idnits 2.17.1 draft-ietf-avtcore-rtp-circuit-breakers-15.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 updates RFC3550, but the abstract doesn't seem to mention this, which it should. Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year (Using the creation date from RFC3550, updated by this document, for RFC5378 checks: 1998-04-07) -- 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 (April 30, 2016) is 2916 days in the past. Is this intentional? Checking references for intended status: Proposed Standard ---------------------------------------------------------------------------- (See RFCs 3967 and 4897 for information about using normative references to lower-maturity documents in RFCs) == Outdated reference: A later version (-18) exists of draft-ietf-tsvwg-rtcweb-qos-15 -- Obsolete informational reference (is this intentional?): RFC 5405 (Obsoleted by RFC 8085) Summary: 0 errors (**), 0 flaws (~~), 2 warnings (==), 4 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 AVTCORE Working Group C. Perkins 3 Internet-Draft University of Glasgow 4 Updates: 3550 (if approved) V. Singh 5 Intended status: Standards Track callstats.io 6 Expires: November 1, 2016 April 30, 2016 8 Multimedia Congestion Control: Circuit Breakers for Unicast RTP Sessions 9 draft-ietf-avtcore-rtp-circuit-breakers-15 11 Abstract 13 The Real-time Transport Protocol (RTP) is widely used in telephony, 14 video conferencing, and telepresence applications. Such applications 15 are often run on best-effort UDP/IP networks. If congestion control 16 is not implemented in these applications, then network congestion can 17 lead to uncontrolled packet loss, and a resulting deterioration of 18 the user's multimedia experience. The congestion control algorithm 19 acts as a safety measure, stopping RTP flows from using excessive 20 resources, and protecting the network from overload. At the time of 21 this writing, however, while there are several proprietary solutions, 22 there is no standard algorithm for congestion control of interactive 23 RTP flows. 25 This document does not propose a congestion control algorithm. It 26 instead defines a minimal set of RTP circuit breakers: conditions 27 under which an RTP sender needs to stop transmitting media data, to 28 protect the network from excessive congestion. It is expected that, 29 in the absence of long-lived excessive congestion, RTP applications 30 running on best-effort IP networks will be able to operate without 31 triggering these circuit breakers. Future RTP congestion control 32 specifications will be expected to operate within the constraints 33 defined by these circuit breakers. 35 Status of This Memo 37 This Internet-Draft is submitted in full conformance with the 38 provisions of BCP 78 and BCP 79. 40 Internet-Drafts are working documents of the Internet Engineering 41 Task Force (IETF). Note that other groups may also distribute 42 working documents as Internet-Drafts. The list of current Internet- 43 Drafts is at http://datatracker.ietf.org/drafts/current/. 45 Internet-Drafts are draft documents valid for a maximum of six months 46 and may be updated, replaced, or obsoleted by other documents at any 47 time. It is inappropriate to use Internet-Drafts as reference 48 material or to cite them other than as "work in progress." 49 This Internet-Draft will expire on November 1, 2016. 51 Copyright Notice 53 Copyright (c) 2016 IETF Trust and the persons identified as the 54 document authors. All rights reserved. 56 This document is subject to BCP 78 and the IETF Trust's Legal 57 Provisions Relating to IETF Documents 58 (http://trustee.ietf.org/license-info) in effect on the date of 59 publication of this document. Please review these documents 60 carefully, as they describe your rights and restrictions with respect 61 to this document. Code Components extracted from this document must 62 include Simplified BSD License text as described in Section 4.e of 63 the Trust Legal Provisions and are provided without warranty as 64 described in the Simplified BSD License. 66 Table of Contents 68 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 2 69 2. Background . . . . . . . . . . . . . . . . . . . . . . . . . 3 70 3. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . 6 71 4. RTP Circuit Breakers for Systems Using the RTP/AVP Profile . 7 72 4.1. RTP/AVP Circuit Breaker #1: RTCP Timeout . . . . . . . . 9 73 4.2. RTP/AVP Circuit Breaker #2: Media Timeout . . . . . . . . 10 74 4.3. RTP/AVP Circuit Breaker #3: Congestion . . . . . . . . . 12 75 4.4. RTP/AVP Circuit Breaker #4: Media Usability . . . . . . . 15 76 4.5. Ceasing Transmission . . . . . . . . . . . . . . . . . . 16 77 5. RTP Circuit Breakers and the RTP/AVPF and RTP/SAVPF Profiles 17 78 6. Impact of RTCP Extended Reports (XR) . . . . . . . . . . . . 18 79 7. Impact of Explicit Congestion Notification (ECN) . . . . . . 19 80 8. Impact of Bundled Media and Layered Coding . . . . . . . . . 19 81 9. Security Considerations . . . . . . . . . . . . . . . . . . . 20 82 10. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 20 83 11. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 20 84 12. References . . . . . . . . . . . . . . . . . . . . . . . . . 20 85 12.1. Normative References . . . . . . . . . . . . . . . . . . 21 86 12.2. Informative References . . . . . . . . . . . . . . . . . 21 87 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 24 89 1. Introduction 91 The Real-time Transport Protocol (RTP) [RFC3550] is widely used in 92 voice-over-IP, video teleconferencing, and telepresence systems. 93 Many of these systems run over best-effort UDP/IP networks, and can 94 suffer from packet loss and increased latency if network congestion 95 occurs. Designing effective RTP congestion control algorithms, to 96 adapt the transmission of RTP-based media to match the available 97 network capacity, while also maintaining the user experience, is a 98 difficult but important problem. Many such congestion control and 99 media adaptation algorithms have been proposed, but to date there is 100 no consensus on the correct approach, or even that a single standard 101 algorithm is desirable. 103 This memo does not attempt to propose a new RTP congestion control 104 algorithm. Instead, we propose a small set of RTP circuit breakers: 105 mechanisms that terminate RTP flows in conditions under which there 106 is general agreement that serious network congestion is occurring. 107 The RTP circuit breakers proposed in this memo are a specific 108 instance of the general class of network transport circuit breakers 109 [I-D.ietf-tsvwg-circuit-breaker], designed to act as a protection 110 mechanism of last resort to avoid persistent excessive congestion. 111 It is expected that future standards-track congestion control 112 algorithms for RTP will operate within the envelope defined by this 113 memo. 115 2. Background 117 We consider congestion control for unicast RTP traffic flows. This 118 is the problem of adapting the transmission of an audio/visual data 119 flow, encapsulated within an RTP transport session, from one sender 120 to one receiver, so that it does not use more capacity than is 121 available along the network path. Such adaptation needs to be done 122 in a way that limits the disruption to the user experience caused by 123 both packet loss and excessive rate changes. Congestion control for 124 multicast flows is outside the scope of this memo. Multicast traffic 125 needs different solutions, since the available capacity estimator for 126 a group of receivers will differ from that for a single receiver, and 127 because multicast congestion control has to consider issues of 128 fairness across groups of receivers that do not apply to unicast 129 flows. 