate, resulting in a value that does not reflect 115 current information about the state of the network path the flow is 116 using. Use of such an invalid cwnd may result in reduced application 117 performance and/or could significantly contribute to network 118 congestion. 120 [RFC2861] proposed a solution to these issues in an experimental 121 method known as Congestion Window Validation (CWV). CWV was intended 122 to help reduce cases where TCP accumulated an invalid cwnd. The use 123 and drawbacks of using CWV with an application are discussed in 124 Section 2. 126 Section 3 defines relevant terminology. 128 Section 4 specifies an alternative to CWV that seeks to address the 129 same issues, but does this in a way that is expected to mitigate the 130 impact on an application that varies its sending rate. The method 131 described applies to both a rate-limited and an idle condition. 133 2. Reviewing experience with TCP-CWV 135 RFC 2861 described a simple modification to the TCP congestion 136 control algorithm that decayed the cwnd after the transition to a 137 "sufficiently-long" idle period. This used the slow-start threshold 138 (ssthresh) to save information about the previous value of the 139 congestion window. The approach relaxed the standard TCP behaviour 140 [RFC5681] for an idle session, intended to improve application 141 performance. CWV also modified the behaviour for a rate-limited 142 session where a sender transmitted at a rate less than allowed by 143 cwnd. 145 RFC 2861 has been implemented in some mainstream operating systems as 146 the default behaviour [Bis08]. Analysis (e.g. [Bis10] [Fai12]) has 147 shown that a TCP sender using CWV is able to use available capacity 148 on a shared path after an idle period. This can benefit some 149 applications, especially over long delay paths, when compared to the 150 slow-start restart specified by standard TCP. However, CWV would 151 only benefit an application if the idle period were less than several 152 Retransmission Time Out (RTO) intervals [RFC6298], since the 153 behaviour would otherwise be the same as for standard TCP, which 154 resets the cwnd to the RTCP Restart Window (RW) after this period. 156 Experience with CWV suggests that although CWV benefits the network 157 in a rate-limited scenario (reducing the probability of network 158 congestion), the behaviour can be too conservative for many common 159 rate-limited applications. This mechanism does not therefore offer 160 the desirable increase in application performance for rate-limited 161 applications and it is unclear whether applications actually use this 162 mechanism in the general Internet. 164 It is therefore concluded that CWV is often a poor solution for many 165 rate-limited applications. It has the correct motivation, but has 166 the wrong approach to solving this problem. 168 3. Terminology 170 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 171 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 172 document are to be interpreted as described in [RFC2119]. 174 The document assumes familiarity with the terminology of TCP 175 congestion control [RFC5681]. 177 The following new terminology is introduced: 179 Validated phase: The phase where the cwnd reflects a current estimate 180 of the available path capacity. 182 Non-validated phase: The phase where the cwnd reflects a previous 183 measurement of the available path capacity. 185 Non-validated period, NVP: The maximum period for which cwnd is 186 preserved in the non-validated phase. 188 Rate-limited: A TCP flow that does not consume more than one half of 189 cwnd, and hence operates in the non-validated phase. 191 pipe ACK: The measured volume of data that was acknowledged by the 192 network per RTT. 194 4. An updated TCP response to idle and application-limited periods 196 This section proposes an update to the TCP congestion control 197 behaviour during an idle or rate-limited period. The new method 198 permits a TCP sender to preserve the cwnd when an application becomes 199 idle for a period of time (to be known as the non-validated period, 200 NVP, see section 5). The period where actual usage is less than 201 allowed by cwnd, is named as the non-validated phase. This method 202 allows an application to resume transmission at a previous rate 203 without incurring the delay of slow-start. However, if the TCP 204 sender experiences congestion using the preserved cwnd, it is 205 required to immediately reset the cwnd to an appropriate value 206 specified by the method. If a sender does not take advantage of the 207 preserved cwnd within the NVP, the value of cwnd is reduced, ensuring 208 the value better reflects the capacity that was recently actually 209 used. 211 The method requires that the TCP SACK option [RFC3517]is enabled. 212 This allows the sender to select an appropriate value for the cwnd 213 following a congestion event that is based on the measured path 214 capacity, and better reflects the fair-share. A similar approach was 215 proposed by TCP Jump Start [Liu07], as a congestion response after 216 more rapid opening of a TCP connection. 