idnits 2.17.1 draft-ietf-tcpm-newcwv-08.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 (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 23, 2015) is 3350 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 (==), 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) R. Secchi 5 Updates: 5681 (if approved) University of Aberdeen 6 Intended status: Experimental February 23, 2015 7 Expires: August 27, 2015 9 Updating TCP to support Rate-Limited Traffic 10 draft-ietf-tcpm-newcwv-08 12 Abstract 14 This document updates RFC 5681 to address issues that arise when TCP 15 is used to support traffic that exhibits periods where the sending 16 rate is limited by the application rather than the congestion window. 17 It provides an experimental update to TCP that allows a TCP sender to 18 restart quickly following a rate-limited interval. This method is 19 expected to benefit applications that send rate-limited traffic using 20 TCP, while also providing an appropriate response if congestion is 21 experienced. 23 It also evaluates the Experimental specification of TCP Congestion 24 Window Validation, CWV, defined in RFC 2861, and concludes that RFC 25 2861 sought to address important issues, but failed to deliver a 26 widely used solution. This document therefore recommends that the 27 status of RFC 2861 is moved from Experimental to Historic, and that 28 it is replaced by the current specification. 30 Status of This Memo 32 This Internet-Draft is submitted in full conformance with the 33 provisions of BCP 78 and BCP 79. 35 Internet-Drafts are working documents of the Internet Engineering 36 Task Force (IETF). Note that other groups may also distribute 37 working documents as Internet-Drafts. The list of current Internet- 38 Drafts is at http://datatracker.ietf.org/drafts/current/. 40 Internet-Drafts are draft documents valid for a maximum of six months 41 and may be updated, replaced, or obsoleted by other documents at any 42 time. It is inappropriate to use Internet-Drafts as reference 43 material or to cite them other than as "work in progress." 45 This Internet-Draft will expire on August 27, 2015. 47 Copyright Notice 49 Copyright (c) 2015 IETF Trust and the persons identified as the 50 document authors. All rights reserved. 52 This document is subject to BCP 78 and the IETF Trust's Legal 53 Provisions Relating to IETF Documents 54 (http://trustee.ietf.org/license-info) in effect on the date of 55 publication of this document. Please review these documents 56 carefully, as they describe your rights and restrictions with respect 57 to this document. Code Components extracted from this document must 58 include Simplified BSD License text as described in Section 4.e of 59 the Trust Legal Provisions and are provided without warranty as 60 described in the Simplified BSD License. 62 Table of Contents 64 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 2 65 1.1. Standards Status of this Document . . . . . . . . . . . . 4 66 2. Reviewing experience with TCP-CWV . . . . . . . . . . . . . . 5 67 3. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . 6 68 4.1. Initialisation . . . . . . . . . . . . . . . . . . . . . 8 69 4.2. Estimating the validated capacity supported by a path . . 8 70 4.3. Preserving cwnd during a rate-limited period. . . . . . . 9 71 4.4. TCP congestion control during the non-validated phase . . 9 72 4.4.1. Response to congestion in the non-validated phase . . 10 73 4.4.2. Sender burst control during the non-validated phase . 12 74 4.4.3. Adjustment at the end of the non-validated phase . . 12 75 4.5. Examples of Implementation . . . . . . . . . . . . . . . 13 76 4.5.1. Implementing the pipeACK measurement . . . . . . . . 13 77 4.5.2. Implementing detection of the cwnd-limited condition 14 78 5. Determining a safe period to preserve cwnd . . . . . . . . . 15 79 6. Security Considerations . . . . . . . . . . . . . . . . . . . 16 80 7. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 16 81 8. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . 16 82 9. Author Notes . . . . . . . . . . . . . . . . . . . . . . . . 16 83 9.1. Other related work . . . . . . . . . . . . . . . . . . . 16 84 9.2. Revision notes . . . . . . . . . . . . . . . . . . . . . 18 85 10. References . . . . . . . . . . . . . . . . . . . . . . . . . 21 86 10.1. Normative References . . . . . . . . . . . . . . . . . . 21 87 10.2. Informative References . . . . . . . . . . . . . . . . . 22 88 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 23 90 1. Introduction 92 TCP is used to support a range of application behaviours. The TCP 93 congestion window (cwnd) controls the number of unacknowledged 94 packets/bytes that a TCP flow may have in the network at any time, a 95 value known as the FlightSize [RFC5681]. A bulk application will 96 always have data available to transmit. The rate at which it sends 97 is therefore limited by the maximum permitted by the receiver 98 advertised window and the sender congestion window (cwnd). In 99 contrast, a rate-limited application will experience periods when the 100 sender is either idle or is unable to send at the maximum rate 101 permitted by the cwnd. The update in this document targets the 102 operation of TCP in such rate-limited cases. 