idnits 2.17.1 draft-ietf-avtcore-rtp-circuit-breakers-16.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 (June 11, 2016) is 2868 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(See RFCs 3967 and 4897 for information about using normative references to lower-maturity documents in RFCs) == Outdated reference: A later version (-13) exists of draft-ietf-rmcat-nada-02 == Outdated reference: A later version (-13) exists of draft-ietf-rmcat-scream-cc-04 == Outdated reference: A later version (-18) exists of draft-ietf-tsvwg-rtcweb-qos-17 == Outdated reference: A later version (-01) exists of draft-khademi-tcpm-alternativebackoff-ecn-00 -- Obsolete informational reference (is this intentional?): RFC 5405 (Obsoleted by RFC 8085) Summary: 0 errors (**), 0 flaws (~~), 5 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: December 13, 2016 June 11, 2016 8 Multimedia Congestion Control: Circuit Breakers for Unicast RTP Sessions 9 draft-ietf-avtcore-rtp-circuit-breakers-16 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. To avoid triggering the RTP 32 circuit breaker, any standards-track congestion control algorithms 33 defined for RTP will need to operate within the envelope set by these 34 RTP circuit breaker algorithms. 36 Status of This Memo 38 This Internet-Draft is submitted in full conformance with the 39 provisions of BCP 78 and BCP 79. 41 Internet-Drafts are working documents of the Internet Engineering 42 Task Force (IETF). Note that other groups may also distribute 43 working documents as Internet-Drafts. The list of current Internet- 44 Drafts is at http://datatracker.ietf.org/drafts/current/. 46 Internet-Drafts are draft documents valid for a maximum of six months 47 and may be updated, replaced, or obsoleted by other documents at any 48 time. It is inappropriate to use Internet-Drafts as reference 49 material or to cite them other than as "work in progress." 51 This Internet-Draft will expire on December 13, 2016. 53 Copyright Notice 55 Copyright (c) 2016 IETF Trust and the persons identified as the 56 document authors. All rights reserved. 58 This document is subject to BCP 78 and the IETF Trust's Legal 59 Provisions Relating to IETF Documents 60 (http://trustee.ietf.org/license-info) in effect on the date of 61 publication of this document. Please review these documents 62 carefully, as they describe your rights and restrictions with respect 63 to this document. Code Components extracted from this document must 64 include Simplified BSD License text as described in Section 4.e of 65 the Trust Legal Provisions and are provided without warranty as 66 described in the Simplified BSD License. 68 Table of Contents 70 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 2 71 2. Background . . . . . . . . . . . . . . . . . . . . . . . . . 3 72 3. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . 6 73 4. RTP Circuit Breakers for Systems Using the RTP/AVP Profile . 7 74 4.1. RTP/AVP Circuit Breaker #1: RTCP Timeout . . . . . . . . 10 75 4.2. RTP/AVP Circuit Breaker #2: Media Timeout . . . . . . . . 11 76 4.3. RTP/AVP Circuit Breaker #3: Congestion . . . . . . . . . 12 77 4.4. RTP/AVP Circuit Breaker #4: Media Usability . . . . . . . 16 78 4.5. Ceasing Transmission . . . . . . . . . . . . . . . . . . 17 79 5. RTP Circuit Breakers and the RTP/AVPF and RTP/SAVPF Profiles 17 80 6. Impact of RTCP Extended Reports (XR) . . . . . . . . . . . . 19 81 7. Impact of Explicit Congestion Notification (ECN) . . . . . . 19 82 8. Impact of Bundled Media and Layered Coding . . . . . . . . . 20 83 9. Security Considerations . . . . . . . . . . . . . . . . . . . 21 84 10. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 21 85 11. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 21 86 12. References . . . . . . . . . . . . . . . . . . . . . . . . . 22 87 12.1. Normative References . . . . . . . . . . . . . . . . . . 22 88 12.2. Informative References . . . . . . . . . . . . . . . . . 22 89 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 26 91 1. Introduction 93 The Real-time Transport Protocol (RTP) [RFC3550] is widely used in 94 voice-over-IP, video teleconferencing, and telepresence systems. 95 Many of these systems run over best-effort UDP/IP networks, and can 96 suffer from packet loss and increased latency if network congestion 97 occurs. Designing effective RTP congestion control algorithms, to 98 adapt the transmission of RTP-based media to match the available 99 network capacity, while also maintaining the user experience, is a 100 difficult but important problem. Many such congestion control and 101 media adaptation algorithms have been proposed, but to date there is 102 no consensus on the correct approach, or even that a single standard 103 algorithm is desirable. 105 This memo does not attempt to propose a new RTP congestion control 106 algorithm. Instead, we propose a small set of RTP circuit breakers: 107 mechanisms that terminate RTP flows in conditions under which there 108 is general agreement that serious network congestion is occurring. 109 The RTP circuit breakers proposed in this memo are a specific 110 instance of the general class of network transport circuit breakers 111 [I-D.ietf-tsvwg-circuit-breaker], designed to act as a protection 112 mechanism of last resort to avoid persistent excessive congestion. 113 To avoid triggering the RTP circuit breaker, any standards-track 114 congestion control algorithms defined for RTP will need to operate 115 within the envelope set by the RTP circuit breaker algorithms defined 116 by this memo. 118 2. Background 120 We consider congestion control for unicast RTP traffic flows. This 121 is the problem of adapting the transmission of an audio/visual data 122 flow, encapsulated within an RTP transport session, from one sender 123 to one receiver, so that it does not use more capacity than is 124 available along the network path. Such adaptation needs to be done 125 in a way that limits the disruption to the user experience caused by 126 both packet loss and excessive rate changes. Congestion control for 127 multicast flows is outside the scope of this memo. Multicast traffic 128 needs different solutions, since the available capacity estimator for 129 a group of receivers will differ from that for a single receiver, and 130 because multicast congestion control has to consider issues of 131 fairness across groups of receivers that do not apply to unicast 132 flows. 134 Congestion control for unicast RTP traffic can be implemented in one 135 of two places in the protocol stack. One approach is to run the RTP 136 traffic over a congestion controlled transport protocol, for example 137 over TCP, and to adapt the media encoding to match the dictates of 138 the transport-layer congestion control algorithm. This is safe for 139 the network, but can be suboptimal for the media quality unless the 140 transport protocol is designed to support real-time media flows. We 141 do not consider this class of applications further in this memo, as 142 their network safety is guaranteed by the underlying transport. 