131 Congestion control for unicast RTP traffic can be implemented in one 132 of two places in the protocol stack. One approach is to run the RTP 133 traffic over a congestion controlled transport protocol, for example 134 over TCP, and to adapt the media encoding to match the dictates of 135 the transport-layer congestion control algorithm. This is safe for 136 the network, but can be suboptimal for the media quality unless the 137 transport protocol is designed to support real-time media flows. We 138 do not consider this class of applications further in this memo, as 139 their network safety is guaranteed by the underlying transport. 141 Alternatively, RTP flows can be run over a non-congestion controlled 142 transport protocol, for example UDP, performing rate adaptation at 143 the application layer based on RTP Control Protocol (RTCP) feedback. 144 With a well-designed, network-aware, application, this allows highly 145 effective media quality adaptation, but there is potential to cause 146 persistent congestion in the network if the application does not 147 adapt its sending rate in a timely and effective manner. We consider 148 this class of applications in this memo. 150 Congestion control relies on monitoring the delivery of a media flow, 151 and responding to adapt the transmission of that flow when there are 152 signs that the network path is congested. Network congestion can be 153 detected in one of three ways: 1) a receiver can infer the onset of 154 congestion by observing an increase in one-way delay caused by queue 155 build-up within the network; 2) if Explicit Congestion Notification 156 (ECN) [RFC3168] is supported, the network can signal the presence of 157 congestion by marking packets using ECN Congestion Experienced (CE) 158 marks (this could potentially be augmented by mechanisms such as 159 ConEX [RFC7713], or other future protocol extensions for network 160 signalling of congestion); or 3) in the extreme case, congestion will 161 cause packet loss that can be detected by observing a gap in the 162 received RTP sequence numbers. 164 Once the onset of congestion is observed, the receiver has to send 165 feedback to the sender to indicate that the transmission rate needs 166 to be reduced. How the sender reduces the transmission rate is 167 highly dependent on the media codec being used, and is outside the 168 scope of this memo. 170 There are several ways in which a receiver can send feedback to a 171 media sender within the RTP framework: 173 o The base RTP specification [RFC3550] defines RTCP Reception Report 174 (RR) packets to convey reception quality feedback information, and 175 Sender Report (SR) packets to convey information about the media 176 transmission. RTCP SR packets contain data that can be used to 177 reconstruct media timing at a receiver, along with a count of the 178 total number of octets and packets sent. RTCP RR packets report 179 on the fraction of packets lost in the last reporting interval, 180 the cumulative number of packets lost, the highest sequence number 181 received, and the inter-arrival jitter. The RTCP RR packets also 182 contain timing information that allows the sender to estimate the 183 network round trip time (RTT) to the receivers. RTCP reports are 184 sent periodically, with the reporting interval being determined by 185 the number of SSRCs used in the session and a configured session 186 bandwidth estimate (the number of synchronisation sources (SSRCs) 187 used is usually two in a unicast session, one for each 188 participant, but can be greater if the participants send multiple 189 media streams). The interval between reports sent from each 190 receiver tends to be on the order of a few seconds on average, 191 although it varies with the session bandwidth, and sub-second 192 reporting intervals are possible in high bandwidth sessions, and 193 it is randomised to avoid synchronisation of reports from multiple 194 receivers. RTCP RR packets allow a receiver to report ongoing 195 network congestion to the sender. However, if a receiver detects 196 the onset of congestion part way through a reporting interval, the 197 base RTP specification contains no provision for sending the RTCP 198 RR packet early, and the receiver has to wait until the next 199 scheduled reporting interval. 201 o The RTCP Extended Reports (XR) [RFC3611] allow reporting of more 202 complex and sophisticated reception quality metrics, but do not 203 change the RTCP timing rules. RTCP extended reports of potential 204 interest for congestion control purposes are the extended packet 205 loss, discard, and burst metrics [RFC3611], [RFC7002], [RFC7097], 206 [RFC7003], [RFC6958]; and the extended delay metrics [RFC6843], 207 [RFC6798]. Other RTCP Extended Reports that could be helpful for 208 congestion control purposes might be developed in future. 210 o Rapid feedback about the occurrence of congestion events can be 211 achieved using the Extended RTP Profile for RTCP-Based Feedback 212 (RTP/AVPF) [RFC4585] (or its secure variant, RTP/SAVPF [RFC5124]) 213 in place of the RTP/AVP profile [RFC3551]. This modifies the RTCP 214 timing rules to allow RTCP reports to be sent early, in some cases 215 immediately, provided the RTCP transmission rate keeps within its 216 bandwidth allocation. It also defines transport-layer feedback 217 messages, including negative acknowledgements (NACKs), that can be 218 used to report on specific congestion events. RTP Codec Control 219 Messages [RFC5104] extend the RTP/AVPF profile with additional 220 feedback messages that can be used to influence that way in which 221 rate adaptation occurs, but do not further change the dynamics of 222 how rapidly feedback can be sent. Use of the RTP/AVPF profile is 223 dependent on signalling. 225 o Finally, Explicit Congestion Notification (ECN) for RTP over UDP 226 [RFC6679] can be used to provide feedback on the number of packets 227 that received an ECN Congestion Experienced (CE) mark. This RTCP 228 extension builds on the RTP/AVPF profile to allow rapid congestion 229 feedback when ECN is supported. 231 In addition to these mechanisms for providing feedback, the sender 232 can include an RTP header extension in each packet to record packet 233 transmission times [RFC5450]. Accurate transmission timestamps can 234 be helpful for estimating queuing delays, to get an early indication 235 of the onset of congestion. 237 Taken together, these various mechanisms allow receivers to provide 238 feedback on the senders when congestion events occur, with varying 239 degrees of timeliness and accuracy. The key distinction is between 240 systems that use only the basic RTCP mechanisms, without RTP/AVPF 241 rapid feedback, and those that use the RTP/AVPF extensions to respond 242 to congestion more rapidly. 244 3. Terminology 246 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 247 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 248 document are to be interpreted as described in RFC 2119 [RFC2119]. 249 This interpretation of these key words applies only when written in 250 ALL CAPS. Mixed- or lower-case uses of these key words are not to be 251 interpreted as carrying special significance in this memo. 253 The definition of the RTP circuit breaker is specified in terms of 254 the following variables: 256 o Td is the deterministic RTCP reporting interval, as defined in 257 Section 6.3.1 of [RFC3550]. 259 o Tdr is the sender's estimate of the deterministic RTCP reporting 260 interval, Td, calculated by a receiver of the data it is sending. 