104 Standard TCP [RFC5681] states that a TCP sender SHOULD set cwnd to no 105 more than the Restart Window (RW) before beginning transmission, if 106 the TCP sender has not sent data in an interval exceeding the 107 retransmission timeout, i..e when an application becomes idle. 108 [RFC2861] noted that this TCP behaviour was not always observed in 109 current implementations. Experiments [Bis08] confirm this to still 110 be the case. 112 Congestion Window Validation, CWV, introduced the terminology of 113 "application limited periods". This document describes any time that 114 an application limits the sending rate, rather than being limited by 115 the transport, as "rate-limited". This update improves support for 116 applications that vary their transmission rate, either with (short) 117 idle periods between transmission or by changing the rate the 118 application sends. These applications are characterised by the TCP 119 FlightSize often being less than cwnd. Many Internet applications 120 exhibit this behaviour, including web browsing, http-based adaptive 121 streaming, applications that support query/response type protocols, 122 network file sharing, and live video transmission. Many such 123 applications currently avoid using long-lived (persistent) TCP 124 connections (e.g. [RFC2616] servers typically support persistent 125 HTTP connections, but do not enable this by default). Such 126 applications often instead either use a succession of short TCP 127 transfers or use UDP. 129 Standard TCP does not impose additional restrictions on the growth of 130 the congestion window when a TCP sender is unable to send at the 131 maximum rate allowed by the cwnd. In this case the rate-limited 132 sender may grow a cwnd far beyond that corresponding to the current 133 transmit rate, resulting in a value that does not reflect current 134 information about the state of the network path the flow is using. 135 Use of such an invalid cwnd may result in reduced application 136 performance and/or could significantly contribute to network 137 congestion. 139 [RFC2861] proposed a solution to these issues in an experimental 140 method known as CWV. CWV was intended to help reduce cases where TCP 141 accumulated an invalid (inappropriately large) cwnd. The use and 142 drawbacks of using the CWV algorithm in RFC 2861 with an application 143 are discussed in Section 2. 145 Section 3 defines relevant terminology. 147 Section 4 specifies an alternative to CWV that seeks to address the 148 same issues, but does this in a way that is expected to mitigate the 149 impact on an application that varies its sending rate. The updated 150 method applies to the rate-limited conditions (including both an 151 application-limited and idle sender). 153 The goals of this update are: 155 o To not change the behaviour of a TCP sender that performs bulk 156 transfers that consume the cwnd. 158 o To provide a method that co-exists with Standard TCP and other 159 flows that use this updated method. 161 o To reduce transfer latency for applications that change their rate 162 over short intervals of time. 164 o To avoid a TCP sender growing a large "non-validated" cwnd, when 165 it has not recently sent using this cwnd. 167 o To remove the incentive for ad-hoc application or network stack 168 methods (such as "padding") solely to maintain a large cwnd for 169 future transmission. 171 o To incentivise the use of long-lived connections, rather than a 172 succession of short-lived flows, benefiting both flows and network 173 when actual congestion is encountered. 175 Section 5 describes the rationale for selecting the safe period to 176 preserve the cwnd. 178 1.1. Standards Status of this Document 180 This document was produced by the TCP Maintenance and Minor 181 Extensions (tcpm) working group. 183 The document updates and obsoletes the methods described in 184 [RFC2861]. It recommends a set of mechanisms, including the use of 185 pacing during a non-validated period. The updated mechanisms are 186 intended to have a less aggressive congestion impact than would be 187 exhibited by a standard TCP sender. 189 The specification in this draft is classified as "Experimental" 190 pending experience with deployed implementations of the methods. 192 2. Reviewing experience with TCP-CWV 194 [RFC2861] described a simple modification to the TCP congestion 195 control algorithm that decayed the cwnd after the transition to a 196 "sufficiently-long" idle period. This used the slow-start threshold 197 (ssthresh) to save information about the previous value of the 198 congestion window. The approach relaxed the standard TCP behaviour 199 [RFC5681] for an idle session, intended to improve application 200 performance. CWV also modified the behaviour where a sender 201 transmitted at a rate less than allowed by cwnd. 203 [RFC2861] proposed two set of responses, one after an "application- 204 limited" and one after an "idle period". Although this distinction 205 was argued, in practice differentiating the two conditions was found 206 problematic in actual networks (e.g.[Bis10]). This offers 207 predictable performance for long on-off periods (>>1 RTT), or slowly 208 varying rate-based traffic, the performance could be unpredictable 209 for variable-rate traffic and depended both upon whether an accurate 210 RTT had been obtained and the pattern of application traffic relative 211 to the measured RTT. 213 Many applications can and often do vary their transmission over a 214 wide range of rates. Using [RFC2861] such applications often 215 experienced varying performance, which made it hard for application 216 developers to predict the TCP latency even when using a path with 217 stable network characteristics. We argue that an attempt to classify 218 application behaviour as application-limited or idle is problematic 219 and also inappropriate. This document therefore explicitly avoids 220 trying to differentiate these two cases, instead treating all rate- 221 limited traffic uniformly. 223 [RFC2861] has been implemented in some mainstream operating systems 224 as the default behaviour [Bis08]. Analysis (e.g. [Bis10] [Fai12]) 225 has shown that a TCP sender using CWV is able to use available 226 capacity on a shared path after an idle period. This can benefit 227 variable-rate applications, especially over long delay paths, when 228 compared to the slow-start restart specified by standard TCP. 229 However, CWV would only benefit an application if the idle period 230 were less than several Retransmission Time Out (RTO) intervals 231 [RFC6298], since the behaviour would otherwise be the same as for 232 standard TCP, which resets the cwnd to the TCP Restart Window after 233 this period. 