idnits 2.17.1 draft-ietf-tcpm-newcwv-09.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. Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year -- The document date (March 6, 2015) is 3310 days in the past. Is this intentional? Checking references for intended status: Experimental ---------------------------------------------------------------------------- ** Obsolete normative reference: RFC 793 (Obsoleted by RFC 9293) ** Obsolete normative reference: RFC 2861 (Obsoleted by RFC 7661) -- Obsolete informational reference (is this intentional?): RFC 2616 (Obsoleted by RFC 7230, RFC 7231, RFC 7232, RFC 7233, RFC 7234, RFC 7235) Summary: 2 errors (**), 0 flaws (~~), 1 warning (==), 3 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) R. Secchi 5 Intended status: Experimental University of Aberdeen 6 Expires: September 7, 2015 March 6, 2015 8 Updating TCP to support Rate-Limited Traffic 9 draft-ietf-tcpm-newcwv-09 11 Abstract 13 This document provides a mechanism to address issues that arise when 14 TCP is used to support traffic that exhibits periods where the 15 sending rate is limited by the application rather than the congestion 16 window. It provides an experimental update to TCP that allows a TCP 17 sender to restart quickly following a rate-limited interval. This 18 method is expected to benefit applications that send rate-limited 19 traffic using TCP, while also providing an appropriate response if 20 congestion is experienced. 22 It also evaluates the Experimental specification of TCP Congestion 23 Window Validation, CWV, defined in RFC 2861, and concludes that RFC 24 2861 sought to address important issues, but failed to deliver a 25 widely used solution. This document therefore recommends that the 26 status of RFC 2861 is moved from Experimental to Historic, and that 27 it is replaced by the current specification. 29 Status of This Memo 31 This Internet-Draft is submitted in full conformance with the 32 provisions of BCP 78 and BCP 79. 34 Internet-Drafts are working documents of the Internet Engineering 35 Task Force (IETF). Note that other groups may also distribute 36 working documents as Internet-Drafts. The list of current Internet- 37 Drafts is at http://datatracker.ietf.org/drafts/current/. 39 Internet-Drafts are draft documents valid for a maximum of six months 40 and may be updated, replaced, or obsoleted by other documents at any 41 time. It is inappropriate to use Internet-Drafts as reference 42 material or to cite them other than as "work in progress." 44 This Internet-Draft will expire on September 7, 2015. 46 Copyright Notice 48 Copyright (c) 2015 IETF Trust and the persons identified as the 49 document authors. All rights reserved. 51 This document is subject to BCP 78 and the IETF Trust's Legal 52 Provisions Relating to IETF Documents 53 (http://trustee.ietf.org/license-info) in effect on the date of 54 publication of this document. Please review these documents 55 carefully, as they describe your rights and restrictions with respect 56 to this document. Code Components extracted from this document must 57 include Simplified BSD License text as described in Section 4.e of 58 the Trust Legal Provisions and are provided without warranty as 59 described in the Simplified BSD License. 61 Table of Contents 63 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 2 64 1.1. Standards Status of this Document . . . . . . . . . . . . 4 65 2. Reviewing experience with TCP-CWV . . . . . . . . . . . . . . 5 66 3. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . 6 67 4.1. Initialisation . . . . . . . . . . . . . . . . . . . . . 8 68 4.2. Estimating the validated capacity supported by a path . . 8 69 4.3. Preserving cwnd during a rate-limited period. . . . . . . 9 70 4.4. TCP congestion control during the non-validated phase . . 9 71 4.4.1. Response to congestion in the non-validated phase . . 10 72 4.4.2. Sender burst control during the non-validated phase . 12 73 4.4.3. Adjustment at the end of the non-validated phase . . 12 74 4.5. Examples of Implementation . . . . . . . . . . . . . . . 13 75 4.5.1. Implementing the pipeACK measurement . . . . . . . . 13 76 4.5.2. Implementing detection of the cwnd-limited condition 14 77 5. Determining a safe period to preserve cwnd . . . . . . . . . 15 78 6. Security Considerations . . . . . . . . . . . . . . . . . . . 16 79 7. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 16 80 8. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . 16 81 9. Author Notes . . . . . . . . . . . . . . . . . . . . . . . . 16 82 9.1. Other related work . . . . . . . . . . . . . . . . . . . 16 83 10. Revision notes . . . . . . . . . . . . . . . . . . . . . . . 18 84 11. References . . . . . . . . . . . . . . . . . . . . . . . . . 21 85 11.1. Normative References . . . . . . . . . . . . . . . . . . 21 86 11.2. Informative References . . . . . . . . . . . . . . . . . 22 87 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 23 89 1. Introduction 91 TCP is used to support a range of application behaviours. The TCP 92 congestion window (cwnd) controls the number of unacknowledged 93 packets/bytes that a TCP flow may have in the network at any time, a 94 value known as the FlightSize [RFC5681]. A bulk application will 95 always have data available to transmit. The rate at which it sends 96 is therefore limited by the maximum permitted by the receiver 97 advertised window and the sender congestion window (cwnd). In 98 contrast, a rate-limited application will experience periods when the 99 sender is either idle or is unable to send at the maximum rate 100 permitted by the cwnd. The update in this document targets the 101 operation of TCP in such rate-limited cases. 