144 Alternatively, RTP flows can be run over a non-congestion controlled 145 transport protocol, for example UDP, performing rate adaptation at 146 the application layer based on RTP Control Protocol (RTCP) feedback. 147 With a well-designed, network-aware, application, this allows highly 148 effective media quality adaptation, but there is potential to cause 149 persistent congestion in the network if the application does not 150 adapt its sending rate in a timely and effective manner. We consider 151 this class of applications in this memo. 153 Congestion control relies on monitoring the delivery of a media flow, 154 and responding to adapt the transmission of that flow when there are 155 signs that the network path is congested. Network congestion can be 156 detected in one of three ways: 1) a receiver can infer the onset of 157 congestion by observing an increase in one-way delay caused by queue 158 build-up within the network; 2) if Explicit Congestion Notification 159 (ECN) [RFC3168] is supported, the network can signal the presence of 160 congestion by marking packets using ECN Congestion Experienced (CE) 161 marks (this could potentially be augmented by mechanisms such as 162 ConEX [RFC7713], or other future protocol extensions for network 163 signalling of congestion); or 3) in the extreme case, congestion will 164 cause packet loss that can be detected by observing a gap in the 165 received RTP sequence numbers. 167 Once the onset of congestion is observed, the receiver has to send 168 feedback to the sender to indicate that the transmission rate needs 169 to be reduced. How the sender reduces the transmission rate is 170 highly dependent on the media codec being used, and is outside the 171 scope of this memo. 173 There are several ways in which a receiver can send feedback to a 174 media sender within the RTP framework: 176 o The base RTP specification [RFC3550] defines RTCP Reception Report 177 (RR) packets to convey reception quality feedback information, and 178 Sender Report (SR) packets to convey information about the media 179 transmission. RTCP SR packets contain data that can be used to 180 reconstruct media timing at a receiver, along with a count of the 181 total number of octets and packets sent. RTCP RR packets report 182 on the fraction of packets lost in the last reporting interval, 183 the cumulative number of packets lost, the highest sequence number 184 received, and the inter-arrival jitter. The RTCP RR packets also 185 contain timing information that allows the sender to estimate the 186 network round trip time (RTT) to the receivers. RTCP reports are 187 sent periodically, with the reporting interval being determined by 188 the number of SSRCs used in the session and a configured session 189 bandwidth estimate (the number of synchronisation sources (SSRCs) 190 used is usually two in a unicast session, one for each 191 participant, but can be greater if the participants send multiple 192 media streams). The interval between reports sent from each 193 receiver tends to be on the order of a few seconds on average, 194 although it varies with the session bandwidth, and sub-second 195 reporting intervals are possible in high bandwidth sessions, and 196 it is randomised to avoid synchronisation of reports from multiple 197 receivers. RTCP RR packets allow a receiver to report ongoing 198 network congestion to the sender. However, if a receiver detects 199 the onset of congestion part way through a reporting interval, the 200 base RTP specification contains no provision for sending the RTCP 201 RR packet early, and the receiver has to wait until the next 202 scheduled reporting interval. 204 o The RTCP Extended Reports (XR) [RFC3611] allow reporting of more 205 complex and sophisticated reception quality metrics, but do not 206 change the RTCP timing rules. RTCP extended reports of potential 207 interest for congestion control purposes are the extended packet 208 loss, discard, and burst metrics [RFC3611], [RFC7002], [RFC7097], 209 [RFC7003], [RFC6958]; and the extended delay metrics [RFC6843], 210 [RFC6798]. Other RTCP Extended Reports that could be helpful for 211 congestion control purposes might be developed in future. 213 o Rapid feedback about the occurrence of congestion events can be 214 achieved using the Extended RTP Profile for RTCP-Based Feedback 215 (RTP/AVPF) [RFC4585] (or its secure variant, RTP/SAVPF [RFC5124]) 216 in place of the RTP/AVP profile [RFC3551]. This modifies the RTCP 217 timing rules to allow RTCP reports to be sent early, in some cases 218 immediately, provided the RTCP transmission rate keeps within its 219 bandwidth allocation. It also defines transport-layer feedback 220 messages, including negative acknowledgements (NACKs), that can be 221 used to report on specific congestion events. RTP Codec Control 222 Messages [RFC5104] extend the RTP/AVPF profile with additional 223 feedback messages that can be used to influence that way in which 224 rate adaptation occurs, but do not further change the dynamics of 225 how rapidly feedback can be sent. Use of the RTP/AVPF profile is 226 dependent on signalling. 228 o Finally, Explicit Congestion Notification (ECN) for RTP over UDP 229 [RFC6679] can be used to provide feedback on the number of packets 230 that received an ECN Congestion Experienced (CE) mark. This RTCP 231 extension builds on the RTP/AVPF profile to allow rapid congestion 232 feedback when ECN is supported. 234 In addition to these mechanisms for providing feedback, the sender 235 can include an RTP header extension in each packet to record packet 236 transmission times [RFC5450]. Accurate transmission timestamps can 237 be helpful for estimating queuing delays, to get an early indication 238 of the onset of congestion. 240 Taken together, these various mechanisms allow receivers to provide 241 feedback on the senders when congestion events occur, with varying 242 degrees of timeliness and accuracy. The key distinction is between 243 systems that use only the basic RTCP mechanisms, without RTP/AVPF 244 rapid feedback, and those that use the RTP/AVPF extensions to respond 245 to congestion more rapidly. 247 3. Terminology 249 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 250 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 251 document are to be interpreted as described in RFC 2119 [RFC2119]. 252 This interpretation of these key words applies only when written in 253 ALL CAPS. Mixed- or lower-case uses of these key words are not to be 254 interpreted as carrying special significance in this memo. 256 The definition of the RTP circuit breaker is specified in terms of 257 the following variables: 259 o Td is the deterministic RTCP reporting interval, as defined in 260 Section 6.3.1 of [RFC3550]. 262 o Tdr is the sender's estimate of the deterministic RTCP reporting 263 interval, Td, calculated by a receiver of the data it is sending. 