235 To enable better performance for variable-rate applications with TCP, 236 some operating systems have chosen to support non-standard methods, 237 or applications have resorted to "padding" streams by sending dummy 238 data to maintain their sending rate when they have no data to 239 transmit. Although transmitting redundant data across a network path 240 provides good evidence that the path can sustain data at the offered 241 rate, padding also consumes network capacity and reduces the 242 opportunity for congestion-free statistical multiplexing. For 243 variable-rate flows, the benefits of statistical multiplexing can be 244 significant and it is therefore a goal to find a viable alternative 245 to padding streams. 247 Experience with [RFC2861] suggests that although the CWV method 248 benefited the network in a rate-limited scenario (reducing the 249 probability of network congestion), the behaviour was too 250 conservative for many common rate-limited applications. This 251 mechanism did not therefore offer the desirable increase in 252 application performance for rate-limited applications and it is 253 unclear whether applications actually use this mechanism in the 254 general Internet. 256 It is therefore concluded that CWV, as defined in [RFC2861], was 257 often a poor solution for many rate-limited applications. It had the 258 correct motivation, but had the wrong approach to solving this 259 problem. 261 3. Terminology 263 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 264 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 265 document are to be interpreted as described in [RFC2119]. 267 The document assumes familiarity with the terminology of TCP 268 congestion control [RFC5681]. 270 The following terminology is used in this document: 272 cwnd-limited: A TCP flow that has sent the maximum number of segments 273 permitted by the cwnd, where the application utilises the allowed 274 sending rate (see Section 4.5.2). 276 pipeACK sample: A measure of the volume of data acknowledged by the 277 network within an RTT. 279 pipeACK variable: A variable that measures the available capacity 280 using the set of pipeACK samples. 282 pipeACK Sampling Period: The maximum period that a measured pipeACK 283 sample may influence the pipeACK variable. 285 Non-validated phase: The phase where the cwnd reflects a previous 286 measurement of the available path capacity. 288 Non-validated period, NVP: The maximum period for which cwnd is 289 preserved in the non-validated phase. 291 Rate-limited: A TCP flow that does not consume more than one half of 292 cwnd, and hence operates in the non-validated phase. This includes 293 periods when an application is either idle or chooses to send at a 294 rate less than the maximum permitted by the cwnd. 296 Validated phase: The phase where the cwnd reflects a current estimate 297 of the available path capacity. 299 4. A New Congestion Window Validation method 301 This section proposes an update to the TCP congestion control 302 behaviour during a rate-limited interval. This new method 303 intentionally does not differentiate between times when the sender 304 has become idle or chooses to send at a rate less than the maximum 305 allowed by the cwnd. 307 The period where actual usage is less than allowed by cwnd, is named 308 the non-validated phase. The update allows an application in the 309 non-validated phase to resume transmission at a previous rate without 310 incurring the delay of slow-start. However, if the TCP sender 311 experiences congestion using the preserved cwnd, it is required to 312 immediately reset the cwnd to an appropriate value specified by the 313 method. If a sender does not take advantage of the preserved cwnd 314 within the Non-validated period, NVP, the value of cwnd is reduced, 315 ensuring the value better reflects the capacity that was recently 316 actually used. 318 It is expected that this update will satisfy the requirements of many 319 rate-limited applications and at the same time provide an appropriate 320 method for use in the Internet. New-CWV reduces this incentive for 321 an application to send "padding" data simply to keep transport 322 congestion state. 324 The method is specified in following subsections and is expected to 325 encourage applications and TCP stacks to use standards-based 326 congestion control methods. It may also encourage the use of long- 327 lived connections where this offers benefit (such as persistent 328 http). 330 4.1. Initialisation 332 A sender starts a TCP connection in the validated phase and 333 initialises the pipeACK variable to the "undefined" value. This 334 value inhibits use of the value in cwnd calculations. 336 4.2. Estimating the validated capacity supported by a path 338 [RFC6675] defines a variable, FlightSize, that indicates the 339 instantaneous amount of data that has been sent, but not cumulatively 340 acknowledged. In this method a new variable "pipeACK" is introduced 341 to measure the acknowledged size of the network pipe. This is used 342 to determine if the sender has validated the cwnd. pipeACK differs 343 from FlightSize in that it is evaluated over a window of acknowledged 344 data, rather than reflecting the amount of data outstanding. 346 A sender determines a pipeACK sample by measuring the volume of data 347 that was acknowledged by the network over the period of a measured 348 Round Trip Time (RTT). Using the variables defined in [RFC6675], a 349 value could be measured by caching the value of HighACK and after one 350 RTT measuring the difference between the cached HighACK value and the 351 current HighACK value. Other equivalent methods may be used. 353 A sender is not required to continuously update the pipeACK variable 354 after each received ACK, but SHOULD perform a pipeACK sample at least 355 once per RTT when it has sent unacknowledged segments. 357 The pipeACK variable MAY consider multiple pipeACK samples over the 358 pipeACK Sampling Period. The value of the pipeACK variable MUST NOT 359 exceed the maximum (highest value) within the sampling period. This 360 specification defines the pipeACK Sampling Period as Max(3*RTT, 1 361 second). This period enables a sender to compensate for large 362 fluctuations in the sending rate, where there may be pauses in 363 transmission, and allows the pipeACK variable to reflect the largest 364 recently measured pipeACK sample. 