103 Standard TCP [RFC5681] states that a TCP sender SHOULD set cwnd to no 104 more than the Restart Window (RW) before beginning transmission, if 105 the TCP sender has not sent data in an interval exceeding the 106 retransmission timeout, i.e., when an application becomes idle. 107 [RFC2861] noted that this TCP behaviour was not always observed in 108 current implementations. Experiments [Bis08] confirm this to still 109 be the case. 111 Congestion Window Validation, CWV, introduced the terminology of 112 "application limited periods". This document describes any time that 113 an application limits the sending rate, rather than being limited by 114 the transport, as "rate-limited". This update improves support for 115 applications that vary their transmission rate, either with (short) 116 idle periods between transmission or by changing the rate the 117 application sends. These applications are characterised by the TCP 118 FlightSize often being less than cwnd. Many Internet applications 119 exhibit this behaviour, including web browsing, http-based adaptive 120 streaming, applications that support query/response type protocols, 121 network file sharing, and live video transmission. Many such 122 applications currently avoid using long-lived (persistent) TCP 123 connections (e.g. [RFC2616] servers typically support persistent 124 HTTP connections, but do not enable this by default). Such 125 applications often instead either use a succession of short TCP 126 transfers or use UDP. 128 Standard TCP does not impose additional restrictions on the growth of 129 the congestion window when a TCP sender is unable to send at the 130 maximum rate allowed by the cwnd. In this case the rate-limited 131 sender may grow a cwnd far beyond that corresponding to the current 132 transmit rate, resulting in a value that does not reflect current 133 information about the state of the network path the flow is using. 134 Use of such an invalid cwnd may result in reduced application 135 performance and/or could significantly contribute to network 136 congestion. 138 [RFC2861] proposed a solution to these issues in an experimental 139 method known as CWV. CWV was intended to help reduce cases where TCP 140 accumulated an invalid (inappropriately large) cwnd. The use and 141 drawbacks of using the CWV algorithm in RFC 2861 with an application 142 are discussed in Section 2. 144 Section 3 defines relevant terminology. 146 Section 4 specifies an alternative to CWV that seeks to address the 147 same issues, but does this in a way that is expected to mitigate the 148 impact on an application that varies its sending rate. The updated 149 method applies to the rate-limited conditions (including both an 150 application-limited and idle sender). 152 The goals of this update are: 154 o To not change the behaviour of a TCP sender that performs bulk 155 transfers that consume the cwnd. 157 o To provide a method that co-exists with Standard TCP and other 158 flows that use this updated method. 160 o To reduce transfer latency for applications that change their rate 161 over short intervals of time. 163 o To avoid a TCP sender growing a large "non-validated" cwnd, when 164 it has not recently sent using this cwnd. 166 o To remove the incentive for ad-hoc application or network stack 167 methods (such as "padding") solely to maintain a large cwnd for 168 future transmission. 170 o To incentivise the use of long-lived connections, rather than a 171 succession of short-lived flows, benefiting both flows and network 172 when actual congestion is encountered. 174 Section 5 describes the rationale for selecting the safe period to 175 preserve the cwnd. 177 1.1. Standards Status of this Document 179 This document was produced by the TCP Maintenance and Minor 180 Extensions (tcpm) working group. 182 The document updates and obsoletes the methods described in 183 [RFC2861]. It recommends a set of mechanisms, including the use of 184 pacing during a non-validated period. The updated mechanisms are 185 intended to have a less aggressive congestion impact than would be 186 exhibited by a standard TCP sender. 188 The specification in this draft is classified as "Experimental" 189 pending experience with deployed implementations of the methods. 191 2. Reviewing experience with TCP-CWV 193 [RFC2861] described a simple modification to the TCP congestion 194 control algorithm that decayed the cwnd after the transition to a 195 "sufficiently-long" idle period. This used the slow-start threshold 196 (ssthresh) to save information about the previous value of the 197 congestion window. The approach relaxed the standard TCP behaviour 198 [RFC5681] for an idle session, intended to improve application 199 performance. CWV also modified the behaviour where a sender 200 transmitted at a rate less than allowed by cwnd. 202 [RFC2861] proposed two set of responses, one after an "application- 203 limited" and one after an "idle period". Although this distinction 204 was argued, in practice differentiating the two conditions was found 205 problematic in actual networks (e.g.[Bis10]). This offers 206 predictable performance for long on-off periods (>>1 RTT), or slowly 207 varying rate-based traffic, the performance could be unpredictable 208 for variable-rate traffic and depended both upon whether an accurate 209 RTT had been obtained and the pattern of application traffic relative 210 to the measured RTT. 212 Many applications can and often do vary their transmission over a 213 wide range of rates. Using [RFC2861] such applications often 214 experienced varying performance, which made it hard for application 215 developers to predict the TCP latency even when using a path with 216 stable network characteristics. We argue that an attempt to classify 217 application behaviour as application-limited or idle is problematic 218 and also inappropriate. This document therefore explicitly avoids 219 trying to differentiate these two cases, instead treating all rate- 220 limited traffic uniformly. 222 [RFC2861] has been implemented in some mainstream operating systems 223 as the default behaviour [Bis08]. Analysis (e.g. [Bis10] [Fai12]) 224 has shown that a TCP sender using CWV is able to use available 225 capacity on a shared path after an idle period. This can benefit 226 variable-rate applications, especially over long delay paths, when 227 compared to the slow-start restart specified by standard TCP. 228 However, CWV would only benefit an application if the idle period 229 were less than several Retransmission Time Out (RTO) intervals 230 [RFC6298], since the behaviour would otherwise be the same as for 231 standard TCP, which resets the cwnd to the TCP Restart Window after 232 this period. 