218 It is expected that this update will satisfy the requirements of many 219 rate-limited applications and at the same time provide an appropriate 220 method for use in the Internet. It also reduces the incentive for an 221 application to send data simply to keep transport congestion state. 222 (This is sometimes known as "padding"). 224 The new method does not differentiate between times when the sender 225 has become idle or rate-limited. This is partly a response to 226 recognition that some applications wish to transmit at a rate less 227 than allowed by the sender cwnd, and that it can be hard to make a 228 distinction between rate-limited and idle behaviour. This is 229 expected to encourage applications and TCP stacks to use standards- 230 based congestion control methods. It may also encourage the use of 231 long-lived connections where this offers benefit (such as persistent 232 http). 234 The method is specified in following subsections. 236 4.1. A method for preserving cwnd during the idle and application- 237 limited periods. 239 The method described in this document updates [RFC5681]. Use of the 240 method REQUIRES a TCP sender and the corresponding receiver to enable 241 the TCP SACK option [RFC3517]. 243 [RFC5681] defines a variable, FlightSize, that indicates the amount 244 of outstanding data in the network. This is assumed to be equal to 245 the value of Pipe calculated based on the pipe algorithm [RFC3517]. 246 In RFC5681 this value is used during loss recovery, whereas in this 247 method a new variable "pipeACK" is introduced and used to determine 248 if the sender has validated the cwnd. 250 The value of pipeACK is initialised to the maxium value. This value 251 is used to inhibt entering the nonvalidated phase until the first 252 measurement of pipeACK completes. 254 A sender is not required to continuously track the pipeACK value, but 255 MUST set this variable to the volume of data that was acknowledged by 256 the network per measured Round Trip Time (RTT), with a sampling 257 period of not less than one measurement for Min(RTT, 1 second). 258 Using the variables defined in [RFC3517]. This could be implemented 259 by caching the value of HighACK and after one RTT assigning pipeACK 260 to the difference between the cached HighACK value and the current 261 HighACK value. Other equivalent methods may be used. 263 4.2. The nonvalidated phase 265 The updated method creates a new TCP sender phase that captures 266 whether the cwnd reflects a validated or non-validated value. The 267 phases are defined as: 269 o Validated phase: pipeACK >=(1/2)*cwnd. This is the normal phase, 270 where cwnd is expected to be an approximate indication of the 271 available capacity currently available along the network path, and 272 the standard methods are used to increase cwnd (currently 273 [RFC5681]). The rule for transitioning to the non-validated phase 274 is specified in section 4.3. 276 o Non-validated phase: pipeACK <(1/2)*cwnd. This is the phase where 277 the cwnd has a value based on a previous measurement of the 278 available capacity, and the usage of this capacity has not been 279 validated in the previous RTT. That is, when it is not known 280 whether the cwnd reflects the currently available capacity along 281 the network path. The mechanisms to be used in this phase seek to 282 determine a safe value for cwnd and an appropriate reaction to 283 congestion. These mechanisms are specified in section 4.3. 285 A sender starts a TCP connection in the Validated phase. 287 The value 1/2 was selected to reduce the effects of variations in the 288 measured pipeACK, and to allow the sender some flexibility in when it 289 sends data. 291 4.3. TCP congestion control during the nonvalidated phase 293 A TCP sender MUST enter the non-validated phase when the measured 294 pipeACK is less than (1/2)*cwnd. 296 A TCP sender that enters the non-validated phase will preserve the 297 cwnd (i.e., this neither grows nor reduces while the sender remains 298 in this phase). The phase is concluded after a fixed period of time 299 (the NVP, as explained in section 4.3.2) or when the sender transmits 300 sufficient data so that pipeACK > (1/2)*cwnd (i.e. it is no longer 301 rate-limited). 303 The behaviour in the non-validated phase is specified as: 305 o The cwnd is not increased when ACK packets are received in this 306 phase. 308 o If the sender receives an indication of congestion while in the 309 non-validated phase (i.e. detects loss, or an Explicit Congestion 310 Notification, ECN, mark [RFC3168]), the sender MUST exit the non- 311 validated phase (reducing the cwnd as defined in section 4.3.1). 313 o If the Retransmission Time Out (RTO) expires while in the non- 314 validated phase, the sender MUST exit the non-validated phase. It 315 then resumes using the Standard TCP RTO mechanism [RFC5681]. (The 316 resulting reduction of cwnd described in section 4.3.2 is 317 appropriate, since any accumulated path history is considered 318 unreliable). 