366 When no measurements are available, the pipeACK variable is set to 367 the "undefined value". This value is used to inhibit entering the 368 non-validated phase until the first new measurement of a pipeACK 369 sample. 371 The pipeACK variable MUST NOT be updated during TCP Fast Recovery. 372 That is, the sender stops collecting pipeACK samples during loss 373 recovery. The method RECOMMENDS that the TCP SACK option [RFC2018] 374 is enabled and the method defined on [RFC6675]is used to recover 375 missing segments. This allows the sender to more accurately 376 determine the number of missing bytes during the loss recovery phase, 377 and using this method will result in a more appropriate cwnd 378 following loss. 380 4.3. Preserving cwnd during a rate-limited period. 382 The updated method creates a new TCP sender phase that captures 383 whether the cwnd reflects a validated or non-validated value. The 384 phases are defined as: 386 o Validated phase: pipeACK >=(1/2)*cwnd, or pipeACK is undefined. 387 This is the normal phase, where cwnd is expected to be an 388 approximate indication of the capacity currently available along 389 the network path, and the standard methods are used to increase 390 cwnd (currently [RFC5681]). 392 o Non-validated phase: pipeACK <(1/2)*cwnd. This is the phase where 393 the cwnd has a value based on a previous measurement of the 394 available capacity, and the usage of this capacity has not been 395 validated in the pipeACK Sampling Period. That is, when it is not 396 known whether the cwnd reflects the currently available capacity 397 along the network path. The mechanisms to be used in this phase 398 seek to determine a safe value for cwnd and an appropriate 399 reaction to congestion. 401 Note: A threshold is needed to determine whether a sender is in the 402 validated or non-validated phase. A standard TCP sender in slow- 403 start is permitted to double its FlightSize from one RTT to the next. 404 This motivated the choice of a threshold value of 1/2. This 405 threshold ensures a sender does not further increase the cwnd as long 406 as the FlightSize is less than (1/2*cwnd). Furthermore, a sender 407 with a FlightSize less than (1/2*cwnd) may in the next RTT be 408 permitted by the cwnd to send at a rate that more than doubles the 409 FlightSize, and hence this case needs to be regarded as non-validated 410 and a sender therefore needs to employ additional mechanisms while in 411 this phase. 413 4.4. TCP congestion control during the non-validated phase 415 A TCP sender MUST enter the non-validated phase when the pipeACK is 416 less than (1/2)*cwnd. 418 A TCP sender that enters the non-validated phase SHOULD preserve the 419 cwnd (i.e., this neither grows nor reduces while the sender remains 420 in this phase). If the sender receives an indication of congestion, 421 it uses the method described below. The phase is concluded after a 422 fixed period of time (the NVP, as explained in Section 4.4.3) or when 423 the sender transmits sufficient data so that pipeACK > (1/2)*cwnd 424 (i.e. the sender is no longer rate-limited). 426 The behaviour in the non-validated phase is specified as: 428 o A sender determines whether to increase the cwnd based upon 429 whether it is cwnd-limited (see Section 4.5.2): 431 o 433 * A sender that is cwnd-limited MAY use the standard TCP method 434 to increase cwnd (i.e. a TCP sender that fully utilises the 435 cwnd is permitted to increase cwnd each received ACK using 436 standard methods). 438 * A sender that is not cwnd-limited MUST NOT increase the cwnd 439 when ACK packets are received in this phase. 441 o If the sender receives an indication of congestion while in the 442 non-validated phase (i.e., detects loss), the sender MUST exit the 443 non-validated phase (reducing the cwnd as defined in 444 Section 4.4.1). 446 o If the Retransmission Time Out (RTO) expires while in the non- 447 validated phase, the sender MUST exit the non-validated phase. It 448 then resumes using the standard TCP RTO mechanism [RFC5681]. 450 o A sender with a pipeACK variable greater than (1/2)*cwnd SHOULD 451 enter the validated phase. (A rate-limited sender will not 452 normally be impacted by whether it is in a validated or non- 453 validated phase, since it will normally not consume the entire 454 cwnd. However a change to the validated phase will release the 455 sender from constraints on the growth of cwnd, and restore the use 456 of the standard congestion response.) 458 The cwnd-limited behaviour may be triggered during a transient 459 condition that occurs when a sender is in the non-validated phase and 460 receives an ACK that acknowledges received data, the cwnd was fully 461 utilised, and more data is awaiting transmission than may be sent 462 with the current cwnd. The sender is then allowed to use the 463 standard method to increase the cwnd. (Note, if the sender succeeds 464 in sending these new segments, the updated cwnd and pipeACK variables 465 will eventually result in a transition to the validated phase.) 467 4.4.1. Response to congestion in the non-validated phase 469 Reception of congestion feedback while in the non-validated phase is 470 interpreted as an indication that it was inappropriate for the sender 471 to use the preserved cwnd. The sender is therefore required to 472 quickly reduce the rate to avoid further congestion. Since the cwnd 473 does not have a validated value, a new cwnd value must be selected 474 based on the utilised rate. 476 A sender that detects a packet-drop MUST record the current 477 FlightSize in the variable LossFlightSize and MUST calculate a safe 478 cwnd for loss recovery using the method below: 480 cwnd = (Max(pipeACK,LossFlightSize))/2. 482 The pipeACK value is not updated during loss recoverySection 4.2. If 483 there is a valid pipeACK value, the new cwnd is adjusted to reflect 484 that a non-validated cwnd may be larger than the actual FlightSize, 485 or recently used FlightSize (recorded in pipeACK). The updated cwnd 486 therefore prevents overshoot by a sender significantly increasing its 487 transmission rate during the recovery period. 