261 Tdr is not known at the sender, but can be estimated by executing 262 the algorithm in Section 6.2 of [RFC3550] using the average RTCP 263 packet size seen at the sender, the number of members reported in 264 the receiver's SR/RR report blocks, and whether the receiver is 265 sending SR or RR packets. Tdr is recalculated when each new RTCP 266 SR/RR report is received, but the media timeout circuit breaker 267 (see Section 4.2) is only reconsidered when Tdr increases. 269 o Tr is the network round-trip time, calculated by the sender using 270 the algorithm in Section 6.4.1 of [RFC3550] and smoothed using an 271 exponentially weighted moving average as Tr = (0.8 * Tr) + (0.2 * 272 Tr_new) where Tr_new is the latest RTT estimate obtained from an 273 RTCP report. The weight is chosen so old estimates decay over k 274 intervals. 276 o k is the non-reporting threshold (see Section 4.2). 278 o Tf is the media framing interval at the sender. For applications 279 sending at a constant frame rate, Tf is the inter-frame interval. 280 For applications that switch between a small set of possible frame 281 rates, for example when sending speech with comfort noise, where 282 comfort noise frames are sent less often than speech frames, Tf is 283 set to the longest of the inter-frame intervals of the different 284 frame rates. For applications that send periodic frames but 285 dynamically vary their frame rate, Tf is set to the largest inter- 286 frame interval used in the last 10 seconds. For applications that 287 send less than one frame every 10 seconds, or that have no concept 288 of periodic frames (e.g., text conversation [RFC4103], or pointer 289 events [RFC2862]), Tf is set to the time interval since the 290 previous frame when each frame is sent. 292 o G is the frame group size. That is, the number of frames that are 293 coded together based on a particular sending rate setting. If the 294 codec used by the sender can change its rate on each frame, G = 1; 295 otherwise G is set to the number of frames before the codec can 296 adjust to the new rate. For codecs that have the concept of a 297 group-of-pictures (GoP), G is likely the GoP length. 299 o T_rr_interval is the minimal interval between RTCP reports, as 300 defined in Section 3.4 of [RFC4585]; it is only meaningful for 301 implementations of RTP/AVPF profile [RFC4585] or the RTP/SAVPF 302 profile [RFC5124]. 304 o X is the estimated throughput a TCP connection would achieve over 305 a path, in bytes per second. 307 o s is the size of RTP packets being sent, in bytes. If the RTP 308 packets being sent vary in size, then the average size over the 309 packet comprising the last 4 * G frames MUST be used (this is 310 intended to be comparable to the four loss intervals used in 311 [RFC5348]). 313 o p is the loss event rate, between 0.0 and 1.0, that would be seen 314 by a TCP connection over a particular path. When used in the RTP 315 congestion circuit breaker, this is approximated as described in 316 Section 4.3. 318 o t_RTO is the retransmission timeout value that would be used by a 319 TCP connection over a particular path, in seconds. This MUST be 320 approximated using t_RTO = 4 * Tr when used as part of the RTP 321 congestion circuit breaker. 323 o b is the number of packets that are acknowledged by a single TCP 324 acknowledgement. Following [RFC5348], it is RECOMMENDED that the 325 value b = 1 is used as part of the RTP congestion circuit breaker. 327 4. RTP Circuit Breakers for Systems Using the RTP/AVP Profile 329 The feedback mechanisms defined in [RFC3550] and available under the 330 RTP/AVP profile [RFC3551] are the minimum that can be assumed for a 331 baseline circuit breaker mechanism that is suitable for all unicast 332 applications of RTP. Accordingly, for an RTP circuit breaker to be 333 useful, it needs to be able to detect that an RTP flow is causing 334 excessive congestion using only basic RTCP features, without needing 335 RTCP XR feedback or the RTP/AVPF profile for rapid RTCP reports. 337 RTCP is a fundamental part of the RTP protocol, and the mechanisms 338 described here rely on the implementation of RTCP. Implementations 339 that claim to support RTP, but that do not implement RTCP, will be 340 unable to use the circuit breaker mechanisms described in this memo. 341 Such implementations SHOULD NOT be used on networks that might be 342 subject to congestion unless equivalent mechanisms are defined using 343 some non-RTCP feedback channel to report congestion and signal 344 circuit breaker conditions. 346 The RTCP timeout circuit breaker (Section 4.1) will trigger if an 347 implementation of this memo attempts to interwork with an endpoint 348 that does not support RTCP. Implementations that sometimes need to 349 interwork with endpoints that do not support RTCP need to disable the 350 RTP circuit breakers if they don't receive some confirmation via 351 signalling that the remote endpoint implements RTCP (the presence of 352 an SDP "a=rtcp:" attribute in an answer might be such an indication). 353 The RTP Circuit Breaker SHOULD NOT be disabled on networks that might 354 be subject to congestion, unless equivalent mechanisms are defined 355 using some non-RTCP feedback channel to report congestion and signal 356 circuit breaker conditions [I-D.ietf-tsvwg-circuit-breaker]. 358 Three potential congestion signals are available from the basic RTCP 359 SR/RR packets and are reported for each SSRC in the RTP session: 361 1. The sender can estimate the network round-trip time once per RTCP 362 reporting interval, based on the contents and timing of RTCP SR 363 and RR packets. 365 2. Receivers report a jitter estimate (the statistical variance of 366 the RTP data packet inter-arrival time) calculated over the RTCP 367 reporting interval. Due to the nature of the jitter calculation 368 ([RFC3550], section 6.4.4), the jitter is only meaningful for RTP 369 flows that send a single data packet for each RTP timestamp value 370 (i.e., audio flows, or video flows where each packet comprises 371 one video frame). 373 3. Receivers report the fraction of RTP data packets lost during the 374 RTCP reporting interval, and the cumulative number of RTP packets 375 lost over the entire RTP session. 377 These congestion signals limit the possible circuit breakers, since 378 they give only limited visibility into the behaviour of the network. 380 RTT estimates are widely used in congestion control algorithms, as a 381 proxy for queuing delay measures in delay-based congestion control or 382 to determine connection timeouts. RTT estimates derived from RTCP SR 383 and RR packets sent according to the RTP/AVP timing rules are too 384 infrequent to be useful for congestion control, and don't give enough 385 information to distinguish a delay change due to routing updates from 386 queuing delay caused by congestion. Accordingly, we cannot use the 387 RTT estimate alone as an RTP circuit breaker. 389 Increased jitter can be a signal of transient network congestion, but 390 in the highly aggregated form reported in RTCP RR packets, it offers 391 insufficient information to estimate the extent or persistence of 392 congestion. Jitter reports are a useful early warning of potential 393 network congestion, but provide an insufficiently strong signal to be 394 used as a circuit breaker. 396 The remaining congestion signals are the packet loss fraction and the 397 cumulative number of packets lost. If considered carefully, and over 398 an appropriate time frame to distinguish transient problems from long 399 term issues [I-D.ietf-tsvwg-circuit-breaker], these can be effective 400 indicators that persistent excessive congestion is occurring in 401 networks where packet loss is primarily due to queue overflows, 402 although loss caused by non-congestive packet corruption can distort 403 the result in some networks. TCP congestion control [RFC5681] 404 intentionally tries to fill the router queues, and uses the resulting 405 packet loss as congestion feedback. An RTP flow competing with TCP 406 traffic will therefore expect to see a non-zero packet loss fraction, 407 and some variation in queuing latency, in normal operation when 408 sharing a path with other flows, that needs to be accounted for when 409 determining the circuit breaker threshold 410 [I-D.ietf-tsvwg-circuit-breaker]. This behaviour of TCP is reflected 411 in the congestion circuit breaker below, and will affect the design 412 of any RTP congestion control protocol. 