320 o A sender that measures a pipeACK greater than (1/2)*cwnd SHOULD 321 enter the validated phase. (A rate-limited sender will not 322 normally be impacted by whether it is in a validated or non- 323 validate phase, since it will normally not consume the entire 324 cwnd. However a change to the validated phase will release the 325 sender from constraints on the growth of cwnd, and restore the use 326 of the standard congestion response.) 328 4.3.1. Response to congestion in the nonvalidated phase 330 Reception of congestion feedback while in the non-validated phase is 331 interpreted as an indication that it was inappropriate for the sender 332 to use the preserved cwnd. The sender is therefore required to 333 quickly reduce the rate to avoid further congestion. Since the cwnd 334 does not have a validated value, a new cwnd value must be selected 335 based on the utilised rate. 337 A sender that detects a packet-drop or receives an ECN marked packet 338 MUST calculate a safe cwnd, by setting it to the value specified in 339 Section 3.2 of [RFC5681]. 341 At the end of the recovery phase, the TCP sender MUST reset the cwnd 342 using the method below: 343 cwnd = ((FlightSize - R)/2). 345 Where, R is the volume of data that was reported as unacknowledged by 346 the SACK information. This follows the method proposed for Jump 347 Start [Liu07]. 349 The inclusion of the term R makes this adjustment more conservative 350 than standard TCP. (This is required, since the sender may have sent 351 more segments than a Standard TCP sender would have done. The 352 additional reduction is beneficial when the FlightSize significantly 353 overshoots the available path capacity incurring significant loss, 354 for instance an intense traffic burst following a non-validated 355 period.) 357 If the sender implements a method that allows it to identify the 358 number of ECN-marked segments within a window that were observed by 359 the receiver, the sender SHOULD use the method above, further 360 reducing R by the number of marked segments. 362 The sender MUST also re-initialise the pipeACK variable to the maxium 363 value. This ensures that standard TCP methods are used immediately 364 after completing loss recovery. 366 4.3.2. Adjustment at the end of the nonvalidated phase 368 During the non-validated phase, a sender can produce bursts of data 369 of up to the cwnd in size. While this is no different to standard 370 TCP, it is desirable to control the maximum burst size, e.g. by 371 setting a burst size limit, using a pacing algorithm, or some other 372 method [Hug01]. 374 An application that remains in the non-validated phase for a period 375 greater than the NVP is required to adjust its congestion control 376 state. If the sender exits the non-validated phase after this 377 period, it MUST update the ssthresh: 379 ssthresh = max(ssthresh, 3*cwnd/4). 381 (This adjustment of ssthresh ensures that the sender records that it 382 has safely sustained the present rate. The change is beneficial to 383 rate-limited flows that encounter occasional congestion, and could 384 otherwise suffer an unwanted additional delay in recovering the 385 sending rate.) 387 The sender MUST then update cwnd to be not greater than: 389 cwnd = max(1/2*cwnd, IW). 391 Where IW is the TCP inital window [RFC5681]. 393 (This adjustment ensures that sender responds conservatively at the 394 end of the non-validated phase by reducing the cwnd to better reflect 395 the current sending rate of the sender. The cwnd update does not 396 take into account FlightSize or pipeACK because these values only 397 reflect data during the last RTT and do not reflect the average or 398 peak sending rate.) 400 After completing this adjustment, the sender MAY re-enter the non- 401 validated phase, if required (see section 4.2). 403 5. Determining a safe period to preserve cwnd 405 This section documents the rationale for selecting the maximum period 406 that cwnd may be preserved, known as the non-validated period, NVP. 408 Limiting the period that cwnd may be preserved avoids undesirable 409 side effects that would result if the cwnd were to be kept 410 unecessarily high for an arbitrary long period, which was a part of 411 the problem that CWV originally attempted to address. The period a 412 sender may safely preserve the cwnd, is a function of the period that 413 a network path is expected to sustain the capacity reflected by cwnd. 414 There is no ideal choice for this time. 416 A period of five minutes was chosen for this NVP. This is a 417 compromise that was larger than the idle intervals of common 418 applications, but not sufficiently larger than the period for which 419 the capacity of an Internet path may commonly be regarded as stable. 420 The capacity of wired networks is usually relatively stable for 421 periods of several minutes and that load stability increases with the 422 capacity. This suggests that cwnd may be preserved for at least a 423 few minutes. 