489 At the end of the recovery phase, the TCP sender MUST reset the cwnd 490 using the method below: 492 cwnd = (Max(pipeACK,LossFlightSize) - R)/2. 494 Where R is the volume of data that was successfully retransmitted 495 during the recovery phase. This counts segments retransmitted and 496 considered lost by the pipe estimation algorithm at the end of 497 recovery. It does not include the additional cost of multiple 498 retransmission of the same data. 500 The calculated cwnd value MUST NOT be reduced below 1 MSS. 502 After completing the loss recovery phase, the sender MUST re- 503 initialise the pipeACK variable to the "undefined" value. This 504 ensures that standard TCP methods are used immediately after 505 completing loss recovery until a new pipeACK value can be determined. 507 ssthresh is adjusted using the standard TCP method. 509 Note: The adjustment by reducing cwnd by the volume of data not sent 510 (R) follows the method proposed for Jump Start [Liu07]. The 511 inclusion of the term R makes the adjustment more conservative than 512 standard TCP. This is required, since a sender in the non-validated 513 state may increase the rate more than a standard TCP would have done 514 relative to what was sent in the last RTT (i.e., more than doubled 515 the number of segments in flight relative to what it sent in the last 516 RTT). The additional reduction after congestion is beneficial when 517 the LossFlightSize has significantly overshot the available path 518 capacity incurring significant loss (e.g. following a change of path 519 characteristics or when additional traffic has taken a larger share 520 of the network bottleneck during a period when the sender transmits 521 less). 523 Note: The pipeACK value is only valid during a non-validated phase, 524 and therefore does not exceed cwnd/2. If LossFlightSize and R were 525 small, then this can result in the final cwnd after loss recovery 526 being 1/4 of the cwnd on detection of congestion. This reduction is 527 conservative, and pipeACK is reset to undefined. Subsequent updates 528 to cwnd do not therefore reflect pipeACK history before any 529 congestion event. 531 4.4.2. Sender burst control during the non-validated phase 533 TCP congestion control allows a sender to accumulate a cwnd that 534 would allow it to send a burst of segments with a total size up to 535 the difference between the FlightsSize and cwnd. Such bursts can 536 impact other flows that share a network bottleneck and/or may induce 537 congestion when buffering is limited. 539 Various methods have been proposed to control the sender burstiness 540 [Hug01], [All05]. For example, TCP can limit the number of new 541 segments it sends per received ACK. This is effective when a flow of 542 ACKs is received, but can not be used to control a sender that has 543 not send appreciable data in the previous RTT [All05]. 545 This document recommends using a method to avoid line-rate bursts 546 after an idle or rate-limited interval when there is less reliable 547 information about the capacity of the network path: A TCP sender in 548 the non-validated phase SHOULD control the maximum burst size, e.g. 549 using a rate-based pacing algorithm in which a sender paces out the 550 cwnd over its estimate of the RTT, or some other method, to prevent 551 many segments being transmitted contiguously at line-rate. The most 552 appropriate method(s) to implement pacing depend on the design of the 553 TCP/IP stack, speed of interface and whether hardware support (such 554 as TCP Segment Offload, TSO) is used. The present document does not 555 recommend any specific method. 557 4.4.3. Adjustment at the end of the non-validated phase 559 An application that remains in the non-validated phase for a period 560 greater than the NVP is required to adjust its congestion control 561 state. If the sender exits the non-validated phase after this 562 period, it MUST update the ssthresh: 564 ssthresh = max(ssthresh, 3*cwnd/4). 566 (This adjustment of ssthresh ensures that the sender records that it 567 has safely sustained the present rate. The change is beneficial to 568 rate-limited flows that encounter occasional congestion, and could 569 otherwise suffer an unwanted additional delay in recovering the 570 sending rate.) 572 The sender MUST then update cwnd to be not greater than: 574 cwnd = max((1/2)*cwnd, IW). 576 Where IW is the appropriate TCP initial window, used by the TCP 577 sender (e.g. [RFC5681]). 579 Note: This adjustment ensures that the sender responds conservatively 580 after remaining in the non-validated phase for more than the non- 581 validated period. In this case, it reduces the cwnd by a factor of 582 two from the preserved value. This adjustment is helpful when flows 583 accumulate but do not use a large cwnd, and seeks to mitigate the 584 impact when these flows later resume transmission. This could for 585 instance mitigate the impact if multiple high-rate application flows 586 were to become idle over an extended period of time and then were 587 simultaneously awakened by an external event. 589 4.5. Examples of Implementation 591 This section provides informative examples of implementation methods. 592 Implementations may choose to use other methods that comply with the 593 normative requirements. 595 4.5.1. Implementing the pipeACK measurement 597 A pipeACK sample may be measured once each RTT. This reduces the 598 sender processing burden for calculating after each acknowledgement 599 and also reduces storage requirements at the sender. 601 Since application behaviour can be bursty using CWV, it may be 602 desirable to implement a maximum filter to accumulate the measured 603 values so that the pipeACK variable records the largest pipeACK 604 sample within the pipeACK Sampling Period. One simple way to 605 implement this is to divide the pipeACK Sampling Period into several 606 (e.g. 5) equal length measurement periods. The sender then records 607 the start time for each measurement period and the highest measured 608 pipeACK sample. At the end of the measurement period, any 609 measurement(s) that are older than the pipeACK Sampling Period are 610 discarded. The pipeACK variable is then assigned the largest of the 611 set of the highest measured values. 