234 To enable better performance for variable-rate applications with TCP, 235 some operating systems have chosen to support non-standard methods, 236 or applications have resorted to "padding" streams by sending dummy 237 data to maintain their sending rate when they have no data to 238 transmit. Although transmitting redundant data across a network path 239 provides good evidence that the path can sustain data at the offered 240 rate, padding also consumes network capacity and reduces the 241 opportunity for congestion-free statistical multiplexing. For 242 variable-rate flows, the benefits of statistical multiplexing can be 243 significant and it is therefore a goal to find a viable alternative 244 to padding streams. 246 Experience with [RFC2861] suggests that although the CWV method 247 benefited the network in a rate-limited scenario (reducing the 248 probability of network congestion), the behaviour was too 249 conservative for many common rate-limited applications. This 250 mechanism did not therefore offer the desirable increase in 251 application performance for rate-limited applications and it is 252 unclear whether applications actually use this mechanism in the 253 general Internet. 255 It is therefore concluded that CWV, as defined in [RFC2861], was 256 often a poor solution for many rate-limited applications. It had the 257 correct motivation, but had the wrong approach to solving this 258 problem. 260 3. Terminology 262 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 263 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 264 document are to be interpreted as described in [RFC2119]. 266 The document assumes familiarity with the terminology of TCP 267 congestion control [RFC5681]. 269 The following terminology is used in this document: 271 cwnd-limited: A TCP flow that has sent the maximum number of segments 272 permitted by the cwnd, where the application utilises the allowed 273 sending rate (see Section 4.5.2). 275 pipeACK sample: A measure of the volume of data acknowledged by the 276 network within an RTT. 278 pipeACK variable: A variable that measures the available capacity 279 using the set of pipeACK samples. 281 pipeACK Sampling Period: The maximum period that a measured pipeACK 282 sample may influence the pipeACK variable. 284 Non-validated phase: The phase where the cwnd reflects a previous 285 measurement of the available path capacity. 287 Non-validated period, NVP: The maximum period for which cwnd is 288 preserved in the non-validated phase. 290 Rate-limited: A TCP flow that does not consume more than one half of 291 cwnd, and hence operates in the non-validated phase. This includes 292 periods when an application is either idle or chooses to send at a 293 rate less than the maximum permitted by the cwnd. 295 Validated phase: The phase where the cwnd reflects a current estimate 296 of the available path capacity. 298 4. A New Congestion Window Validation method 300 This section proposes an update to the TCP congestion control 301 behaviour during a rate-limited interval. This new method 302 intentionally does not differentiate between times when the sender 303 has become idle or chooses to send at a rate less than the maximum 304 allowed by the cwnd. 306 The period where actual usage is less than allowed by cwnd, is named 307 the non-validated phase. The update allows an application in the 308 non-validated phase to resume transmission at a previous rate without 309 incurring the delay of slow-start. However, if the TCP sender 310 experiences congestion using the preserved cwnd, it is required to 311 immediately reset the cwnd to an appropriate value specified by the 312 method. If a sender does not take advantage of the preserved cwnd 313 within the Non-validated period, NVP, the value of cwnd is reduced, 314 ensuring the value better reflects the capacity that was recently 315 actually used. 317 It is expected that this update will satisfy the requirements of many 318 rate-limited applications and at the same time provide an appropriate 319 method for use in the Internet. New-CWV reduces this incentive for 320 an application to send "padding" data simply to keep transport 321 congestion state. 323 The method is specified in following subsections and is expected to 324 encourage applications and TCP stacks to use standards-based 325 congestion control methods. It may also encourage the use of long- 326 lived connections where this offers benefit (such as persistent 327 http). 329 4.1. Initialisation 331 A sender starts a TCP connection in the validated phase and 332 initialises the pipeACK variable to the "undefined" value. This 333 value inhibits use of the value in cwnd calculations. 335 4.2. Estimating the validated capacity supported by a path 337 [RFC6675] defines a variable, FlightSize, that indicates the 338 instantaneous amount of data that has been sent, but not cumulatively 339 acknowledged. In this method a new variable "pipeACK" is introduced 340 to measure the acknowledged size of the network pipe. This is used 341 to determine if the sender has validated the cwnd. pipeACK differs 342 from FlightSize in that it is evaluated over a window of acknowledged 343 data, rather than reflecting the amount of data outstanding. 345 A sender determines a pipeACK sample by measuring the volume of data 346 that was acknowledged by the network over the period of a measured 347 Round Trip Time (RTT). Using the variables defined in [RFC6675], a 348 value could be measured by caching the value of HighACK and after one 349 RTT measuring the difference between the cached HighACK value and the 350 current HighACK value. Other equivalent methods may be used. 352 A sender is not required to continuously update the pipeACK variable 353 after each received ACK, but SHOULD perform a pipeACK sample at least 354 once per RTT when it has sent unacknowledged segments. 356 The pipeACK variable MAY consider multiple pipeACK samples over the 357 pipeACK Sampling Period. The value of the pipeACK variable MUST NOT 358 exceed the maximum (highest value) within the sampling period. This 359 specification defines the pipeACK Sampling Period as Max(3*RTT, 1 360 second). This period enables a sender to compensate for large 361 fluctuations in the sending rate, where there may be pauses in 362 transmission, and allows the pipeACK variable to reflect the largest 363 recently measured pipeACK sample. 