425 There are cases where the TCP throughput exhibits significant 426 variability over a time less than five minutes. Examples could 427 include wireless topologies, where TCP rate variations may fluctuate 428 on the order of a few seconds as a consequence of medium access 429 protocol instabilities. Mobility changes may also impact TCP 430 performance over short time scales. Senders that observe such rapid 431 changes in the path characteristic may also experience increased 432 congestion with the new method, however such variation would likely 433 also impact TCP's behaviour when supporting interactive and bulk 434 applications. 436 Routing algorithms may modify the network path, disrupting the RTT 437 measurement and changing the capacity available to a TCP connection, 438 however such changes do not often occur within a time frame of a few 439 minutes. 441 The value of five minutes is therefore expected to be sufficient for 442 most current applications. Simulation studies (e.g. [Bis11]) also 443 suggest that for many practical applications, the performance using 444 this value will not be significantly different to that observed using 445 a non-standard method that does not reset the cwnd after idle. 447 Finally, other TCP sender mechanisms have used a 5 minute timer, and 448 there could be simplifications in some implementations by reusing the 449 same interval. TCP defines a default user timeout of 5 minutes 450 [RFC0793] i.e. how long transmitted data may remain unacknowledged 451 before a connection is forcefully closed. 453 6. Security Considerations 455 General security considerations concerning TCP congestion control are 456 discussed in [RFC5681]. This document describes an algorithm that 457 updates one aspect of the congestion control procedures, and so the 458 considerations described in RFC 5681 also apply to this algorithm. 460 7. IANA Considerations 462 There are no IANA considerations. 464 8. Acknowledgments 466 The authors acknowledge the contributions of Dr I Biswas and Dr R 467 Secchi in supporting the evaluation of CWV and for their help in 468 developing the mechanisms proposed in this draft. We also 469 acknowledge comments received from the Internet Congestion Control 470 Research Group, in particular Yuchung Cheng, Mirja Kuehlewind, and 471 Joe Touch. 473 9. Author Notes 475 9.1. Other related work 477 There are several issues to be discussed more widely: 479 o Should the method explicitly state a procedure for limiting 480 burstiness or pacing? 482 This is often regarded as good practice, but is not presently a 483 formal part of TCP. draft-hughes-restart-00.txt provides some 484 discussion of this topic. 486 o There are potential interactions with the proposal to raise the 487 TCP initial Window to ten segments, do these cases need to be 488 elaborated? 490 This relates to draft-ietf-tcpm-initcwnd. 492 The two methods have different functions and different response 493 to loss/congestion. 495 IW=10 proposes an experimental update to TCP that would allow 496 faster opening of the cwnd, and also a large (same size) 497 restart window. This approach is based on the assumption that 498 many forward paths can sustain bursts of up to ten segments 499 without (appreciable) loss. Such a significant increase in 500 cwnd must be matched with an equally large reduction of cwnd if 501 loss/congestion is detected, and such a congestion indication 502 is likely to require future use of IW=10 to be disabled for 503 this path for some time. This guards against the unwanted 504 behaviour of a series of short flows continuously flooding a 505 network path without network congestion feedback. 507 In contrast, new-CWV proposes a standards-track update with a 508 rationale that relies on recent previous path history to select 509 an appropriate cwnd after restart. 511 The behaviour differs in three ways: 513 1) For applications that send little initially, new-cwv may 514 constrain more than IW=10, but would not require the connection 515 to reset any path information when a restart incurred loss. In 516 contrast, new-cwv would allow the TCP connection to preserve 517 the cached cwnd, any loss, would impact cwnd, but not impact 518 other flows. 520 2) For applications that utilise more capacity than provided by 521 a cwnd=10, this method would permit a larger restart window 522 compared to a restart using IW=10. This is justified by the 523 recent path history. 525 3) new-CWV is attended to also be used for rate-limited 526 applications, where the application sends, but does not seek to 527 fully utilise the cwnd. In this case, new-cwv constrains the 528 cwnd to that justified by the recent path history. The 529 performance trade-offs are hence different, and it would be 530 possible to enable new-cwv when also using IW=10, and yield the 531 benefits of this. 533 o There is potential overlap with the Laminar proposal 534 (draft-mathis-tcpm-tcp-laminar) 536 The current draft was intended as a standards-track update to 537 TCP, rather than a new transport variant. At least, it would 538 be good to understand how the two interact and whether there is 539 a possibility of a single method. 