613 +----------+----------+ +----------+---...... 614 | Sample A | Sample B | No | Sample C | Sample D 615 | | | Sample | | 616 | |\ 5 | | | | 617 | | | | | | /\ 4 | 618 | | | | |\ 3 | | | \ | 619 | | \ | | \--- | | / \ | /| 2 620 |/ \------| - | | / \------/ \... 621 +----------+---------\+----/ /----+/---------+-------------> Time 623 <------------------------------------------------| 624 Sampling Period Current Time 626 Figure 1: Example of measuring pipeACK samples 628 Figure 1 shows an example of how measurement samples may be 629 collected. At the time represented by the figure new samples are 630 being accumulated into sample D. Three previous samples also fall 631 within the pipeACK Sampling Period: A, B, and C. There was also a 632 period of inactivity between samples B and C during which no 633 measurements were taken. The current value of the pipeACK variable 634 will be 5, the maximum across all samples. 636 After one further measurement period, Sample A will be discarded, 637 since it then is older than the pipeACK Sampling Period and the 638 pipeACK variable will be recalculated, Its value will be the larger 639 of Sample C or the final value accumulated in Sample D. 641 Note: the pipeACK Sampling Period and the NVP period do not 642 necessarily require a new timer to be implemented. An alternative is 643 to record a timestamp when the sender enters the NVP. Each time a 644 sender transmits a new segment, this timestamp may be used to 645 determine if the NVP period has expired. If the period expires, the 646 sender may take into account how many units of the NVP period have 647 passed and make one reduction (as defined in Section 4.4.3) for each 648 NVP period. 650 4.5.2. Implementing detection of the cwnd-limited condition 652 A method is required to detect the cwnd-limited condition (see 653 Section 4.4. This is used to detect a condition where a sender in 654 the non-validated phase receives an ACK, but the size of cwnd 655 prevents sending more new data. 657 In simple terms this condition is true only when the TCP sender's 658 FlightSize is equal to or larger than the cwnd. However, an 659 implementation must consider other constraints on the way in which 660 cwnd variable is used, for instance the need to support methods such 661 as the Nagle Algorithm and TCP Segment Offload (TSO). This can 662 result in a sender becoming cwnd-limited when the cwnd is nearly, 663 rather than completely, equal to the FlightSize. 665 5. Determining a safe period to preserve cwnd 667 This section documents the rationale for selecting the maximum period 668 that cwnd may be preserved, known as the non-validated period, NVP. 670 Limiting the period that cwnd may be preserved avoids undesirable 671 side effects that would result if the cwnd were to be kept 672 unnecessarily high for an arbitrary long period, which was a part of 673 the problem that CWV originally attempted to address. The period a 674 sender may safely preserve the cwnd, is a function of the period that 675 a network path is expected to sustain the capacity reflected by cwnd. 676 There is no ideal choice for this time. 678 A period of five minutes was chosen for this NVP. This is a 679 compromise that was larger than the idle intervals of common 680 applications, but not sufficiently larger than the period for which 681 the capacity of an Internet path may commonly be regarded as stable. 682 The capacity of wired networks is usually relatively stable for 683 periods of several minutes and that load stability increases with the 684 capacity. This suggests that cwnd may be preserved for at least a 685 few minutes. 687 There are cases where the TCP throughput exhibits significant 688 variability over a time less than five minutes. Examples could 689 include wireless topologies, where TCP rate variations may fluctuate 690 on the order of a few seconds as a consequence of medium access 691 protocol instabilities. Mobility changes may also impact TCP 692 performance over short time scales. Senders that observe such rapid 693 changes in the path characteristic may also experience increased 694 congestion with the new method, however such variation would likely 695 also impact TCP's behaviour when supporting interactive and bulk 696 applications. 698 Routing algorithms may modify the network path, disrupting the RTT 699 measurement and changing the capacity available to a TCP connection, 700 however such changes do not usually occur within a time frame of a 701 few minutes. 703 The value of five minutes is therefore expected to be sufficient for 704 most current applications. Simulation studies (e.g. [Bis11]) also 705 suggest that for many practical applications, the performance using 706 this value will not be significantly different to that observed using 707 a non-standard method that does not reset the cwnd after idle. 709 Finally, other TCP sender mechanisms have used a 5 minute timer, and 710 there could be simplifications in some implementations by reusing the 711 same interval. TCP defines a default user timeout of 5 minutes 712 [RFC0793] i.e. how long transmitted data may remain unacknowledged 713 before a connection is forcefully closed. 715 6. Security Considerations 717 General security considerations concerning TCP congestion control are 718 discussed in [RFC5681]. This document describes an algorithm that 719 updates one aspect of the congestion control procedures, and so the 720 considerations described in RFC 5681 also apply to this algorithm. 722 7. IANA Considerations 724 There are no IANA considerations. 726 8. Acknowledgments 728 The authors acknowledge the contributions of Dr I Biswas, Mr Ziaul 729 Hossain in supporting the evaluation of CWV and for their help in 730 developing the mechanisms proposed in this draft. We also 731 acknowledge comments received from the Internet Congestion Control 732 Research Group, in particular Yuchung Cheng, Mirja Kuehlewind, Joe 733 Touch, and Mark Allman. This work was part-funded by the European 734 Community under its Seventh Framework Programme through the Reducing 735 Internet Transport Latency (RITE) project (ICT-317700). 