264 Tdr is not known at the sender, but can be estimated by executing 265 the algorithm in Section 6.2 of [RFC3550] using the average RTCP 266 packet size seen at the sender, the number of members reported in 267 the receiver's SR/RR report blocks, and whether the receiver is 268 sending SR or RR packets. Tdr is recalculated when each new RTCP 269 SR/RR report is received, but the media timeout circuit breaker 270 (see Section 4.2) is only reconsidered when Tdr increases. 272 o Tr is the network round-trip time, calculated by the sender using 273 the algorithm in Section 6.4.1 of [RFC3550] and smoothed using an 274 exponentially weighted moving average as Tr = (0.8 * Tr) + (0.2 * 275 Tr_new) where Tr_new is the latest RTT estimate obtained from an 276 RTCP report. The weight is chosen so old estimates decay over k 277 intervals. 279 o k is the non-reporting threshold (see Section 4.2). 281 o Tf is the media framing interval at the sender. For applications 282 sending at a constant frame rate, Tf is the inter-frame interval. 283 For applications that switch between a small set of possible frame 284 rates, for example when sending speech with comfort noise, where 285 comfort noise frames are sent less often than speech frames, Tf is 286 set to the longest of the inter-frame intervals of the different 287 frame rates. For applications that send periodic frames but 288 dynamically vary their frame rate, Tf is set to the largest inter- 289 frame interval used in the last 10 seconds. For applications that 290 send less than one frame every 10 seconds, or that have no concept 291 of periodic frames (e.g., text conversation [RFC4103], or pointer 292 events [RFC2862]), Tf is set to the time interval since the 293 previous frame when each frame is sent. 295 o G is the frame group size. That is, the number of frames that are 296 coded together based on a particular sending rate setting. If the 297 codec used by the sender can change its rate on each frame, G = 1; 298 otherwise G is set to the number of frames before the codec can 299 adjust to the new rate. For codecs that have the concept of a 300 group-of-pictures (GoP), G is likely the GoP length. 302 o T_rr_interval is the minimal interval between RTCP reports, as 303 defined in Section 3.4 of [RFC4585]; it is only meaningful for 304 implementations of RTP/AVPF profile [RFC4585] or the RTP/SAVPF 305 profile [RFC5124]. 307 o X is the estimated throughput a TCP connection would achieve over 308 a path, in bytes per second. 310 o s is the size of RTP packets being sent, in bytes. If the RTP 311 packets being sent vary in size, then the average size over the 312 packet comprising the last 4 * G frames MUST be used (this is 313 intended to be comparable to the four loss intervals used in 314 [RFC5348]). 316 o p is the loss event rate, between 0.0 and 1.0, that would be seen 317 by a TCP connection over a particular path. When used in the RTP 318 congestion circuit breaker, this is approximated as described in 319 Section 4.3. 321 o t_RTO is the retransmission timeout value that would be used by a 322 TCP connection over a particular path, in seconds. This MUST be 323 approximated using t_RTO = 4 * Tr when used as part of the RTP 324 congestion circuit breaker. 326 o b is the number of packets that are acknowledged by a single TCP 327 acknowledgement. Following [RFC5348], it is RECOMMENDED that the 328 value b = 1 is used as part of the RTP congestion circuit breaker. 330 4. RTP Circuit Breakers for Systems Using the RTP/AVP Profile 332 The feedback mechanisms defined in [RFC3550] and available under the 333 RTP/AVP profile [RFC3551] are the minimum that can be assumed for a 334 baseline circuit breaker mechanism that is suitable for all unicast 335 applications of RTP. Accordingly, for an RTP circuit breaker to be 336 useful, it needs to be able to detect that an RTP flow is causing 337 excessive congestion using only basic RTCP features, without needing 338 RTCP XR feedback or the RTP/AVPF profile for rapid RTCP reports. 340 RTCP is a fundamental part of the RTP protocol, and the mechanisms 341 described here rely on the implementation of RTCP. Implementations 342 that claim to support RTP, but that do not implement RTCP, will be 343 unable to use the circuit breaker mechanisms described in this memo. 344 Such implementations SHOULD NOT be used on networks that might be 345 subject to congestion unless equivalent mechanisms are defined using 346 some non-RTCP feedback channel to report congestion and signal 347 circuit breaker conditions. 349 The RTCP timeout circuit breaker (Section 4.1) will trigger if an 350 implementation of this memo attempts to interwork with an endpoint 351 that does not support RTCP. Implementations that sometimes need to 352 interwork with endpoints that do not support RTCP need to disable the 353 RTP circuit breakers if they don't receive some confirmation via 354 signalling that the remote endpoint implements RTCP (the presence of 355 an SDP "a=rtcp:" attribute in an answer might be such an indication). 356 The RTP Circuit Breaker SHOULD NOT be disabled on networks that might 357 be subject to congestion, unless equivalent mechanisms are defined 358 using some non-RTCP feedback channel to report congestion and signal 359 circuit breaker conditions [I-D.ietf-tsvwg-circuit-breaker]. 361 Three potential congestion signals are available from the basic RTCP 362 SR/RR packets and are reported for each SSRC in the RTP session: 364 1. The sender can estimate the network round-trip time once per RTCP 365 reporting interval, based on the contents and timing of RTCP SR 366 and RR packets. 368 2. Receivers report a jitter estimate (the statistical variance of 369 the RTP data packet inter-arrival time) calculated over the RTCP 370 reporting interval. Due to the nature of the jitter calculation 371 ([RFC3550], section 6.4.4), the jitter is only meaningful for RTP 372 flows that send a single data packet for each RTP timestamp value 373 (i.e., audio flows, or video flows where each packet comprises 374 one video frame). 376 3. Receivers report the fraction of RTP data packets lost during the 377 RTCP reporting interval, and the cumulative number of RTP packets 378 lost over the entire RTP session. 380 These congestion signals limit the possible circuit breakers, since 381 they give only limited visibility into the behaviour of the network. 383 RTT estimates are widely used in congestion control algorithms, as a 384 proxy for queuing delay measures in delay-based congestion control or 385 to determine connection timeouts. RTT estimates derived from RTCP SR 386 and RR packets sent according to the RTP/AVP timing rules are too 387 infrequent to be useful for congestion control, and don't give enough 388 information to distinguish a delay change due to routing updates from 389 queuing delay caused by congestion. Accordingly, we cannot use the 390 RTT estimate alone as an RTP circuit breaker. 