365 When no measurements are available, the pipeACK variable is set to 366 the "undefined value". This value is used to inhibit entering the 367 non-validated phase until the first new measurement of a pipeACK 368 sample. 370 The pipeACK variable MUST NOT be updated during TCP Fast Recovery. 371 That is, the sender stops collecting pipeACK samples during loss 372 recovery. The method RECOMMENDS that the TCP SACK option [RFC2018] 373 is enabled and the method defined on [RFC6675]is used to recover 374 missing segments. This allows the sender to more accurately 375 determine the number of missing bytes during the loss recovery phase, 376 and using this method will result in a more appropriate cwnd 377 following loss. 379 4.3. Preserving cwnd during a rate-limited period. 381 The updated method creates a new TCP sender phase that captures 382 whether the cwnd reflects a validated or non-validated value. The 383 phases are defined as: 385 o Validated phase: pipeACK >=(1/2)*cwnd, or pipeACK is undefined. 386 This is the normal phase, where cwnd is expected to be an 387 approximate indication of the capacity currently available along 388 the network path, and the standard methods are used to increase 389 cwnd (currently [RFC5681]). 391 o Non-validated phase: pipeACK <(1/2)*cwnd. This is the phase where 392 the cwnd has a value based on a previous measurement of the 393 available capacity, and the usage of this capacity has not been 394 validated in the pipeACK Sampling Period. That is, when it is not 395 known whether the cwnd reflects the currently available capacity 396 along the network path. The mechanisms to be used in this phase 397 seek to determine a safe value for cwnd and an appropriate 398 reaction to congestion. 400 Note: A threshold is needed to determine whether a sender is in the 401 validated or non-validated phase. A standard TCP sender in slow- 402 start is permitted to double its FlightSize from one RTT to the next. 403 This motivated the choice of a threshold value of 1/2. This 404 threshold ensures a sender does not further increase the cwnd as long 405 as the FlightSize is less than (1/2*cwnd). Furthermore, a sender 406 with a FlightSize less than (1/2*cwnd) may in the next RTT be 407 permitted by the cwnd to send at a rate that more than doubles the 408 FlightSize, and hence this case needs to be regarded as non-validated 409 and a sender therefore needs to employ additional mechanisms while in 410 this phase. 412 4.4. TCP congestion control during the non-validated phase 414 A TCP sender MUST enter the non-validated phase when the pipeACK is 415 less than (1/2)*cwnd. 417 A TCP sender that enters the non-validated phase SHOULD preserve the 418 cwnd (i.e., this neither grows nor reduces while the sender remains 419 in this phase). If the sender receives an indication of congestion, 420 it uses the method described below. The phase is concluded after a 421 fixed period of time (the NVP, as explained in Section 4.4.3) or when 422 the sender transmits sufficient data so that pipeACK > (1/2)*cwnd 423 (i.e., the sender is no longer rate-limited). 425 The behaviour in the non-validated phase is specified as: 427 o A sender determines whether to increase the cwnd based upon 428 whether it is cwnd-limited (see Section 4.5.2): 430 o 432 * A sender that is cwnd-limited MAY use the standard TCP method 433 to increase cwnd (i.e., a TCP sender that fully utilises the 434 cwnd is permitted to increase cwnd each received ACK using 435 standard methods). 437 * A sender that is not cwnd-limited MUST NOT increase the cwnd 438 when ACK packets are received in this phase. 440 o If the sender receives an indication of congestion while in the 441 non-validated phase (i.e., detects loss), the sender MUST exit the 442 non-validated phase (reducing the cwnd as defined in 443 Section 4.4.1). 445 o If the Retransmission Time Out (RTO) expires while in the non- 446 validated phase, the sender MUST exit the non-validated phase. It 447 then resumes using the standard TCP RTO mechanism [RFC5681]. 449 o A sender with a pipeACK variable greater than (1/2)*cwnd SHOULD 450 enter the validated phase. (A rate-limited sender will not 451 normally be impacted by whether it is in a validated or non- 452 validated phase, since it will normally not consume the entire 453 cwnd. However a change to the validated phase will release the 454 sender from constraints on the growth of cwnd, and restore the use 455 of the standard congestion response.) 457 The cwnd-limited behaviour may be triggered during a transient 458 condition that occurs when a sender is in the non-validated phase and 459 receives an ACK that acknowledges received data, the cwnd was fully 460 utilised, and more data is awaiting transmission than may be sent 461 with the current cwnd. The sender is then allowed to use the 462 standard method to increase the cwnd. (Note, if the sender succeeds 463 in sending these new segments, the updated cwnd and pipeACK variables 464 will eventually result in a transition to the validated phase.) 466 4.4.1. Response to congestion in the non-validated phase 468 Reception of congestion feedback while in the non-validated phase is 469 interpreted as an indication that it was inappropriate for the sender 470 to use the preserved cwnd. The sender is therefore required to 471 quickly reduce the rate to avoid further congestion. Since the cwnd 472 does not have a validated value, a new cwnd value must be selected 473 based on the utilised rate. 475 A sender that detects a packet-drop MUST record the current 476 FlightSize in the variable LossFlightSize and MUST calculate a safe 477 cwnd for loss recovery using the method below: 479 cwnd = (Max(pipeACK,LossFlightSize))/2. 481 The pipeACK value is not updated during loss recoverySection 4.2. If 482 there is a valid pipeACK value, the new cwnd is adjusted to reflect 483 that a non-validated cwnd may be larger than the actual FlightSize, 484 or recently used FlightSize (recorded in pipeACK). The updated cwnd 485 therefore prevents overshoot by a sender significantly increasing its 486 transmission rate during the recovery period. 