737 9. Author Notes 739 RFC-Editor note: please remove this section prior to publication. 741 9.1. Other related work 743 RFC-Editor note: please remove this section prior to publication. 745 There are several issues to be discussed more widely: 747 o There are potential interactions with the Experimental update in 748 [RFC6928] that raises the TCP initial Window to ten segments, do 749 these cases need to be elaborated? 751 This relates to the Experimental specification for increasing 752 the TCP IW defined in RFC 6928. 754 The two methods have different functions and different response 755 to loss/congestion. 757 RFC 6928 proposes an experimental update to TCP that would 758 increase the IW to ten segments. This would allow faster 759 opening of the cwnd, and also a large (same size) restart 760 window. This approach is based on the assumption that many 761 forward paths can sustain bursts of up to ten segments without 762 (appreciable) loss. Such a significant increase in cwnd must 763 be matched with an equally large reduction of cwnd if loss/ 764 congestion is detected, and such a congestion indication is 765 likely to require future use of IW=10 to be disabled for this 766 path for some time. This guards against the unwanted behaviour 767 of a series of short flows continuously flooding a network path 768 without network congestion feedback. 770 In contrast, this document proposes an update with a rationale 771 that relies on recent previous path history to select an 772 appropriate cwnd after restart. 774 The behaviour differs in three ways: 776 1) For applications that send little initially, new-cwv may 777 constrain more than RFC 6928, but would not require the 778 connection to reset any path information when a restart 779 incurred loss. In contrast, new-cwv would allow the TCP 780 connection to preserve the cached cwnd, any loss, would impact 781 cwnd, but not impact other flows. 783 2) For applications that utilise more capacity than provided by 784 a cwnd of 10 segments, this method would permit a larger 785 restart window compared to a restart using the method in RFC 786 6928. This is justified by the recent path history. 788 3) new-CWV is attended to also be used for rate-limited 789 applications, where the application sends, but does not seek to 790 fully utilise the cwnd. In this case, new-cwv constrains the 791 cwnd to that justified by the recent path history. The 792 performance trade-offs are hence different, and it would be 793 possible to enable new-cwv when also using the method in RFC 794 6928, and yield benefits. 796 o There is potential overlap with the Laminar proposal (draft- 797 mathis-tcpm-tcp-laminar) 798 The current draft was intended as a standards-track update to 799 TCP, rather than a new transport variant. At least, it would 800 be good to understand how the two interact and whether there is 801 a possibility of a single method. 803 o There is potential performance loss in loss of a short burst 804 (off list with M Allman) 806 A sender can transmit several segments then become idle. If 807 the first segments are all ACK'ed the ssthresh collapses to a 808 small value (no new data is sent by the idle sender). Loss of 809 the later data results in congestion (e.g. maybe a RED drop or 810 some other cause, rather than the maximum rate of this flow). 811 When the sender performs loss recovery it may have an 812 appreciable pipeACK and cwnd, but a very low FlightSize - the 813 Standard algorithm results in an unusually low cwnd ((1/2)* 814 FlightSize). 816 A constant rate flow would have maintained a FlightSize 817 appropriate to pipeACK (cwnd if it is a bulk flow). 819 This could be fixed by adding a new state variable? It could 820 also be argued this is a corner case (e.g. loss of only the 821 last segments would have resulted in RTO), the impact could be 822 significant. 824 o There is potential interaction with TCP Control Block Sharing(M 825 Welzl) 827 An application that is non-validated can accumulate a cwnd that 828 is larger than the actual capacity. Is this a fair value to 829 use in TCB sharing? 831 We propose that TCB sharing should use the pipeACK in place of 832 cwnd when a TCP sender is in the Non-validated phase. This 833 value better reflects the capacity that the flow has utilised 834 in the network path. 836 9.2. Revision notes 838 RFC-Editor note: please remove this section prior to publication. 840 Draft 03 was submitted to ICCRG to receive comments and feedback. 842 Draft 04 contained the first set of clarifications after feedback: 844 o Changed name to application limited and used the term rate-limited 845 in all places. 847 o Added justification and many minor changes suggested on the list. 849 o Added text to tie-in with more accurate ECN marking. 851 o Added ref to Hug01 853 Draft 05 contained various updates: 855 o New text to redefine how to measure the acknowledged pipe, 856 differentiating this from the FlightSize, and hence avoiding 857 previous issues with infrequent large bursts of data not being 858 validated. A key point new feature is that pipeACK only triggers 859 leaving the NVP after the size of the pipe has been acknowledged. 860 This removed the need for hysteresis. 862 o Reduction values were changed to 1/2, following analysis of 863 suggestions from ICCRG. This also sets the "target" cwnd as twice 864 the used rate for non-validated case. 866 o Introduced a symbolic name (NVP) to denote the 5 minute period. 868 Draft 06 contained various updates: 870 o Required reset of pipeACK after congestion. 872 o Added comment on the effect of congestion after a short burst (M. 873 Allman). 875 o Correction of minor Typos. 877 WG draft 00 contained various updates: 879 o Updated initialisation of pipeACK to maximum value. 881 o Added note on intended status still to be determined. 