392 Increased jitter can be a signal of transient network congestion, but 393 in the highly aggregated form reported in RTCP RR packets, it offers 394 insufficient information to estimate the extent or persistence of 395 congestion. Jitter reports are a useful early warning of potential 396 network congestion, but provide an insufficiently strong signal to be 397 used as a circuit breaker. 399 The remaining congestion signals are the packet loss fraction and the 400 cumulative number of packets lost. If considered carefully, and over 401 an appropriate time frame to distinguish transient problems from long 402 term issues [I-D.ietf-tsvwg-circuit-breaker], these can be effective 403 indicators that persistent excessive congestion is occurring in 404 networks where packet loss is primarily due to queue overflows, 405 although loss caused by non-congestive packet corruption can distort 406 the result in some networks. TCP congestion control [RFC5681] 407 intentionally tries to fill the router queues, and uses the resulting 408 packet loss as congestion feedback. An RTP flow competing with TCP 409 traffic will therefore expect to see a non-zero packet loss fraction, 410 and some variation in queuing latency, in normal operation when 411 sharing a path with other flows, that needs to be accounted for when 412 determining the circuit breaker threshold 413 [I-D.ietf-tsvwg-circuit-breaker]. This behaviour of TCP is reflected 414 in the congestion circuit breaker below, and will affect the design 415 of any RTP congestion control protocol. 417 Two packet loss regimes can be observed: 1) RTCP RR packets show a 418 non-zero packet loss fraction, while the extended highest sequence 419 number received continues to increment; and 2) RR packets show a loss 420 fraction of zero, but the extended highest sequence number received 421 does not increment even though the sender has been transmitting RTP 422 data packets. The former corresponds to the TCP congestion avoidance 423 state, and indicates a congested path that is still delivering data; 424 the latter corresponds to a TCP timeout, and is most likely due to a 425 path failure. A third condition is that data is being sent but no 426 RTCP feedback is received at all, corresponding to a failure of the 427 reverse path. We derive circuit breaker conditions for these loss 428 regimes in the following. 430 4.1. RTP/AVP Circuit Breaker #1: RTCP Timeout 432 An RTCP timeout can occur when RTP data packets are being sent, but 433 there are no RTCP reports returned from the receiver. This is either 434 due to a failure of the receiver to send RTCP reports, or a failure 435 of the return path that is preventing those RTCP reporting from being 436 delivered. In either case, it is not safe to continue transmission, 437 since the sender has no way of knowing if it is causing congestion. 439 An RTP sender that has not received any RTCP SR or RTCP RR packets 440 reporting on the SSRC it is using, for a time period of at least 441 three times its deterministic RTCP reporting interval, Td, without 442 the randomization factor, and using the fixed minimum interval of 443 Tmin=5 seconds, SHOULD cease transmission (see Section 4.5). The 444 rationale for this choice of timeout is as described in Section 6.2 445 of [RFC3550] ("so that implementations which do not use the reduced 446 value for transmitting RTCP packets are not timed out by other 447 participants prematurely"), as updated by Section 6.1.4 of 448 [I-D.ietf-avtcore-rtp-multi-stream] to account for the use of the 449 RTP/AVPF profile [RFC4585] or the RTP/SAVPF profile [RFC5124]. 451 To reduce the risk of premature timeout, implementations SHOULD NOT 452 configure the RTCP bandwidth such that Td is larger than 5 seconds. 453 Similarly, implementations that use the RTP/AVPF profile [RFC4585] or 454 the RTP/SAVPF profile [RFC5124] SHOULD NOT configure T_rr_interval to 455 values larger than 4 seconds (the reduced limit for T_rr_interval 456 follows Section 6.1.3 of [I-D.ietf-avtcore-rtp-multi-stream]). 458 The choice of three RTCP reporting intervals as the timeout is made 459 following Section 6.3.5 of RFC 3550 [RFC3550]. This specifies that 460 participants in an RTP session will timeout and remove an RTP sender 461 from the list of active RTP senders if no RTP data packets have been 462 received from that RTP sender within the last two RTCP reporting 463 intervals. Using a timeout of three RTCP reporting intervals is 464 therefore large enough that the other participants will have timed 465 out the sender if a network problem stops the data packets it is 466 sending from reaching the receivers, even allowing for loss of some 467 RTCP packets. 469 If a sender is transmitting a large number of RTP media streams, such 470 that the corresponding RTCP SR or RR packets are too large to fit 471 into the network MTU, the receiver will generate RTCP SR or RR 472 packets in a round-robin manner. In this case, the sender SHOULD 473 treat receipt of an RTCP SR or RR packet corresponding to any SSRC it 474 sent on the same 5-tuple of source and destination IP address, port, 475 and protocol, as an indication that the receiver and return path are 476 working, preventing the RTCP timeout circuit breaker from triggering. 478 4.2. RTP/AVP Circuit Breaker #2: Media Timeout 480 If RTP data packets are being sent, but the RTCP SR or RR packets 481 reporting on that SSRC indicate a non-increasing extended highest 482 sequence number received, this is an indication that those RTP data 483 packets are not reaching the receiver. This could be a short-term 484 issue affecting only a few RTP packets, perhaps caused by a slow to 485 open firewall or a transient connectivity problem, but if the issue 486 persists, it is a sign of a more ongoing and significant problem (a 487 "media timeout"). 489 The time needed to declare a media timeout depends on the parameters 490 Tdr, Tr, Tf, and on the non-reporting threshold k. The value of k is 491 chosen so that when Tdr is large compared to Tr and Tf, receipt of at 492 least k RTCP reports with non-increasing extended highest sequence 493 number received gives reasonable assurance that the forward path has 494 failed, and that the RTP data packets have not been lost by chance. 495 The RECOMMENDED value for k is 5 reports. 497 When Tdr < Tf, then RTP data packets are being sent at a rate less 498 than one per RTCP reporting interval of the receiver, so the extended 499 highest sequence number received can be expected to be non-increasing 500 for some receiver RTCP reporting intervals. Similarly, when Tdr < 501 Tr, some receiver RTCP reporting intervals might pass before the RTP 502 data packets arrive at the receiver, also leading to reports where 503 the extended highest sequence number received is non-increasing. 504 Both issues require the media timeout interval to be scaled relative 505 to the threshold, k. 507 The media timeout RTP circuit breaker is therefore as follows. When 508 starting sending, calculate MEDIA_TIMEOUT using: 510 MEDIA_TIMEOUT = ceil(k * max(Tf, Tr, Tdr) / Tdr) 512 When a sender receives an RTCP packet that indicates reception of the 513 media it has been sending, then it cancels the media timeout circuit 514 breaker. If it is still sending, then it MUST calculate a new value 515 for MEDIA_TIMEOUT, and set a new media timeout circuit breaker. 