488 At the end of the recovery phase, the TCP sender MUST reset the cwnd 489 using the method below: 491 cwnd = (Max(pipeACK,LossFlightSize) - R)/2. 493 Where R is the volume of data that was successfully retransmitted 494 during the recovery phase. This counts segments retransmitted and 495 considered lost by the pipe estimation algorithm at the end of 496 recovery. It does not include the additional cost of multiple 497 retransmission of the same data. 499 The calculated cwnd value MUST NOT be reduced below 1 MSS. 501 After completing the loss recovery phase, the sender MUST re- 502 initialise the pipeACK variable to the "undefined" value. This 503 ensures that standard TCP methods are used immediately after 504 completing loss recovery until a new pipeACK value can be determined. 506 ssthresh is adjusted using the standard TCP method. 508 Note: The adjustment by reducing cwnd by the volume of data not sent 509 (R) follows the method proposed for Jump Start [Liu07]. The 510 inclusion of the term R makes the adjustment more conservative than 511 standard TCP. This is required, since a sender in the non-validated 512 state may increase the rate more than a standard TCP would have done 513 relative to what was sent in the last RTT (i.e., more than doubled 514 the number of segments in flight relative to what it sent in the last 515 RTT). The additional reduction after congestion is beneficial when 516 the LossFlightSize has significantly overshot the available path 517 capacity incurring significant loss (e.g. following a change of path 518 characteristics or when additional traffic has taken a larger share 519 of the network bottleneck during a period when the sender transmits 520 less). 522 Note: The pipeACK value is only valid during a non-validated phase, 523 and therefore does not exceed cwnd/2. If LossFlightSize and R were 524 small, then this can result in the final cwnd after loss recovery 525 being 1/4 of the cwnd on detection of congestion. This reduction is 526 conservative, and pipeACK is reset to undefined. Subsequent updates 527 to cwnd do not therefore reflect pipeACK history before any 528 congestion event. 530 4.4.2. Sender burst control during the non-validated phase 532 TCP congestion control allows a sender to accumulate a cwnd that 533 would allow it to send a burst of segments with a total size up to 534 the difference between the FlightsSize and cwnd. Such bursts can 535 impact other flows that share a network bottleneck and/or may induce 536 congestion when buffering is limited. 538 Various methods have been proposed to control the sender burstiness 539 [Hug01], [All05]. For example, TCP can limit the number of new 540 segments it sends per received ACK. This is effective when a flow of 541 ACKs is received, but can not be used to control a sender that has 542 not send appreciable data in the previous RTT [All05]. 544 This document recommends using a method to avoid line-rate bursts 545 after an idle or rate-limited interval when there is less reliable 546 information about the capacity of the network path: A TCP sender in 547 the non-validated phase SHOULD control the maximum burst size, e.g. 548 using a rate-based pacing algorithm in which a sender paces out the 549 cwnd over its estimate of the RTT, or some other method, to prevent 550 many segments being transmitted contiguously at line-rate. The most 551 appropriate method(s) to implement pacing depend on the design of the 552 TCP/IP stack, speed of interface and whether hardware support (such 553 as TCP Segment Offload, TSO) is used. The present document does not 554 recommend any specific method. 556 4.4.3. Adjustment at the end of the non-validated phase 558 An application that remains in the non-validated phase for a period 559 greater than the NVP is required to adjust its congestion control 560 state. If the sender exits the non-validated phase after this 561 period, it MUST update the ssthresh: 563 ssthresh = max(ssthresh, 3*cwnd/4). 565 (This adjustment of ssthresh ensures that the sender records that it 566 has safely sustained the present rate. The change is beneficial to 567 rate-limited flows that encounter occasional congestion, and could 568 otherwise suffer an unwanted additional delay in recovering the 569 sending rate.) 571 The sender MUST then update cwnd to be not greater than: 573 cwnd = max((1/2)*cwnd, IW). 575 Where IW is the appropriate TCP initial window, used by the TCP 576 sender (e.g. [RFC5681]). 578 Note: This adjustment ensures that the sender responds conservatively 579 after remaining in the non-validated phase for more than the non- 580 validated period. In this case, it reduces the cwnd by a factor of 581 two from the preserved value. This adjustment is helpful when flows 582 accumulate but do not use a large cwnd, and seeks to mitigate the 583 impact when these flows later resume transmission. This could for 584 instance mitigate the impact if multiple high-rate application flows 585 were to become idle over an extended period of time and then were 586 simultaneously awakened by an external event. 588 4.5. Examples of Implementation 590 This section provides informative examples of implementation methods. 591 Implementations may choose to use other methods that comply with the 592 normative requirements. 594 4.5.1. Implementing the pipeACK measurement 596 A pipeACK sample may be measured once each RTT. This reduces the 597 sender processing burden for calculating after each acknowledgement 598 and also reduces storage requirements at the sender. 600 Since application behaviour can be bursty using CWV, it may be 601 desirable to implement a maximum filter to accumulate the measured 602 values so that the pipeACK variable records the largest pipeACK 603 sample within the pipeACK Sampling Period. One simple way to 604 implement this is to divide the pipeACK Sampling Period into several 605 (e.g. 5) equal length measurement periods. The sender then records 606 the start time for each measurement period and the highest measured 607 pipeACK sample. At the end of the measurement period, any 608 measurement(s) that are older than the pipeACK Sampling Period are 609 discarded. The pipeACK variable is then assigned the largest of the 610 set of the highest measured values. 