541 o There is potential performance loss in loss of a short burst 542 (off list with M Allman) 544 A sender can transmit several segments then become idle. If 545 the first segments are all ACK'ed the ssthresh collapses to a 546 small value (no new data is sent by the idle sender). Loss of 547 the later data results in congestion (e.g. maybe a RED drop or 548 some other cause, rather than the peak rate of this flow). 549 When performs loss recovery it may have an appreciable pipeACK 550 and cwnd, but a very low flight size - the Standard algorithm 551 results in an unusually low cwnd (1/2 Flight size). 553 A constant rate flow would have maintained a flight size 554 appropriate to pipeACK (cwnd if it is a bulk flow). 556 This could be fixed by adding a new state variable? It could 557 also be argued this is a corner case (e.g. loss of only the 558 last segments would have resulted in RTO), the impact could be 559 significant. 561 9.2. Revision notes 563 RFC-Editor note: please remove this section prior to publication. 565 Draft 03 was submitted to ICCRG to receive comments and feedback. 567 Draft 04 contained the first set of clarifications after feedback: 569 o Changed name to application limited and used the term rate-limited 570 in all places. 572 o Added justification and many minor changes suggested on the list. 574 o Added text to tie-in with more accurate ECN marking. 576 o Added ref to Hug01 578 Draft 05 contained various updates: 580 o New text to redefine how to measure the acknowledged pipe, 581 differentiating this from the FlightSize, and hence avoiding 582 previous issues with infrequent large bursts of data not being 583 validated. A key point new feature is that pipeACK only triggers 584 leaving the NVP after the size of the pipe has been acknowledged. 585 This removed the need for hysteresis. 587 o Reduction values were changed to 1/2, following analysis of 588 suggestions from ICCRG. This also sets the "target" cwnd as twice 589 the used rate for non-validated case. 591 o Introduced a symbolic name (NVP) to denote the 5 minute period. 593 Draft 06 contained various updates: 595 o Required reset of pipeACK after congestion. 597 o Added comment on the effect of congestion after a short burst (M. 598 Allman). 600 o Correction of minor Typos. 602 WG draft 01 contained various updates: 604 o Updaed initialisation of pipeACK to maximum value. 606 o Added note on intended status still to be determined. 608 10. References 610 10.1. Normative References 612 [RFC0793] Postel, J., "Transmission Control Protocol", STD 7, 613 RFC 793, September 1981. 615 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 616 Requirement Levels", BCP 14, RFC 2119, March 1997. 618 [RFC2861] Handley, M., Padhye, J., and S. Floyd, "TCP Congestion 619 Window Validation", RFC 2861, June 2000. 621 [RFC3168] Ramakrishnan, K., Floyd, S., and D. Black, "The Addition 622 of Explicit Congestion Notification (ECN) to IP", 623 RFC 3168, September 2001. 625 [RFC3517] Blanton, E., Allman, M., Fall, K., and L. Wang, "A 626 Conservative Selective Acknowledgment (SACK)-based Loss 627 Recovery Algorithm for TCP", RFC 3517, April 2003. 629 [RFC5681] Allman, M., Paxson, V., and E. Blanton, "TCP Congestion 630 Control", RFC 5681, September 2009. 632 [RFC6298] Paxson, V., Allman, M., Chu, J., and M. Sargent, 633 "Computing TCP's Retransmission Timer", RFC 6298, 634 June 2011. 636 10.2. Informative References 638 [Bis08] Biswas and Fairhurst, "A Practical Evaluation of 639 Congestion Window Validation Behaviour, 9th Annual 640 Postgraduate Symposium in the Convergence of 641 Telecommunications, Networking and Broadcasting (PGNet), 642 Liverpool, UK", June 2008. 644 [Bis10] Biswas, Sathiaseelan, Secchi, and Fairhurst, "Analysing 645 TCP for Bursty Traffic, Int'l J. of Communications, 646 Network and System Sciences, 7(3)", June 2010. 648 [Bis11] Biswas, "PhD Thesis, Internet congestion control for 649 variable rate TCP traffic, School of Engineering, 650 University of Aberdeen", June 2011. 652 [Fai12] Fairhurst, Biswas, Biswas, and Biswas, "Enhancing TCP 653 Performance to support Variable-Rate Traffic, 2nd Capacity 654 Sharing Workshop, ACM CoNEXT, Nice, France, 10th December 655 2012.", June 2008. 657 [Hug01] Hughes, Touch, and Heidemann, "√√Issues in TCP 658 Slow-Start Restart After Idle (Work-in-Progress)", 659 December 2001. 661 [Liu07] Liu, Allman, Jiny, and Wang, "Congestion Control without a 662 Startup Phase, 5th International Workshop on Protocols for 663 Fast Long-Distance Networks (PFLDnet), Los Angeles, 664 California, USA", February 2007. 666 Authors' Addresses 668 Godred Fairhurst 669 University of Aberdeen 670 School of Engineering 671 Fraser Noble Building 672 Aberdeen, Scotland AB24 3UE 673 UK 675 Email: gorry@erg.abdn.ac.uk 676 URI: http://www.erg.abdn.ac.uk 678 Arjuna Sathiaseelan 679 University of Aberdeen 680 School of Engineering 681 Fraser Noble Building 682 Aberdeen, Scotland AB24 3UE 683 UK 685 Email: arjuna@erg.abdn.ac.uk 686 URI: http://www.erg.abdn.ac.uk