517 If a sender receives an RTCP packet indicating that its media was not 518 received, it MUST calculate a new value for MEDIA_TIMEOUT. If the 519 new value is larger than the previous, it replaces MEDIA_TIMEOUT with 520 the new value, extending the media timeout circuit breaker; otherwise 521 it keeps the original value of MEDIA_TIMEOUT. This process is known 522 as reconsidering the media timeout circuit breaker. 524 If MEDIA_TIMEOUT consecutive RTCP packets are received indicating 525 that the media being sent was not received, and the media timeout 526 circuit breaker has not been cancelled, then the media timeout 527 circuit breaker triggers. When the media timeout circuit breaker 528 triggers, the sender SHOULD cease transmission (see Section 4.5). 530 When stopping sending an RTP stream, a sender MUST cancel the 531 corresponding media timeout circuit breaker. 533 4.3. RTP/AVP Circuit Breaker #3: Congestion 535 If RTP data packets are being sent, and the corresponding RTCP SR or 536 RR packets show non-zero packet loss fraction and increasing extended 537 highest sequence number received, then those RTP data packets are 538 arriving at the receiver, but some degree of congestion is occurring. 539 The RTP/AVP profile [RFC3551] states that: 541 If best-effort service is being used, RTP receivers SHOULD monitor 542 packet loss to ensure that the packet loss rate is within 543 acceptable parameters. Packet loss is considered acceptable if a 544 TCP flow across the same network path and experiencing the same 545 network conditions would achieve an average throughput, measured 546 on a reasonable time scale, that is not less than the throughput 547 the RTP flow is achieving. This condition can be satisfied by 548 implementing congestion control mechanisms to adapt the 549 transmission rate (or the number of layers subscribed for a 550 layered multicast session), or by arranging for a receiver to 551 leave the session if the loss rate is unacceptably high. 553 The comparison to TCP cannot be specified exactly, but is intended 554 as an "order-of-magnitude" comparison in time scale and 555 throughput. The time scale on which TCP throughput is measured is 556 the round-trip time of the connection. In essence, this 557 requirement states that it is not acceptable to deploy an 558 application (using RTP or any other transport protocol) on the 559 best-effort Internet which consumes bandwidth arbitrarily and does 560 not compete fairly with TCP within an order of magnitude. 562 The phase "order of magnitude" in the above means within a factor of 563 ten, approximately. In order to implement this, it is necessary to 564 estimate the throughput a bulk TCP connection would achieve over the 565 path. For a long-lived TCP Reno connection, it has been shown that 566 the TCP throughput, X, in bytes per second, can be estimated using 567 [Padhye]: 569 s 570 X = ------------------------------------------------------------- 571 Tr*sqrt(2*b*p/3)+(t_RTO * (3*sqrt(3*b*p/8) * p * (1+32*p*p))) 573 This is the same approach to estimated TCP throughput that is used in 574 [RFC5348]. Under conditions of low packet loss the second term on 575 the denominator is small, so this formula can be approximated with 576 reasonable accuracy as follows [Mathis]: 578 s 579 X = ---------------- 580 Tr*sqrt(2*b*p/3) 582 It is RECOMMENDED that this simplified throughput equation be used, 583 since the reduction in accuracy is small, and it is much simpler to 584 calculate than the full equation. Measurements have shown that the 585 simplified TCP throughput equation is effective as an RTP circuit 586 breaker for multimedia flows sent to hosts on residential networks 587 using ADSL and cable modem links [Singh]. The data shows that the 588 full TCP throughput equation tends to be more sensitive to packet 589 loss and triggers the RTP circuit breaker earlier than the simplified 590 equation. Implementations that desire this extra sensitivity MAY use 591 the full TCP throughput equation in the RTP circuit breaker. Initial 592 measurements in LTE networks have shown that the extra sensitivity is 593 helpful in that environment, with the full TCP throughput equation 594 giving a more balanced circuit breaker response than the simplified 595 TCP equation [Sarker]; other networks might see similar behaviour. 597 No matter what TCP throughput equation is chosen, two parameters need 598 to be estimated and reported to the sender in order to calculate the 599 throughput: the round trip time, Tr, and the loss event rate, p (the 600 packet size, s, is known to the sender). The round trip time can be 601 estimated from RTCP SR and RR packets. This is done too infrequently 602 for accurate statistics, but is the best that can be done with the 603 standard RTCP mechanisms. 605 Report blocks in RTCP SR or RR packets contain the packet loss 606 fraction, rather than the loss event rate, so p cannot be reported 607 (TCP typically treats the loss of multiple packets within a single 608 RTT as one loss event, but RTCP RR packets report the overall 609 fraction of packets lost, and does not report when the packet losses 610 occurred). Using the loss fraction in place of the loss event rate 611 can overestimate the loss. We believe that this overestimate will 612 not be significant, given that we are only interested in order of 613 magnitude comparison ([Floyd] section 3.2.1 shows that the difference 614 is small for steady-state conditions and random loss, but using the 615 loss fraction is more conservative in the case of bursty loss). 617 The congestion circuit breaker is therefore: when a sender that is 618 transmitting at least one RTP packet every max(Tdr, Tr) seconds 619 receives an RTCP SR or RR packet that contains a report block for an 620 SSRC it is using, the sender MUST record the value of the fraction 621 lost field from the report block, and the time since the last report 622 block was received, for that SSRC. If more than CB_INTERVAL (see 623 below) report blocks have been received for that SSRC, the sender 624 MUST calculate the average fraction lost over the last CB_INTERVAL 625 reporting intervals, and then estimate the TCP throughput that would 626 be achieved over the path using the chosen TCP throughput equation 627 and the measured values of the round-trip time, Tr, the loss event 628 rate, p (approximated by the average fraction lost, as is described 629 below), and the packet size, s. The estimate of the TCP throughput, 630 X, is then compared with the actual sending rate of the RTP stream. 631 If the actual sending rate of the RTP stream is more than 10 * X, 632 then the congestion circuit breaker is triggered. 