612 +----------+----------+ +----------+---...... 613 | Sample A | Sample B | No | Sample C | Sample D 614 | | | Sample | | 615 | |\ 5 | | | | 616 | | | | | | /\ 4 | 617 | | | | |\ 3 | | | \ | 618 | | \ | | \--- | | / \ | /| 2 619 |/ \------| - | | / \------/ \... 620 +----------+---------\+----/ /----+/---------+-------------> Time 622 <------------------------------------------------| 623 Sampling Period Current Time 625 Figure 1: Example of measuring pipeACK samples 627 Figure 1 shows an example of how measurement samples may be 628 collected. At the time represented by the figure new samples are 629 being accumulated into sample D. Three previous samples also fall 630 within the pipeACK Sampling Period: A, B, and C. There was also a 631 period of inactivity between samples B and C during which no 632 measurements were taken. The current value of the pipeACK variable 633 will be 5, the maximum across all samples. 635 After one further measurement period, Sample A will be discarded, 636 since it then is older than the pipeACK Sampling Period and the 637 pipeACK variable will be recalculated, Its value will be the larger 638 of Sample C or the final value accumulated in Sample D. 640 Note: the pipeACK Sampling Period and the NVP period do not 641 necessarily require a new timer to be implemented. An alternative is 642 to record a timestamp when the sender enters the NVP. Each time a 643 sender transmits a new segment, this timestamp may be used to 644 determine if the NVP period has expired. If the period expires, the 645 sender may take into account how many units of the NVP period have 646 passed and make one reduction (as defined in Section 4.4.3) for each 647 NVP period. 649 4.5.2. Implementing detection of the cwnd-limited condition 651 A method is required to detect the cwnd-limited condition (see 652 Section 4.4. This is used to detect a condition where a sender in 653 the non-validated phase receives an ACK, but the size of cwnd 654 prevents sending more new data. 656 In simple terms this condition is true only when the TCP sender's 657 FlightSize is equal to or larger than the cwnd. However, an 658 implementation must consider other constraints on the way in which 659 cwnd variable is used, for instance the need to support methods such 660 as the Nagle Algorithm and TCP Segment Offload (TSO). This can 661 result in a sender becoming cwnd-limited when the cwnd is nearly, 662 rather than completely, equal to the FlightSize. 664 5. Determining a safe period to preserve cwnd 666 This section documents the rationale for selecting the maximum period 667 that cwnd may be preserved, known as the non-validated period, NVP. 669 Limiting the period that cwnd may be preserved avoids undesirable 670 side effects that would result if the cwnd were to be kept 671 unnecessarily high for an arbitrary long period, which was a part of 672 the problem that CWV originally attempted to address. The period a 673 sender may safely preserve the cwnd, is a function of the period that 674 a network path is expected to sustain the capacity reflected by cwnd. 675 There is no ideal choice for this time. 677 A period of five minutes was chosen for this NVP. This is a 678 compromise that was larger than the idle intervals of common 679 applications, but not sufficiently larger than the period for which 680 the capacity of an Internet path may commonly be regarded as stable. 681 The capacity of wired networks is usually relatively stable for 682 periods of several minutes and that load stability increases with the 683 capacity. This suggests that cwnd may be preserved for at least a 684 few minutes. 686 There are cases where the TCP throughput exhibits significant 687 variability over a time less than five minutes. Examples could 688 include wireless topologies, where TCP rate variations may fluctuate 689 on the order of a few seconds as a consequence of medium access 690 protocol instabilities. Mobility changes may also impact TCP 691 performance over short time scales. Senders that observe such rapid 692 changes in the path characteristic may also experience increased 693 congestion with the new method, however such variation would likely 694 also impact TCP's behaviour when supporting interactive and bulk 695 applications. 697 Routing algorithms may modify the network path, disrupting the RTT 698 measurement and changing the capacity available to a TCP connection, 699 however such changes do not usually occur within a time frame of a 700 few minutes. 702 The value of five minutes is therefore expected to be sufficient for 703 most current applications. Simulation studies (e.g. [Bis11]) also 704 suggest that for many practical applications, the performance using 705 this value will not be significantly different to that observed using 706 a non-standard method that does not reset the cwnd after idle. 708 Finally, other TCP sender mechanisms have used a 5 minute timer, and 709 there could be simplifications in some implementations by reusing the 710 same interval. TCP defines a default user timeout of 5 minutes 711 [RFC0793] i.e., how long transmitted data may remain unacknowledged 712 before a connection is forcefully closed. 714 6. Security Considerations 716 General security considerations concerning TCP congestion control are 717 discussed in [RFC5681]. This document describes an algorithm that 718 updates one aspect of the congestion control procedures, and so the 719 considerations described in RFC 5681 also apply to this algorithm. 721 7. IANA Considerations 723 There are no IANA considerations. 725 8. Acknowledgments 727 The authors acknowledge the contributions of Dr I Biswas, Mr Ziaul 728 Hossain in supporting the evaluation of CWV and for their help in 729 developing the mechanisms proposed in this draft. We also 730 acknowledge comments received from the Internet Congestion Control 731 Research Group, in particular Yuchung Cheng, Mirja Kuehlewind, Joe 732 Touch, and Mark Allman. This work was part-funded by the European 733 Community under its Seventh Framework Programme through the Reducing 734 Internet Transport Latency (RITE) project (ICT-317700). 