414 Two packet loss regimes can be observed: 1) RTCP RR packets show a 415 non-zero packet loss fraction, while the extended highest sequence 416 number received continues to increment; and 2) RR packets show a loss 417 fraction of zero, but the extended highest sequence number received 418 does not increment even though the sender has been transmitting RTP 419 data packets. The former corresponds to the TCP congestion avoidance 420 state, and indicates a congested path that is still delivering data; 421 the latter corresponds to a TCP timeout, and is most likely due to a 422 path failure. A third condition is that data is being sent but no 423 RTCP feedback is received at all, corresponding to a failure of the 424 reverse path. We derive circuit breaker conditions for these loss 425 regimes in the following. 427 4.1. RTP/AVP Circuit Breaker #1: RTCP Timeout 429 An RTCP timeout can occur when RTP data packets are being sent, but 430 there are no RTCP reports returned from the receiver. This is either 431 due to a failure of the receiver to send RTCP reports, or a failure 432 of the return path that is preventing those RTCP reporting from being 433 delivered. In either case, it is not safe to continue transmission, 434 since the sender has no way of knowing if it is causing congestion. 436 An RTP sender that has not received any RTCP SR or RTCP RR packets 437 reporting on the SSRC it is using, for a time period of at least 438 three times its deterministic RTCP reporting interval, Td, without 439 the randomization factor, and using the fixed minimum interval of 440 Tmin=5 seconds, SHOULD cease transmission (see Section 4.5). The 441 rationale for this choice of timeout is as described in Section 6.2 442 of [RFC3550] ("so that implementations which do not use the reduced 443 value for transmitting RTCP packets are not timed out by other 444 participants prematurely"), as updated by Section 6.1.4 of 445 [I-D.ietf-avtcore-rtp-multi-stream] to account for the use of the 446 RTP/AVPF profile [RFC4585] or the RTP/SAVPF profile [RFC5124]. 448 To reduce the risk of premature timeout, implementations SHOULD NOT 449 configure the RTCP bandwidth such that Td is larger than 5 seconds. 450 Similarly, implementations that use the RTP/AVPF profile [RFC4585] or 451 the RTP/SAVPF profile [RFC5124] SHOULD NOT configure T_rr_interval to 452 values larger than 4 seconds (the reduced limit for T_rr_interval 453 follows Section 6.1.3 of [I-D.ietf-avtcore-rtp-multi-stream]). 455 The choice of three RTCP reporting intervals as the timeout is made 456 following Section 6.3.5 of RFC 3550 [RFC3550]. This specifies that 457 participants in an RTP session will timeout and remove an RTP sender 458 from the list of active RTP senders if no RTP data packets have been 459 received from that RTP sender within the last two RTCP reporting 460 intervals. Using a timeout of three RTCP reporting intervals is 461 therefore large enough that the other participants will have timed 462 out the sender if a network problem stops the data packets it is 463 sending from reaching the receivers, even allowing for loss of some 464 RTCP packets. 466 If a sender is transmitting a large number of RTP media streams, such 467 that the corresponding RTCP SR or RR packets are too large to fit 468 into the network MTU, the receiver will generate RTCP SR or RR 469 packets in a round-robin manner. In this case, the sender SHOULD 470 treat receipt of an RTCP SR or RR packet corresponding to any SSRC it 471 sent on the same 5-tuple of source and destination IP address, port, 472 and protocol, as an indication that the receiver and return path are 473 working, preventing the RTCP timeout circuit breaker from triggering. 475 4.2. RTP/AVP Circuit Breaker #2: Media Timeout 477 If RTP data packets are being sent, but the RTCP SR or RR packets 478 reporting on that SSRC indicate a non-increasing extended highest 479 sequence number received, this is an indication that those RTP data 480 packets are not reaching the receiver. This could be a short-term 481 issue affecting only a few RTP packets, perhaps caused by a slow to 482 open firewall or a transient connectivity problem, but if the issue 483 persists, it is a sign of a more ongoing and significant problem (a 484 "media timeout"). 486 The time needed to declare a media timeout depends on the parameters 487 Tdr, Tr, Tf, and on the non-reporting threshold k. The value of k is 488 chosen so that when Tdr is large compared to Tr and Tf, receipt of at 489 least k RTCP reports with non-increasing extended highest sequence 490 number received gives reasonable assurance that the forward path has 491 failed, and that the RTP data packets have not been lost by chance. 492 The RECOMMENDED value for k is 5 reports. 494 When Tdr < Tf, then RTP data packets are being sent at a rate less 495 than one per RTCP reporting interval of the receiver, so the extended 496 highest sequence number received can be expected to be non-increasing 497 for some receiver RTCP reporting intervals. Similarly, when Tdr < 498 Tr, some receiver RTCP reporting intervals might pass before the RTP 499 data packets arrive at the receiver, also leading to reports where 500 the extended highest sequence number received is non-increasing. 501 Both issues require the media timeout interval to be scaled relative 502 to the threshold, k. 504 The media timeout RTP circuit breaker is therefore as follows. When 505 starting sending, calculate MEDIA_TIMEOUT using: 507 MEDIA_TIMEOUT = ceil(k * max(Tf, Tr, Tdr) / Tdr) 509 When a sender receives an RTCP packet that indicates reception of the 510 media it has been sending, then it cancels the media timeout circuit 511 breaker. If it is still sending, then it MUST calculate a new value 512 for MEDIA_TIMEOUT, and set a new media timeout circuit breaker. 514 If a sender receives an RTCP packet indicating that its media was not 515 received, it MUST calculate a new value for MEDIA_TIMEOUT. If the 516 new value is larger than the previous, it replaces MEDIA_TIMEOUT with 517 the new value, extending the media timeout circuit breaker; otherwise 518 it keeps the original value of MEDIA_TIMEOUT. This process is known 519 as reconsidering the media timeout circuit breaker. 521 If MEDIA_TIMEOUT consecutive RTCP packets are received indicating 522 that the media being sent was not received, and the media timeout 523 circuit breaker has not been cancelled, then the media timeout 524 circuit breaker triggers. When the media timeout circuit breaker 525 triggers, the sender SHOULD cease transmission (see Section 4.5). 527 When stopping sending an RTP stream, a sender MUST cancel the 528 corresponding media timeout circuit breaker. 530 4.3. RTP/AVP Circuit Breaker #3: Congestion 532 If RTP data packets are being sent, and the corresponding RTCP SR or 533 RR packets show non-zero packet loss fraction and increasing extended 534 highest sequence number received, then those RTP data packets are 535 arriving at the receiver, but some degree of congestion is occurring. 