634 The average fraction lost is calculated based on the sum, over the 635 last CB_INTERVAL reporting intervals, of the fraction lost in each 636 reporting interval multiplied by the duration of the corresponding 637 reporting interval, divided by the total duration of the last 638 CB_INTERVAL reporting intervals. The CB_INTERVAL parameter is set 639 to: 641 CB_INTERVAL = 642 ceil(3*min(max(10*G*Tf, 10*Tr, 3*Tdr), max(15, 3*Td))/(3*Tdr)) 644 The parameters that feed into CB_INTERVAL are chosen to give the 645 congestion control algorithm time to react to congestion. They give 646 at least three RTCP reports, ten round trip times, and ten groups of 647 frames to adjust the rate to reduce the congestion to a reasonable 648 level. It is expected that a responsive congestion control algorithm 649 will begin to respond with the next group of frames after it receives 650 indication of congestion, so CB_INTERVAL ought to be a much longer 651 interval than the congestion response. 653 If the RTP/AVPF profile [RFC4585] or the RTP/SAVPF [RFC5124] is used, 654 and the T_rr_interval parameter is used to reduce the frequency of 655 regular RTCP reports, then the value Tdr in the above expression for 656 the CB_INTERVAL parameter MUST be replaced by max(T_rr_interval, 657 Tdr). 659 The CB_INTERVAL parameter is calculated on joining the session, and 660 recalculated on receipt of each RTCP packet, after checking whether 661 the media timeout circuit breaker or the congestion circuit breaker 662 has been triggered. 664 To ensure a timely response to persistent congestion, implementations 665 SHOULD NOT configure the RTCP bandwidth such that Tdr is larger than 666 5 seconds. Similarly, implementations that use the RTP/AVPF profile 667 [RFC4585] or the RTP/SAVPF profile [RFC5124] SHOULD NOT configure 668 T_rr_interval to values larger than 4 seconds (the reduced limit for 669 T_rr_interval follows Section 6.1.3 of 670 [I-D.ietf-avtcore-rtp-multi-stream]). 672 The rationale for enforcing a minimum sending rate below which the 673 congestion circuit breaker will not trigger is to avoid spurious 674 circuit breaker triggers when the number of packets sent per RTCP 675 reporting interval is small, and hence the fraction lost samples are 676 subject to measurement artefacts. The bound of at least one packet 677 every max(Tdr, Tr) seconds is derived from the one packet per RTT 678 minimum sending rate of TCP [RFC5405], adapted for use with RTP where 679 the RTCP reporting interval is decoupled from the network RTT. 681 When the congestion circuit breaker is triggered, the sender SHOULD 682 cease transmission (see Section 4.5). However, if the sender is able 683 to reduce its sending rate by a factor of (approximately) ten, then 684 it MAY first reduce its sending rate by this factor (or some larger 685 amount) to see if that resolves the congestion. If the sending rate 686 is reduced in this way and the congestion circuit breaker triggers 687 again after the next CB_INTERVAL RTCP reporting intervals, the sender 688 MUST then cease transmission. An example of such a rate reduction 689 might be a video conferencing system that backs off to sending audio 690 only, before completely dropping the call. If such a reduction in 691 sending rate resolves the congestion problem, the sender MAY 692 gradually increase the rate at which it sends data after a reasonable 693 amount of time has passed, provided it takes care not to cause the 694 problem to recur ("reasonable" is intentionally not defined here, 695 since it depends on the application, media codec, and congestion 696 control algorithm). 698 The RTCP reporting interval of the media sender does not affect how 699 quickly congestion circuit breaker can trigger. The timing is based 700 on the RTCP reporting interval of the receiver that generates the SR/ 701 RR packets from which the loss rate and RTT estimate are derived 702 (note that RTCP requires all participants in a session to have 703 similar reporting intervals, else the participant timeout rules in 704 [RFC3550] will not work, so this interval is likely similar to that 705 of the sender). If the incoming RTCP SR or RR packets are using a 706 reduced minimum RTCP reporting interval (as specified in Section 6.2 707 of RFC 3550 [RFC3550] or the RTP/AVPF profile [RFC4585]), then that 708 reduced RTCP reporting interval is used when determining if the 709 circuit breaker is triggered. 711 If there are more media streams that can be reported in a single RTCP 712 SR or RR packet, or if the size of a complete RTCP SR or RR packet 713 exceeds the network MTU, then the receiver will report on a subset of 714 sources in each reporting interval, with the subsets selected round- 715 robin across multiple intervals so that all sources are eventually 716 reported [RFC3550]. When generating such round-robin RTCP reports, 717 priority SHOULD be given to reports on sources that have high packet 718 loss rates, to ensure that senders are aware of network congestion 719 they are causing (this is an update to [RFC3550]). 721 4.4. RTP/AVP Circuit Breaker #4: Media Usability 723 Applications that use RTP are generally tolerant to some amount of 724 packet loss. How much packet loss can be tolerated will depend on 725 the application, media codec, and the amount of error correction and 726 packet loss concealment that is applied. There is an upper bound on 727 the amount of loss that can be corrected, however, beyond which the 728 media becomes unusable. Similarly, many applications have some upper 729 bound on the media capture to play-out latency that can be tolerated 730 before the application becomes unusable. The latency bound will 731 depend on the application, but typical values can range from the 732 order of a few hundred milliseconds for voice telephony and 733 interactive conferencing applications, up to several seconds for some 734 video-on-demand systems. 736 As a final circuit breaker, RTP senders SHOULD monitor the reported 737 packet loss and delay to estimate whether the media is likely to be 738 suitable for the intended purpose. If the packet loss rate and/or 739 latency is such that the media has become unusable, and has remained 740 unusable for a significant time period, then the application SHOULD 741 cease transmission. Similarly, receivers SHOULD monitor the quality 742 of the media they receive, and if the quality is unusable for a 743 significant time period, they SHOULD terminate the session. This 744 memo intentionally does not define a bound on the packet loss rate or 745 latency that will result in unusable media, as these are highly 746 application dependent. Similarly, the time period that is considered 747 significant is application dependent, but is likely on the order of 748 seconds, or tens of seconds. 750 Sending media that suffers from such high packet loss or latency that 751 it is unusable at the receiver is both wasteful of resources, and of 752 no benefit to the user of the application. It also is highly likely 753 to be congesting the network, and disrupting other applications. As 754 such, the congestion circuit breaker will almost certainly trigger to 755 stop flows where the media would be unusable due to high packet loss 756 or latency. However, in pathological scenarios where the congestion 757 circuit breaker does not stop the flow, it is desirable to prevent 758 the application sending unnecessary traffic that might disrupt other 759 uses of the network. The role of the media usability circuit breaker 760 is to protect the network in such cases. 