883 WG draft 01 contained: 885 o Added corrections from Richard Scheffenegger. 887 o Raffaello Secchi added to the mechanism, based on implementation 888 experience. 890 o Removed that the requirement for the method to use TCP SACK option 892 o Although it may be desirable to use SACK, this is not essential to 893 the algorithm. 895 o Added the notion of the sampling period to accommodate large rate 896 variations and ensure that the method is stable. This algorithm 897 to be validated through implementation. 899 WG draft 02 contained: 901 o Clarified language around pipeACK variable and pipeACK sample - 902 Feedback from Aris Angelogiannopoulos. 904 WG draft 03 contained: 906 o Editorial corrections - Feedback from Anna Brunstrom. 908 o An adjustment to the procedure at the start and end of Reoloss 909 recovery to align the two equations. 911 o Further clarification of the "undefined" value of the pipeACK 912 variable. 914 WG draft 04 contained: 916 o Editorial corrections. 918 o Introduced the "cwnd-limited" term. 920 o An adjustment to the procedure at the start of a cwnd-limited 921 phase - the new text is intended to ensure that new-cwv is not 922 unnecessarily more conservative than standard TCP when the flow is 923 cwnd-limited. This resolves two issues: first it prevents 924 pathologies in which pipeACK increases slowly and erratically. It 925 also ensures that performance of bulk applications is not 926 significantly impacted when using the method. 928 o Clearly identifies that pacing (or equivalent) is requiring during 929 the NVP to control burstiness. New section added. 931 WG draft 05 contained: 933 o Clarification to first two bullets in Section 4.4 describing cwnd- 934 limited, to explain these are really alternates to the same case. 936 o Section giving implementation examples was restructured to clarify 937 there are two methods described. 939 o Cross References to sections updated - thanks to comments from 940 Martin Winbjoerk and Tim Wicinski. 942 WG draft 06 contained: 944 o The section giving implementation examples was restructured to 945 clarify there are two methods described. 947 o Justification of design decisions. 949 o Re-organised text to improve clarity of argument. 951 WG draft 07 contained: 953 o Updated publication date. 955 o Text on noting that cwnd shouldn't ever be made negative. 957 o Updated text on ECN to clarify the process where R is a reduction 958 based on ECN marks. 960 WG draft 08 contained: 962 o Removed description of how to use Accurate ECN feedback. It is 963 not clear that this document should specify a usage of a mechanism 964 that has not been fully defined. Accurate ECN may lead to 965 different congestion responses and these will need to be defined 966 in the CC specifications for using Accurate ECN. 968 10. References 970 10.1. Normative References 972 [RFC0793] Postel, J., "Transmission Control Protocol", STD 7, RFC 973 793, September 1981. 975 [RFC2018] Mathis, M., Mahdavi, J., Floyd, S., and A. Romanow, "TCP 976 Selective Acknowledgment Options", RFC 2018, October 1996. 978 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 979 Requirement Levels", BCP 14, RFC 2119, March 1997. 981 [RFC2861] Handley, M., Padhye, J., and S. Floyd, "TCP Congestion 982 Window Validation", RFC 2861, June 2000. 984 [RFC5681] Allman, M., Paxson, V., and E. Blanton, "TCP Congestion 985 Control", RFC 5681, September 2009. 987 [RFC6675] Blanton, E., Allman, M., Wang, L., Jarvinen, I., Kojo, M., 988 and Y. Nishida, "A Conservative Loss Recovery Algorithm 989 Based on Selective Acknowledgment (SACK) for TCP", RFC 990 6675, August 2012. 992 10.2. Informative References 994 [All05] Allman, M. and E. Blanton, "Notes on burst mitigation for 995 transport protocols", March 2005. 997 [Bis08] Biswas, I. and G. Fairhurst, "A Practical Evaluation of 998 Congestion Window Validation Behaviour, 9th Annual 999 Postgraduate Symposium in the Convergence of 1000 Telecommunications, Networking and Broadcasting (PGNet), 1001 Liverpool, UK", June 2008. 1003 [Bis10] Biswas, I., Sathiaseelan, A., Secchi, R., and G. 1004 Fairhurst, "Analysing TCP for Bursty Traffic, Int'l J. of 1005 Communications, Network and System Sciences, 7(3)", June 1006 2010. 1008 [Bis11] Biswas, I., "PhD Thesis, Internet congestion control for 1009 variable rate TCP traffic, School of Engineering, 1010 University of Aberdeen", June 2011. 1012 [Fai12] Sathiaseelan, A., Secchi, R., Fairhurst, G., and I. 1013 Biswas, "Enhancing TCP Performance to support Variable- 1014 Rate Traffic, 2nd Capacity Sharing Workshop, ACM CoNEXT, 1015 Nice, France, 10th December 2012.", June 2008. 1017 [Hug01] Hughes, A., Touch, J., and J. Heidemann, "Issues in TCP 1018 Slow-Start Restart After Idle (Work-in-Progress)", 1019 December 2001. 1021 [Liu07] Liu, D., Allman, M., Jiny, S., and L. Wang, "Congestion 1022 Control without a Startup Phase, 5th International 1023 Workshop on Protocols for Fast Long-Distance Networks 1024 (PFLDnet), Los Angeles, California, USA", February 2007. 1026 [RFC2616] Fielding, R., Gettys, J., Mogul, J., Frystyk, H., 1027 Masinter, L., Leach, P., and T. Berners-Lee, "Hypertext 1028 Transfer Protocol -- HTTP/1.1", RFC 2616, June 1999. 1030 [RFC6298] Paxson, V., Allman, M., Chu, J., and M. Sargent, 1031 "Computing TCP's Retransmission Timer", RFC 6298, June 1032 2011. 1034 [RFC6928] Chu, J., Dukkipati, N., Cheng, Y., and M. Mathis, 1035 "Increasing TCP's Initial Window", RFC 6928, April 2013. 1037 Authors' Addresses 1039 Godred Fairhurst 1040 University of Aberdeen 1041 School of Engineering 1042 Fraser Noble Building 1043 Aberdeen, Scotland AB24 3UE 1044 UK 1046 Email: gorry@erg.abdn.ac.uk 1047 URI: http://www.erg.abdn.ac.uk 1049 Arjuna Sathiaseelan 1050 University of Aberdeen 1051 School of Engineering 1052 Fraser Noble Building 1053 Aberdeen, Scotland AB24 3UE 1054 UK 1056 Email: arjuna@erg.abdn.ac.uk 1057 URI: http://www.erg.abdn.ac.uk 1059 Raffaello Secchi 1060 University of Aberdeen 1061 School of Engineering 1062 Fraser Noble Building 1063 Aberdeen, Scotland AB24 3UE 1064 UK 1066 Email: raffaello@erg.abdn.ac.uk 1067 URI: http://www.erg.abdn.ac.uk