536 The RTP/AVP profile [RFC3551] states that: 538 If best-effort service is being used, RTP receivers SHOULD monitor 539 packet loss to ensure that the packet loss rate is within 540 acceptable parameters. Packet loss is considered acceptable if a 541 TCP flow across the same network path and experiencing the same 542 network conditions would achieve an average throughput, measured 543 on a reasonable time scale, that is not less than the throughput 544 the RTP flow is achieving. This condition can be satisfied by 545 implementing congestion control mechanisms to adapt the 546 transmission rate (or the number of layers subscribed for a 547 layered multicast session), or by arranging for a receiver to 548 leave the session if the loss rate is unacceptably high. 550 The comparison to TCP cannot be specified exactly, but is intended 551 as an "order-of-magnitude" comparison in time scale and 552 throughput. The time scale on which TCP throughput is measured is 553 the round-trip time of the connection. In essence, this 554 requirement states that it is not acceptable to deploy an 555 application (using RTP or any other transport protocol) on the 556 best-effort Internet which consumes bandwidth arbitrarily and does 557 not compete fairly with TCP within an order of magnitude. 559 The phase "order of magnitude" in the above means within a factor of 560 ten, approximately. In order to implement this, it is necessary to 561 estimate the throughput a bulk TCP connection would achieve over the 562 path. For a long-lived TCP Reno connection, it has been shown that 563 the TCP throughput, X, in bytes per second, can be estimated using 564 [Padhye]: 566 s 567 X = ------------------------------------------------------------- 568 Tr*sqrt(2*b*p/3)+(t_RTO * (3*sqrt(3*b*p/8) * p * (1+32*p*p))) 570 This is the same approach to estimated TCP throughput that is used in 571 [RFC5348]. Under conditions of low packet loss the second term on 572 the denominator is small, so this formula can be approximated with 573 reasonable accuracy as follows [Mathis]: 575 s 576 X = ---------------- 577 Tr*sqrt(2*b*p/3) 579 It is RECOMMENDED that this simplified throughput equation be used, 580 since the reduction in accuracy is small, and it is much simpler to 581 calculate than the full equation. Measurements have shown that the 582 simplified TCP throughput equation is effective as an RTP circuit 583 breaker for multimedia flows sent to hosts on residential networks 584 using ADSL and cable modem links [Singh]. The data shows that the 585 full TCP throughput equation tends to be more sensitive to packet 586 loss and triggers the RTP circuit breaker earlier than the simplified 587 equation. Implementations that desire this extra sensitivity MAY use 588 the full TCP throughput equation in the RTP circuit breaker. Initial 589 measurements in LTE networks have shown that the extra sensitivity is 590 helpful in that environment, with the full TCP throughput equation 591 giving a more balanced circuit breaker response than the simplified 592 TCP equation [Sarker]; other networks might see similar behaviour. 594 No matter what TCP throughput equation is chosen, two parameters need 595 to be estimated and reported to the sender in order to calculate the 596 throughput: the round trip time, Tr, and the loss event rate, p (the 597 packet size, s, is known to the sender). The round trip time can be 598 estimated from RTCP SR and RR packets. This is done too infrequently 599 for accurate statistics, but is the best that can be done with the 600 standard RTCP mechanisms. 602 Report blocks in RTCP SR or RR packets contain the packet loss 603 fraction, rather than the loss event rate, so p cannot be reported 604 (TCP typically treats the loss of multiple packets within a single 605 RTT as one loss event, but RTCP RR packets report the overall 606 fraction of packets lost, and does not report when the packet losses 607 occurred). Using the loss fraction in place of the loss event rate 608 can overestimate the loss. We believe that this overestimate will 609 not be significant, given that we are only interested in order of 610 magnitude comparison ([Floyd] section 3.2.1 shows that the difference 611 is small for steady-state conditions and random loss, but using the 612 loss fraction is more conservative in the case of bursty loss). 614 The congestion circuit breaker is therefore: when a sender that is 615 transmitting at least one RTP packet every max(Tdr, Tr) seconds 616 receives an RTCP SR or RR packet that contains a report block for an 617 SSRC it is using, the sender MUST record the value of the fraction 618 lost field from the report block, and the time since the last report 619 block was received, for that SSRC. If more than CB_INTERVAL (see 620 below) report blocks have been received for that SSRC, the sender 621 MUST calculate the average fraction lost over the last CB_INTERVAL 622 reporting intervals, and then estimate the TCP throughput that would 623 be achieved over the path using the chosen TCP throughput equation 624 and the measured values of the round-trip time, Tr, the loss event 625 rate, p (approximated by the average fraction lost, as is described 626 below), and the packet size, s. The estimate of the TCP throughput, 627 X, is then compared with the actual sending rate of the RTP stream. 628 If the actual sending rate of the RTP stream is more than 10 * X, 629 then the congestion circuit breaker is triggered. 631 The average fraction lost is calculated based on the sum, over the 632 last CB_INTERVAL reporting intervals, of the fraction lost in each 633 reporting interval multiplied by the duration of the corresponding 634 reporting interval, divided by the total duration of the last 635 CB_INTERVAL reporting intervals. The CB_INTERVAL parameter is set 636 to: 638 CB_INTERVAL = 639 ceil(3*min(max(10*G*Tf, 10*Tr, 3*Tdr), max(15, 3*Td))/(3*Tdr)) 641 The parameters that feed into CB_INTERVAL are chosen to give the 642 congestion control algorithm time to react to congestion. They give 643 at least three RTCP reports, ten round trip times, and ten groups of 644 frames to adjust the rate to reduce the congestion to a reasonable 645 level. It is expected that a responsive congestion control algorithm 646 will begin to respond with the next group of frames after it receives 647 indication of congestion, so CB_INTERVAL ought to be a much longer 648 interval than the congestion response. 650 If the RTP/AVPF profile [RFC4585] or the RTP/SAVPF [RFC5124] is used, 651 and the T_rr_interval parameter is used to reduce the frequency of 652 regular RTCP reports, then the value Tdr in the above expression for 653 the CB_INTERVAL parameter MUST be replaced by max(T_rr_interval, 654 Tdr). 656 The CB_INTERVAL parameter is calculated on joining the session, and 657 recalculated on receipt of each RTCP packet, after checking whether 658 the media timeout circuit breaker or the congestion circuit breaker 659 has been triggered. 661 To ensure a timely response to persistent congestion, implementations 662 SHOULD NOT configure the RTCP bandwidth such that Tdr is larger than 663 5 seconds. Similarly, implementations that use the RTP/AVPF profile 664 [RFC4585] or the RTP/SAVPF profile [RFC5124] SHOULD NOT configure 665 T_rr_interval to values larger than 4 seconds (the reduced limit for 666 T_rr_interval follows Section 6.1.3 of 667 [I-D.ietf-avtcore-rtp-multi-stream]). 