762 4.5. Ceasing Transmission 764 What it means to cease transmission depends on the application. The 765 intention is that the application will stop sending RTP data packets 766 on a particular 5-tuple (transport protocol, source and destination 767 ports, source and destination IP addresses), until whatever network 768 problem that triggered the RTP circuit breaker has dissipated. This 769 could mean stopping a single RTP flow, or it could mean that multiple 770 bundled RTP flows are stopped. RTP flows halted by the circuit 771 breaker SHOULD NOT be restarted automatically unless the sender has 772 received information that the congestion has dissipated, or can 773 reasonably be expected to have dissipated. What could trigger this 774 expectation is necessarily application dependent, but could be, for 775 example, an indication that a competing flow has finished and freed 776 up some capacity, or for an application running on a mobile device, 777 that the device moved to a new location so the flow would traverse a 778 different path if it were restarted. Ideally, a human user will be 779 involved in the decision to try to restart the flow, since that user 780 will eventually give up if the flows repeatedly trigger the circuit 781 breaker. This will help avoid problems with automatic redial systems 782 from congesting the network. 784 It is recognised that the RTP implementation in some systems might 785 not be able to determine if a flow set-up request was initiated by a 786 human user, or automatically by some scripted higher-level component 787 of the system. These implementations MUST rate limit attempts to 788 restart a flow on the same 5-tuple as used by a flow that triggered 789 the circuit breaker, so that the reaction to a triggered circuit 790 breaker lasts for at least the triggering interval 791 [I-D.ietf-tsvwg-circuit-breaker]. 793 The RTP circuit breaker will only trigger, and cease transmission, 794 for media flows subject to long-term persistent congestion. Such 795 flows are likely to have poor quality and usability for some time 796 before the circuit breaker triggers. Implementations can monitor 797 RTCP Reception Report blocks being returned for their media flows, 798 and might find it beneficial to use this information to provide a 799 user interface cue that problems are occurring, in advance of the 800 circuit breaker triggering. 802 5. RTP Circuit Breakers and the RTP/AVPF and RTP/SAVPF Profiles 804 Use of the Extended RTP Profile for RTCP-based Feedback (RTP/AVPF) 805 [RFC4585] allows receivers to send early RTCP reports in some cases, 806 to inform the sender about particular events in the media stream. 807 There are several use cases for such early RTCP reports, including 808 providing rapid feedback to a sender about the onset of congestion. 809 The RTP/SAVPF Profile [RFC5124] is a secure variant of the RTP/AVPF 810 profile, that is treated the same in the context of the RTP circuit 811 breaker. These feedback profiles are often used with non-compound 812 RTCP reports [RFC5506] to reduce the reporting overhead. 814 Receiving rapid feedback about congestion events potentially allows 815 congestion control algorithms to be more responsive, and to better 816 adapt the media transmission to the limitations of the network. It 817 is expected that many RTP congestion control algorithms will adopt 818 the RTP/AVPF profile or the RTP/SAVPF profile for this reason, 819 defining new transport layer feedback reports that suit their 820 requirements. Since these reports are not yet defined, and likely 821 very specific to the details of the congestion control algorithm 822 chosen, they cannot be used as part of the generic RTP circuit 823 breaker. 825 Reduced-size RTCP reports sent under the RTP/AVPF early feedback 826 rules that do not contain an RTCP SR or RR packet MUST be ignored by 827 the congestion circuit breaker (they do not contain the information 828 needed by the congestion circuit breaker algorithm), but MUST be 829 counted as received packets for the RTCP timeout circuit breaker. 830 Reduced-size RTCP reports sent under the RTP/AVPF early feedback 831 rules that contain RTCP SR or RR packets MUST be processed by the 832 congestion circuit breaker as if they were sent as regular RTCP 833 reports, and counted towards the circuit breaker conditions specified 834 in Section 4 of this memo. This will potentially make the RTP 835 circuit breaker trigger earlier than it would if the RTP/AVPF profile 836 was not used. 838 When using ECN with RTP (see Section 7), early RTCP feedback packets 839 can contain ECN feedback reports. The count of ECN-CE marked packets 840 contained in those ECN feedback reports is counted towards the number 841 of lost packets reported if the ECN Feedback Report is sent in a 842 compound RTCP packet along with an RTCP SR/RR report packet. Reports 843 of ECN-CE packets sent as reduced-size RTCP ECN feedback packets 844 without an RTCP SR/RR packet MUST be ignored. 846 These rules are intended to allow the use of low-overhead RTP/AVPF 847 feedback for generic NACK messages without triggering the RTP circuit 848 breaker. This is expected to make such feedback suitable for RTP 849 congestion control algorithms that need to quickly report loss events 850 in between regular RTCP reports. The reaction to reduced-size RTCP 851 SR/RR packets is to allow such algorithms to send feedback that can 852 trigger the circuit breaker, when desired. 854 The RTP/AVPF and RTP/SAVPF profiles include the T_rr_interval 855 parameter that can be used to adjust the regular RTCP reporting 856 interval. The use of the T_rr_interval parameter changes the 857 behaviour of the RTP circuit breaker, as described in Section 4. 859 6. Impact of RTCP Extended Reports (XR) 861 RTCP Extended Report (XR) blocks provide additional reception quality 862 metrics, but do not change the RTCP timing rules. Some of the RTCP 863 XR blocks provide information that might be useful for congestion 864 control purposes, others provide non-congestion-related metrics. 865 With the exception of RTCP XR ECN Summary Reports (see Section 7), 866 the presence of RTCP XR blocks in a compound RTCP packet does not 867 affect the RTP circuit breaker algorithm. For consistency and ease 868 of implementation, only the reception report blocks contained in RTCP 869 SR packets, RTCP RR packets, or RTCP XR ECN Summary Report packets, 870 are used by the RTP circuit breaker algorithm. 872 7. Impact of Explicit Congestion Notification (ECN) 874 The use of ECN for RTP flows does not affect the RTCP timeout circuit 875 breaker (Section 4.1) or the media timeout circuit breaker 876 (Section 4.2), since these are both connectivity checks that simply 877 determinate if any packets are being received. 