736 9. Author Notes 738 RFC-Editor note: please remove this section prior to publication. 740 9.1. Other related work 742 RFC-Editor note: please remove this section prior to publication. 744 There are several issues to be discussed more widely: 746 o There are potential interactions with the Experimental update in 747 [RFC6928] that raises the TCP initial Window to ten segments, do 748 these cases need to be elaborated? 750 This relates to the Experimental specification for increasing 751 the TCP IW defined in RFC 6928. 753 The two methods have different functions and different response 754 to loss/congestion. 756 RFC 6928 proposes an experimental update to TCP that would 757 increase the IW to ten segments. This would allow faster 758 opening of the cwnd, and also a large (same size) restart 759 window. This approach is based on the assumption that many 760 forward paths can sustain bursts of up to ten segments without 761 (appreciable) loss. Such a significant increase in cwnd must 762 be matched with an equally large reduction of cwnd if loss/ 763 congestion is detected, and such a congestion indication is 764 likely to require future use of IW=10 to be disabled for this 765 path for some time. This guards against the unwanted behaviour 766 of a series of short flows continuously flooding a network path 767 without network congestion feedback. 769 In contrast, this document proposes an update with a rationale 770 that relies on recent previous path history to select an 771 appropriate cwnd after restart. 773 The behaviour differs in three ways: 775 1) For applications that send little initially, new-cwv may 776 constrain more than RFC 6928, but would not require the 777 connection to reset any path information when a restart 778 incurred loss. In contrast, new-cwv would allow the TCP 779 connection to preserve the cached cwnd, any loss, would impact 780 cwnd, but not impact other flows. 782 2) For applications that utilise more capacity than provided by 783 a cwnd of 10 segments, this method would permit a larger 784 restart window compared to a restart using the method in RFC 785 6928. This is justified by the recent path history. 787 3) new-CWV is attended to also be used for rate-limited 788 applications, where the application sends, but does not seek to 789 fully utilise the cwnd. In this case, new-cwv constrains the 790 cwnd to that justified by the recent path history. The 791 performance trade-offs are hence different, and it would be 792 possible to enable new-cwv when also using the method in RFC 793 6928, and yield benefits. 795 o There is potential overlap with the Laminar proposal (draft- 796 mathis-tcpm-tcp-laminar) 797 The current draft was intended as a standards-track update to 798 TCP, rather than a new transport variant. At least, it would 799 be good to understand how the two interact and whether there is 800 a possibility of a single method. 802 o There is potential performance loss in loss of a short burst 803 (off list with M Allman) 805 A sender can transmit several segments then become idle. If 806 the first set of segments are all Acknowledged, the ssthresh 807 collapses to a small value (no new data is sent by the idle 808 sender). Loss of the later data results in congestion (e.g., 809 maybe a RED drop or some other cause, rather than the maximum 810 rate of this flow). When the sender performs loss recovery it 811 may have an appreciable pipeACK and cwnd, but a very low 812 FlightSize - the Standard algorithm therefore results in an 813 unusually low cwnd ((1/2)* FlightSize). 815 A constant rate flow would have maintained a FlightSize 816 appropriate to pipeACK (cwnd, if it is a bulk flow). 818 This could be fixed by adding a new state variable? It could 819 also be argued this is a corner case (e.g. loss of only the 820 last segments would have resulted in RTO), the impact could be 821 significant. 823 o There is potential interaction with TCP Control Block Sharing(M 824 Welzl) 826 An application that is non-validated can accumulate a cwnd that 827 is larger than the actual capacity. Is this a fair value to 828 use in TCB sharing? 830 We propose that TCB sharing should use the pipeACK in place of 831 cwnd when a TCP sender is in the Non-validated phase. This 832 value better reflects the capacity that the flow has utilised 833 in the network path. 835 10. Revision notes 837 RFC-Editor note: please remove this section prior to publication. 839 Draft 03 was submitted to ICCRG to receive comments and feedback. 841 Draft 04 contained the first set of clarifications after feedback: 843 o Changed name to application limited and used the term rate-limited 844 in all places. 846 o Added justification and many minor changes suggested on the list. 848 o Added text to tie-in with more accurate ECN marking. 850 o Added ref to Hug01 852 Draft 05 contained various updates: 854 o New text to redefine how to measure the acknowledged pipe, 855 differentiating this from the FlightSize, and hence avoiding 856 previous issues with infrequent large bursts of data not being 857 validated. A key point new feature is that pipeACK only triggers 858 leaving the NVP after the size of the pipe has been acknowledged. 859 This removed the need for hysteresis. 861 o Reduction values were changed to 1/2, following analysis of 862 suggestions from ICCRG. This also sets the "target" cwnd as twice 863 the used rate for non-validated case. 865 o Introduced a symbolic name (NVP) to denote the 5 minute period. 867 Draft 06 contained various updates: 869 o Required reset of pipeACK after congestion. 871 o Added comment on the effect of congestion after a short burst (M. 872 Allman). 874 o Correction of minor Typos. 876 WG draft 00 contained various updates: 878 o Updated initialisation of pipeACK to maximum value. 880 o Added note on intended status still to be determined. 882 WG draft 01 contained: 884 o Added corrections from Richard Scheffenegger. 