669 The rationale for enforcing a minimum sending rate below which the 670 congestion circuit breaker will not trigger is to avoid spurious 671 circuit breaker triggers when the number of packets sent per RTCP 672 reporting interval is small, and hence the fraction lost samples are 673 subject to measurement artefacts. The bound of at least one packet 674 every max(Tdr, Tr) seconds is derived from the one packet per RTT 675 minimum sending rate of TCP [RFC5405], adapted for use with RTP where 676 the RTCP reporting interval is decoupled from the network RTT. 678 When the congestion circuit breaker is triggered, the sender SHOULD 679 cease transmission (see Section 4.5). However, if the sender is able 680 to reduce its sending rate by a factor of (approximately) ten, then 681 it MAY first reduce its sending rate by this factor (or some larger 682 amount) to see if that resolves the congestion. If the sending rate 683 is reduced in this way and the congestion circuit breaker triggers 684 again after the next CB_INTERVAL RTCP reporting intervals, the sender 685 MUST then cease transmission. An example of such a rate reduction 686 might be a video conferencing system that backs off to sending audio 687 only, before completely dropping the call. If such a reduction in 688 sending rate resolves the congestion problem, the sender MAY 689 gradually increase the rate at which it sends data after a reasonable 690 amount of time has passed, provided it takes care not to cause the 691 problem to recur ("reasonable" is intentionally not defined here). 693 The RTCP reporting interval of the media sender does not affect how 694 quickly congestion circuit breaker can trigger. The timing is based 695 on the RTCP reporting interval of the receiver that generates the SR/ 696 RR packets from which the loss rate and RTT estimate are derived 697 (note that RTCP requires all participants in a session to have 698 similar reporting intervals, else the participant timeout rules in 699 [RFC3550] will not work, so this interval is likely similar to that 700 of the sender). If the incoming RTCP SR or RR packets are using a 701 reduced minimum RTCP reporting interval (as specified in Section 6.2 702 of RFC 3550 [RFC3550] or the RTP/AVPF profile [RFC4585]), then that 703 reduced RTCP reporting interval is used when determining if the 704 circuit breaker is triggered. 706 If there are more media streams that can be reported in a single RTCP 707 SR or RR packet, or if the size of a complete RTCP SR or RR packet 708 exceeds the network MTU, then the receiver will report on a subset of 709 sources in each reporting interval, with the subsets selected round- 710 robin across multiple intervals so that all sources are eventually 711 reported [RFC3550]. When generating such round-robin RTCP reports, 712 priority SHOULD be given to reports on sources that have high packet 713 loss rates, to ensure that senders are aware of network congestion 714 they are causing (this is an update to [RFC3550]). 716 4.4. RTP/AVP Circuit Breaker #4: Media Usability 718 Applications that use RTP are generally tolerant to some amount of 719 packet loss. How much packet loss can be tolerated will depend on 720 the application, media codec, and the amount of error correction and 721 packet loss concealment that is applied. There is an upper bound on 722 the amount of loss that can be corrected, however, beyond which the 723 media becomes unusable. Similarly, many applications have some upper 724 bound on the media capture to play-out latency that can be tolerated 725 before the application becomes unusable. The latency bound will 726 depend on the application, but typical values can range from the 727 order of a few hundred milliseconds for voice telephony and 728 interactive conferencing applications, up to several seconds for some 729 video-on-demand systems. 731 As a final circuit breaker, RTP senders SHOULD monitor the reported 732 packet loss and delay to estimate whether the media is likely to be 733 suitable for the intended purpose. If the packet loss rate and/or 734 latency is such that the media has become unusable, and has remained 735 unusable for a significant time period, then the application SHOULD 736 cease transmission. Similarly, receivers SHOULD monitor the quality 737 of the media they receive, and if the quality is unusable for a 738 significant time period, they SHOULD terminate the session. This 739 memo intentionally does not define a bound on the packet loss rate or 740 latency that will result in unusable media, as these are highly 741 application dependent. Similarly, the time period that is considered 742 significant is application dependent, but is likely on the order of 743 seconds, or tens of seconds. 745 Sending media that suffers from such high packet loss or latency that 746 it is unusable at the receiver is both wasteful of resources, and of 747 no benefit to the user of the application. It also is highly likely 748 to be congesting the network, and disrupting other applications. As 749 such, the congestion circuit breaker will almost certainly trigger to 750 stop flows where the media would be unusable due to high packet loss 751 or latency. However, in pathological scenarios where the congestion 752 circuit breaker does not stop the flow, it is desirable to prevent 753 the application sending unnecessary traffic that might disrupt other 754 uses of the network. The role of the media usability circuit breaker 755 is to protect the network in such cases. 757 4.5. Ceasing Transmission 759 What it means to cease transmission depends on the application. The 760 intention is that the application will stop sending RTP data packets 761 to a particular destination 3-tuple (transport protocol, destination 762 port, IP address), until the user makes an explicit attempt to 763 restart the call. It is important that a human user is involved in 764 the decision to try to restart the call, since that user will 765 eventually give up if the calls repeatedly trigger the circuit 766 breaker. This will help avoid problems with automatic redial systems 767 from congesting the network. Accordingly, RTP flows halted by the 768 circuit breaker SHOULD NOT be restarted automatically unless the 769 sender has received information that the congestion has dissipated, 770 or can reasonably be expected to have dissipated. What could trigger 771 this expectation is necessarily application dependent, but could be, 772 for example, an indication that a competing flow has finished and 773 freed up some capacity, or -- for an application that is running on a 774 mobile device -- that the device moved to a new location so the flow 775 would traverse a different path if restarted. 777 It is recognised that the RTP implementation in some systems might 778 not be able to determine if a call set-up request was initiated by a 779 human user, or automatically by some scripted higher-level component 780 of the system. These implementations MUST rate limit attempts to 781 restart a call to the same destination 3-tuple as used by a call that 782 triggered the circuit breaker, so that the reaction to a triggered 783 circuit breaker lasts for at least the triggering interval 784 [I-D.ietf-tsvwg-circuit-breaker]. The chosen rate limit ought to not 785 exceed the rate at which an annoyed human caller might redial a 786 misbehaving phone. 788 The RTP circuit breaker will only trigger, and cease transmission, 789 for media flows subject to long-term persistent congestion. Such 790 flows are likely to have poor quality and usability for some time 791 before the circuit breaker triggers. Implementations can monitor 792 RTCP Reception Report blocks being returned for their media flows, 793 and might find it beneficial to use this information to provide a 794 user interface cue that problems are occurring, in advance of the 795 circuit breaker triggering. 797 5. RTP Circuit Breakers and the RTP/AVPF and RTP/SAVPF Profiles 799 Use of the Extended RTP Profile for RTCP-based Feedback (RTP/AVPF) 800 [RFC4585] allows receivers to send early RTCP reports in some cases, 801 to inform the sender about particular events in the media stream. 