886 o Raffaello Secchi added to the mechanism, based on implementation 887 experience. 889 o Removed that the requirement for the method to use TCP SACK option 891 o Although it may be desirable to use SACK, this is not essential to 892 the algorithm. 894 o Added the notion of the sampling period to accommodate large rate 895 variations and ensure that the method is stable. This algorithm 896 to be validated through implementation. 898 WG draft 02 contained: 900 o Clarified language around pipeACK variable and pipeACK sample - 901 Feedback from Aris Angelogiannopoulos. 903 WG draft 03 contained: 905 o Editorial corrections - Feedback from Anna Brunstrom. 907 o An adjustment to the procedure at the start and end of Reoloss 908 recovery to align the two equations. 910 o Further clarification of the "undefined" value of the pipeACK 911 variable. 913 WG draft 04 contained: 915 o Editorial corrections. 917 o Introduced the "cwnd-limited" term. 919 o An adjustment to the procedure at the start of a cwnd-limited 920 phase - the new text is intended to ensure that new-cwv is not 921 unnecessarily more conservative than standard TCP when the flow is 922 cwnd-limited. This resolves two issues: first it prevents 923 pathologies in which pipeACK increases slowly and erratically. It 924 also ensures that performance of bulk applications is not 925 significantly impacted when using the method. 927 o Clearly identifies that pacing (or equivalent) is requiring during 928 the NVP to control burstiness. New section added. 930 WG draft 05 contained: 932 o Clarification to first two bullets in Section 4.4 describing cwnd- 933 limited, to explain these are really alternates to the same case. 935 o Section giving implementation examples was restructured to clarify 936 there are two methods described. 938 o Cross References to sections updated - thanks to comments from 939 Martin Winbjoerk and Tim Wicinski. 941 WG draft 06 contained: 943 o The section giving implementation examples was restructured to 944 clarify there are two methods described. 946 o Justification of design decisions. 948 o Re-organised text to improve clarity of argument. 950 WG draft 07 contained: 952 o Updated publication date. 954 o Text on noting that cwnd shouldn't ever be made negative. 956 o Updated text on ECN to clarify the process where R is a reduction 957 based on ECN marks. 959 WG draft 08 contained: 961 o Removed description of how to use Accurate ECN feedback. It is 962 not clear that this document should specify a usage of a mechanism 963 that has not been fully defined. Accurate ECN may lead to 964 different congestion responses and these will need to be defined 965 in the CC specifications for using Accurate ECN. 967 WG draft 09 contained: 969 o Removed update to RFC 5681 - the status of the present document is 970 Experimental, and hence this document does not update RFC 5681. 972 11. References 974 11.1. Normative References 976 [RFC0793] Postel, J., "Transmission Control Protocol", STD 7, RFC 977 793, September 1981. 979 [RFC2018] Mathis, M., Mahdavi, J., Floyd, S., and A. Romanow, "TCP 980 Selective Acknowledgment Options", RFC 2018, October 1996. 982 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 983 Requirement Levels", BCP 14, RFC 2119, March 1997. 985 [RFC2861] Handley, M., Padhye, J., and S. Floyd, "TCP Congestion 986 Window Validation", RFC 2861, June 2000. 988 [RFC5681] Allman, M., Paxson, V., and E. Blanton, "TCP Congestion 989 Control", RFC 5681, September 2009. 991 [RFC6675] Blanton, E., Allman, M., Wang, L., Jarvinen, I., Kojo, M., 992 and Y. Nishida, "A Conservative Loss Recovery Algorithm 993 Based on Selective Acknowledgment (SACK) for TCP", RFC 994 6675, August 2012. 996 11.2. Informative References 998 [All05] Allman, M. and E. Blanton, "Notes on burst mitigation for 999 transport protocols", March 2005. 1001 [Bis08] Biswas, I. and G. Fairhurst, "A Practical Evaluation of 1002 Congestion Window Validation Behaviour, 9th Annual 1003 Postgraduate Symposium in the Convergence of 1004 Telecommunications, Networking and Broadcasting (PGNet), 1005 Liverpool, UK", June 2008. 1007 [Bis10] Biswas, I., Sathiaseelan, A., Secchi, R., and G. 1008 Fairhurst, "Analysing TCP for Bursty Traffic, Int'l J. of 1009 Communications, Network and System Sciences, 7(3)", June 1010 2010. 1012 [Bis11] Biswas, I., "PhD Thesis, Internet congestion control for 1013 variable rate TCP traffic, School of Engineering, 1014 University of Aberdeen", June 2011. 1016 [Fai12] Sathiaseelan, A., Secchi, R., Fairhurst, G., and I. 1017 Biswas, "Enhancing TCP Performance to support Variable- 1018 Rate Traffic, 2nd Capacity Sharing Workshop, ACM CoNEXT, 1019 Nice, France, 10th December 2012.", June 2008. 1021 [Hug01] Hughes, A., Touch, J., and J. Heidemann, "Issues in TCP 1022 Slow-Start Restart After Idle (Work-in-Progress)", 1023 December 2001. 1025 [Liu07] Liu, D., Allman, M., Jiny, S., and L. Wang, "Congestion 1026 Control without a Startup Phase, 5th International 1027 Workshop on Protocols for Fast Long-Distance Networks 1028 (PFLDnet), Los Angeles, California, USA", February 2007. 1030 [RFC2616] Fielding, R., Gettys, J., Mogul, J., Frystyk, H., 1031 Masinter, L., Leach, P., and T. Berners-Lee, "Hypertext 1032 Transfer Protocol -- HTTP/1.1", RFC 2616, June 1999. 1034 [RFC6298] Paxson, V., Allman, M., Chu, J., and M. Sargent, 1035 "Computing TCP's Retransmission Timer", RFC 6298, June 1036 2011. 1038 [RFC6928] Chu, J., Dukkipati, N., Cheng, Y., and M. Mathis, 1039 "Increasing TCP's Initial Window", RFC 6928, April 2013. 1041 Authors' Addresses 1043 Godred Fairhurst 1044 University of Aberdeen 1045 School of Engineering 1046 Fraser Noble Building 1047 Aberdeen, Scotland AB24 3UE 1048 UK 1050 Email: gorry@erg.abdn.ac.uk 1051 URI: http://www.erg.abdn.ac.uk 1053 Arjuna Sathiaseelan 1054 University of Aberdeen 1055 School of Engineering 1056 Fraser Noble Building 1057 Aberdeen, Scotland AB24 3UE 1058 UK 1060 Email: arjuna@erg.abdn.ac.uk 1061 URI: http://www.erg.abdn.ac.uk 1063 Raffaello Secchi 1064 University of Aberdeen 1065 School of Engineering 1066 Fraser Noble Building 1067 Aberdeen, Scotland AB24 3UE 1068 UK 1070 Email: raffaello@erg.abdn.ac.uk 1071 URI: http://www.erg.abdn.ac.uk