802 There are several use cases for such early RTCP reports, including 803 providing rapid feedback to a sender about the onset of congestion. 804 The RTP/SAVPF Profile [RFC5124] is a secure variant of the RTP/AVPF 805 profile, that is treated the same in the context of the RTP circuit 806 breaker. These feedback profiles are often used with non-compound 807 RTCP reports [RFC5506] to reduce the reporting overhead. 809 Receiving rapid feedback about congestion events potentially allows 810 congestion control algorithms to be more responsive, and to better 811 adapt the media transmission to the limitations of the network. It 812 is expected that many RTP congestion control algorithms will adopt 813 the RTP/AVPF profile or the RTP/SAVPF profile for this reason, 814 defining new transport layer feedback reports that suit their 815 requirements. Since these reports are not yet defined, and likely 816 very specific to the details of the congestion control algorithm 817 chosen, they cannot be used as part of the generic RTP circuit 818 breaker. 820 Reduced-size RTCP reports sent under the RTP/AVPF early feedback 821 rules that do not contain an RTCP SR or RR packet MUST be ignored by 822 the congestion circuit breaker (they do not contain the information 823 needed by the congestion circuit breaker algorithm), but MUST be 824 counted as received packets for the RTCP timeout circuit breaker. 825 Reduced-size RTCP reports sent under the RTP/AVPF early feedback 826 rules that contain RTCP SR or RR packets MUST be processed by the 827 congestion circuit breaker as if they were sent as regular RTCP 828 reports, and counted towards the circuit breaker conditions specified 829 in Section 4 of this memo. This will potentially make the RTP 830 circuit breaker trigger earlier than it would if the RTP/AVPF profile 831 was not used. 833 When using ECN with RTP (see Section 7), early RTCP feedback packets 834 can contain ECN feedback reports. The count of ECN-CE marked packets 835 contained in those ECN feedback reports is counted towards the number 836 of lost packets reported if the ECN Feedback Report is sent in a 837 compound RTCP packet along with an RTCP SR/RR report packet. Reports 838 of ECN-CE packets sent as reduced-size RTCP ECN feedback packets 839 without an RTCP SR/RR packet MUST be ignored. 841 These rules are intended to allow the use of low-overhead RTP/AVPF 842 feedback for generic NACK messages without triggering the RTP circuit 843 breaker. This is expected to make such feedback suitable for RTP 844 congestion control algorithms that need to quickly report loss events 845 in between regular RTCP reports. The reaction to reduced-size RTCP 846 SR/RR packets is to allow such algorithms to send feedback that can 847 trigger the circuit breaker, when desired. 849 The RTP/AVPF and RTP/SAVPF profiles include the T_rr_interval 850 parameter that can be used to adjust the regular RTCP reporting 851 interval. The use of the T_rr_interval parameter changes the 852 behaviour of the RTP circuit breaker, as described in Section 4. 854 6. Impact of RTCP Extended Reports (XR) 856 RTCP Extended Report (XR) blocks provide additional reception quality 857 metrics, but do not change the RTCP timing rules. Some of the RTCP 858 XR blocks provide information that might be useful for congestion 859 control purposes, others provide non-congestion-related metrics. 860 With the exception of RTCP XR ECN Summary Reports (see Section 7), 861 the presence of RTCP XR blocks in a compound RTCP packet does not 862 affect the RTP circuit breaker algorithm. For consistency and ease 863 of implementation, only the reception report blocks contained in RTCP 864 SR packets, RTCP RR packets, or RTCP XR ECN Summary Report packets, 865 are used by the RTP circuit breaker algorithm. 867 7. Impact of Explicit Congestion Notification (ECN) 869 The use of ECN for RTP flows does not affect the media timeout RTP 870 circuit breaker (Section 4.2) or the RTCP timeout circuit breaker 871 (Section 4.1), since these are both connectivity checks that simply 872 determinate if any packets are being received. 874 ECN-CE marked packets SHOULD be treated as if they were lost for the 875 purposes of congestion control, when determining the optimal media 876 sending rate for an RTP flow. If an RTP sender has negotiated ECN 877 support for an RTP session, and has successfully initiated ECN use on 878 the path to the receiver [RFC6679], then ECN-CE marked packets SHOULD 879 be treated as if they were lost when calculating if the congestion- 880 based RTP circuit breaker (Section 4.3) has been met, unless the RTP 881 implementation can determine that the ECN-CE marking on this path is 882 not reliable. The count of ECN-CE marked RTP packets is returned in 883 RTCP XR ECN summary report packets if support for ECN has been 884 initiated for an RTP session. 886 8. Impact of Bundled Media and Layered Coding 888 The RTP circuit breaker operates on a per-RTP session basis. An RTP 889 sender that participates in several RTP sessions MUST treat each RTP 890 session independently with regards to the RTP circuit breaker. 892 An RTP sender can generate several media streams within a single RTP 893 session, with each stream using a different SSRC. This can happen if 894 bundled media are in use, when using simulcast, or when using layered 895 media coding. By default, each SSRC will be treated independently by 896 the RTP circuit breaker. However, the sender MAY choose to treat the 897 flows (or a subset thereof) as a group, such that a circuit breaker 898 trigger for one flow applies to the group of flows as a whole, and 899 either causes the entire group to cease transmission, or the sending 900 rate of the group to reduce by a factor of ten, depending on the RTP 901 circuit breaker triggered. Grouping flows in this way is expected to 902 be especially useful for layered flows sent using multiple SSRCs, as 903 it allows the layered flow to react as a whole, ceasing transmission 904 on the enhancement layers first to reduce sending rate if necessary, 905 rather than treating each layer independently. Care needs to be 906 taken if the different media streams sent on a single transport layer 907 flow use different DSCP values [RFC7657], 908 [I-D.ietf-tsvwg-rtcweb-qos], since congestion could be experienced 909 differently depending on the DSCP marking. Accordingly, RTP media 910 streams with different DSCP values SHOULD NOT be considered as a 911 group when evaluating the RTP Circuit Breaker conditions. 913 9. Security Considerations 915 The security considerations of [RFC3550] apply. 917 If the RTP/AVPF profile is used to provide rapid RTCP feedback, the 918 security considerations of [RFC4585] apply. If ECN feedback for RTP 919 over UDP/IP is used, the security considerations of [RFC6679] apply. 921 If non-authenticated RTCP reports are used, an on-path attacker can 922 trivially generate fake RTCP packets that indicate high packet loss 923 rates, causing the circuit breaker to trigger and disrupt an RTP 924 session. This is somewhat more difficult for an off-path attacker, 925 due to the need to guess the randomly chosen RTP SSRC value and the 926 RTP sequence number. This attack can be avoided if RTCP packets are 927 authenticated; authentication options are discussed in [RFC7201]. 929 Timely operation of the RTP circuit breaker depends on the choice of 930 RTCP reporting interval. If the receiver has a reporting interval 931 that is overly long, then the responsiveness of the circuit breaker 932 decreases. In the limit, the RTP circuit breaker can be disabled for 933 all practical purposes by configuring an RTCP reporting interval that 934 is many minutes duration. This issue is not specific to the circuit 935 breaker: long RTCP reporting intervals also prevent reception quality 936 reports, feedback messages, codec control messages, etc., from being 937 used. Implementations are expected to impose an upper limit on the 938 RTCP reporting interval they are willing to negotiate (based on the 939 session bandwidth and RTCP bandwidth fraction) when using the RTP 940 circuit breaker, as discussed in Section 4.3. 942 10. IANA Considerations 944 There are no actions for IANA. 946 11. Acknowledgements 948 The authors would like to thank Bernard Aboba, Harald Alvestrand, 949 Spencer Dawkins, Gorry Fairhurst, Nazila Fough, Kevin Gross, Cullen 950 Jennings, Randell Jesup, Jonathan Lennox, Matt Mathis, Stephen 951 McQuistin, Simon Perreault, Eric Rescorla, Abheek Saha, Meral 952 Shirazipour, Fabio Verdicchio, and Magnus Westerlund for their 953 valuable feedback. 955 12. References 956 12.1. Normative References 958 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 959 Requirement Levels", BCP 14, RFC 2119, 960 DOI 10.17487/RFC2119, March 1997, 961 . 963 [RFC3550] Schulzrinne, H., Casner, S., Frederick, R., and V. 964 Jacobson, "RTP: A Transport Protocol for Real-Time 965 Applications", STD 64, RFC 3550, DOI 10.17487/RFC3550, 966 July 2003, . 968 [RFC3551] Schulzrinne, H. and S. Casner, "RTP Profile for Audio and 969 Video Conferences with Minimal Control", STD 65, RFC 3551, 970 DOI 10.17487/RFC3551, July 2003, 971 . 973 [RFC3611] Friedman, T., Ed., Caceres, R., Ed., and A. Clark, Ed., 974 "RTP Control Protocol Extended Reports (RTCP XR)", 975 RFC 3611, DOI 10.17487/RFC3611, November 2003, 976 . 978 [RFC4585] Ott, J., Wenger, S., Sato, N., Burmeister, C., and J. Rey, 979 "Extended RTP Profile for Real-time Transport Control 980 Protocol (RTCP)-Based Feedback (RTP/AVPF)", RFC 4585, 981 DOI 10.17487/RFC4585, July 2006, 982 . 984 [RFC5348] Floyd, S., Handley, M., Padhye, J., and J. Widmer, "TCP 985 Friendly Rate Control (TFRC): Protocol Specification", 986 RFC 5348, DOI 10.17487/RFC5348, September 2008, 987 . 989 [RFC6679] Westerlund, M., Johansson, I., Perkins, C., O'Hanlon, P., 990 and K. Carlberg, "Explicit Congestion Notification (ECN) 991 for RTP over UDP", RFC 6679, DOI 10.17487/RFC6679, August 992 2012, . 994 12.2. Informative References 996 [Floyd] Floyd, S., Handley, M., Padhye, J., and J. Widmer, 997 "Equation-Based Congestion Control for Unicast 998 Applications", Proceedings of the ACM SIGCOMM 999 conference, 2000, DOI 10.1145/347059.347397, August 2000. 1001 [I-D.ietf-avtcore-rtp-multi-stream] 1002 Lennox, J., Westerlund, M., Wu, Q., and C. Perkins, 1003 "Sending Multiple RTP Streams in a Single RTP Session", 1004 draft-ietf-avtcore-rtp-multi-stream-11 (work in progress), 1005 December 2015. 1007 [I-D.ietf-tsvwg-circuit-breaker] 1008 Fairhurst, G., "Network Transport Circuit Breakers", 1009 draft-ietf-tsvwg-circuit-breaker-15 (work in progress), 1010 April 2016. 1012 [I-D.ietf-tsvwg-rtcweb-qos] 1013 Jones, P., Dhesikan, S., Jennings, C., and D. Druta, "DSCP 1014 and other packet markings for WebRTC QoS", draft-ietf- 1015 tsvwg-rtcweb-qos-15 (work in progress), March 2016. 1017 [Mathis] Mathis, M., Semke, J., Mahdavi, J., and T. Ott, "The 1018 macroscopic behavior of the TCP congestion avoidance 1019 algorithm", ACM SIGCOMM Computer Communication 1020 Review 27(3), DOI 10.1145/263932.264023, July 1997. 1022 [Padhye] Padhye, J., Firoiu, V., Towsley, D., and J. Kurose, 1023 "Modeling TCP Throughput: A Simple Model and its Empirical 1024 Validation", Proceedings of the ACM SIGCOMM 1025 conference, 1998, DOI 10.1145/285237.285291, August 1998. 1027 [RFC2862] Civanlar, M. and G. Cash, "RTP Payload Format for Real- 1028 Time Pointers", RFC 2862, DOI 10.17487/RFC2862, June 2000, 1029 . 1031 [RFC3168] Ramakrishnan, K., Floyd, S., and D. Black, "The Addition 1032 of Explicit Congestion Notification (ECN) to IP", 1033 RFC 3168, DOI 10.17487/RFC3168, September 2001, 1034 . 1036 [RFC4103] Hellstrom, G. and P. Jones, "RTP Payload for Text 1037 Conversation", RFC 4103, DOI 10.17487/RFC4103, June 2005, 1038 . 1040 [RFC5104] Wenger, S., Chandra, U., Westerlund, M., and B. Burman, 1041 "Codec Control Messages in the RTP Audio-Visual Profile 1042 with Feedback (AVPF)", RFC 5104, DOI 10.17487/RFC5104, 1043 February 2008, . 1045 [RFC5124] Ott, J. and E. Carrara, "Extended Secure RTP Profile for 1046 Real-time Transport Control Protocol (RTCP)-Based Feedback 1047 (RTP/SAVPF)", RFC 5124, DOI 10.17487/RFC5124, February 1048 2008, . 1050 [RFC5405] Eggert, L. and G. Fairhurst, "Unicast UDP Usage Guidelines 1051 for Application Designers", BCP 145, RFC 5405, 1052 DOI 10.17487/RFC5405, November 2008, 1053 . 1055 [RFC5450] Singer, D. and H. Desineni, "Transmission Time Offsets in 1056 RTP Streams", RFC 5450, DOI 10.17487/RFC5450, March 2009, 1057 . 1059 [RFC5506] Johansson, I. and M. Westerlund, "Support for Reduced-Size 1060 Real-Time Transport Control Protocol (RTCP): Opportunities 1061 and Consequences", RFC 5506, DOI 10.17487/RFC5506, April 1062 2009, . 1064 [RFC5681] Allman, M., Paxson, V., and E. Blanton, "TCP Congestion 1065 Control", RFC 5681, DOI 10.17487/RFC5681, September 2009, 1066 . 1068 [RFC6798] Clark, A. and Q. Wu, "RTP Control Protocol (RTCP) Extended 1069 Report (XR) Block for Packet Delay Variation Metric 1070 Reporting", RFC 6798, DOI 10.17487/RFC6798, November 2012, 1071 . 1073 [RFC6843] Clark, A., Gross, K., and Q. Wu, "RTP Control Protocol 1074 (RTCP) Extended Report (XR) Block for Delay Metric 1075 Reporting", RFC 6843, DOI 10.17487/RFC6843, January 2013, 1076 . 1078 [RFC6958] Clark, A., Zhang, S., Zhao, J., and Q. Wu, Ed., "RTP 1079 Control Protocol (RTCP) Extended Report (XR) Block for 1080 Burst/Gap Loss Metric Reporting", RFC 6958, 1081 DOI 10.17487/RFC6958, May 2013, 1082 . 1084 [RFC7002] Clark, A., Zorn, G., and Q. Wu, "RTP Control Protocol 1085 (RTCP) Extended Report (XR) Block for Discard Count Metric 1086 Reporting", RFC 7002, DOI 10.17487/RFC7002, September 1087 2013, . 1089 [RFC7003] Clark, A., Huang, R., and Q. Wu, Ed., "RTP Control 1090 Protocol (RTCP) Extended Report (XR) Block for Burst/Gap 1091 Discard Metric Reporting", RFC 7003, DOI 10.17487/RFC7003, 1092 September 2013, . 1094 [RFC7097] Ott, J., Singh, V., Ed., and I. Curcio, "RTP Control 1095 Protocol (RTCP) Extended Report (XR) for RLE of Discarded 1096 Packets", RFC 7097, DOI 10.17487/RFC7097, January 2014, 1097 . 1099 [RFC7201] Westerlund, M. and C. Perkins, "Options for Securing RTP 1100 Sessions", RFC 7201, DOI 10.17487/RFC7201, April 2014, 1101 . 1103 [RFC7657] Black, D., Ed. and P. Jones, "Differentiated Services 1104 (Diffserv) and Real-Time Communication", RFC 7657, 1105 DOI 10.17487/RFC7657, November 2015, 1106 . 1108 [RFC7713] Mathis, M. and B. Briscoe, "Congestion Exposure (ConEx) 1109 Concepts, Abstract Mechanism, and Requirements", RFC 7713, 1110 DOI 10.17487/RFC7713, December 2015, 1111 . 1113 [Sarker] Sarker, Z., Singh, V., and C. Perkins, "An Evaluation of 1114 RTP Circuit Breaker Performance on LTE Networks", 1115 Proceedings of the IEEE Infocom workshop on Communication 1116 and Networking Techniques for Contemporary Video, 2014, 1117 April 2014. 1119 [Singh] Singh, V., McQuistin, S., Ellis, M., and C. Perkins, 1120 "Circuit Breakers for Multimedia Congestion Control", 1121 Proceedings of the International Packet Video 1122 Workshop, 2013, DOI 10.1109/PV.2013.6691439, December 1123 2013. 1125 Authors' Addresses 1127 Colin Perkins 1128 University of Glasgow 1129 School of Computing Science 1130 Glasgow G12 8QQ 1131 United Kingdom 1133 Email: csp@csperkins.org 1135 Varun Singh 1136 Nemu Dialogue Systems Oy 1137 Runeberginkatu 4c A 4 1138 Helsinki 00100 1139 Finland 1141 Email: varun.singh@iki.fi 1142 URI: http://www.callstats.io/