879 There is no consensus on what would be the correct response of the 880 congestion circuit breaker (Section 4.3) to ECN-CE marked packets. 881 The guidelines in [RFC3168] and [RFC6679] are that the response to 882 receipt of an ECN-CE marked packet needs to be essentially the same 883 as the response to a lost packet for congestion control purposes. 884 Since the RTP congestion circuit breaker responds to the same 885 congestion signals, this suggests that it ought to consider ECN-CE 886 marked packets as lost packets when calculating the TCP throughput 887 estimate to determine if the congestion circuit breaker triggers. 889 More recent work, however, has suggested that the response to an ECN- 890 CE mark ought to be less severe than the response to packet loss. 891 For example, the TCP ABE proposal 892 [I-D.khademi-tcpm-alternativebackoff-ecn] makes the argument that TCP 893 congestion control ought to back-off less in response to an ECN-CE 894 mark than to packet loss, because networks that generate ECN-CE marks 895 tend to use AQM schemes with much smaller buffers. For RTP 896 congestion control, both NADA [I-D.ietf-rmcat-nada] and SCREAM 897 [I-D.ietf-rmcat-scream-cc] suggest responding differently to ECN-CE 898 marked packets than to lost packets, for quality of experience 899 reasons, but make different proposals for how the response ought to 900 change. Such proposals would imply that a different circuit breaker 901 threshold be used for congestion signalled by ECN-CE marks than for 902 congestion signalled by packet loss, but unfortunately they offer no 903 clear guidance on how the threshold ought to be changed. 905 Finally, there are suggestions that forthcoming AQM proposals 906 [I-D.briscoe-aqm-dualq-coupled] might mark packets with ECN-CE in a 907 significantly more aggressive manner that at present. Any such 908 deployment would likely be incompatible with deployed TCP 909 implementations, so is not a short-term issue, but would require 910 significant changes to the congestion circuit breaker response. 912 Given the above issues, implementations MAY ignore ECN-CE marks when 913 determining if the congestion circuit breaker triggers, since 914 excessive persistent congestion will eventually lead to packet loss 915 that will trigger the circuit breaker. Doing this will protect the 916 network from congestion collapse, but might result in sub-optimal 917 user experience for competing flows that share the bottleneck queue, 918 since that queue will be driven to overflow, inducing high latency. 919 If this is a concern, the only current guidance is for 920 implementations to treat ECN-CE marked packets as equivalent to lost 921 packets, whilst being aware that this might trigger the circuit 922 breaker prematurely in future, depending on how AQM and ECN 923 deployment evolves. Developers that implement a circuit breaker 924 based on ECN-CE marks will need to track future developments in AQM 925 standards and deployed ECN marking behaviour, and ensure their 926 implementations are updated to match. 928 For the media usability circuit breaker (Section 4.4), ECN-CE marked 929 packets arrive at the receiver, and if they arrive in time, they will 930 be decoded and rendered as normal. Accordingly, receipt of such 931 packets ought not affect the usability of the media, and the arrival 932 of RTCP feedback indicating their receipt is not expected to impact 933 the operation of the media usability circuit breaker. 935 8. Impact of Bundled Media and Layered Coding 937 The RTP circuit breaker operates on a per-RTP session basis. An RTP 938 sender that participates in several RTP sessions MUST treat each RTP 939 session independently with regards to the RTP circuit breaker. 941 An RTP sender can generate several media streams within a single RTP 942 session, with each stream using a different SSRC. This can happen if 943 bundled media are in use, when using simulcast, or when using layered 944 media coding. By default, each SSRC will be treated independently by 945 the RTP circuit breaker. However, the sender MAY choose to treat the 946 flows (or a subset thereof) as a group, such that a circuit breaker 947 trigger for one flow applies to the group of flows as a whole, and 948 either causes the entire group to cease transmission, or the sending 949 rate of the group to reduce by a factor of ten, depending on the RTP 950 circuit breaker triggered. Grouping flows in this way is expected to 951 be especially useful for layered flows sent using multiple SSRCs, as 952 it allows the layered flow to react as a whole, ceasing transmission 953 on the enhancement layers first to reduce sending rate if necessary, 954 rather than treating each layer independently. Care needs to be 955 taken if the different media streams sent on a single transport layer 956 flow use different DSCP values [RFC7657], 957 [I-D.ietf-tsvwg-rtcweb-qos], since congestion could be experienced 958 differently depending on the DSCP marking. Accordingly, RTP media 959 streams with different DSCP values SHOULD NOT be considered as a 960 group when evaluating the RTP Circuit Breaker conditions. 962 9. Security Considerations 964 The security considerations of [RFC3550] apply. 966 If the RTP/AVPF profile is used to provide rapid RTCP feedback, the 967 security considerations of [RFC4585] apply. If ECN feedback for RTP 968 over UDP/IP is used, the security considerations of [RFC6679] apply. 970 If non-authenticated RTCP reports are used, an on-path attacker can 971 trivially generate fake RTCP packets that indicate high packet loss 972 rates, causing the circuit breaker to trigger and disrupt an RTP 973 session. This is somewhat more difficult for an off-path attacker, 974 due to the need to guess the randomly chosen RTP SSRC value and the 975 RTP sequence number. This attack can be avoided if RTCP packets are 976 authenticated; authentication options are discussed in [RFC7201]. 978 Timely operation of the RTP circuit breaker depends on the choice of 979 RTCP reporting interval. If the receiver has a reporting interval 980 that is overly long, then the responsiveness of the circuit breaker 981 decreases. In the limit, the RTP circuit breaker can be disabled for 982 all practical purposes by configuring an RTCP reporting interval that 983 is many minutes duration. This issue is not specific to the circuit 984 breaker: long RTCP reporting intervals also prevent reception quality 985 reports, feedback messages, codec control messages, etc., from being 986 used. Implementations are expected to impose an upper limit on the 987 RTCP reporting interval they are willing to negotiate (based on the 988 session bandwidth and RTCP bandwidth fraction) when using the RTP 989 circuit breaker, as discussed in Section 4.3. 991 10. IANA Considerations 993 There are no actions for IANA. 995 11. Acknowledgements 997 The authors would like to thank Bernard Aboba, Harald Alvestrand, Ben 998 Campbell, Alissa Cooper, Spencer Dawkins, Gorry Fairhurst, Stephen 999 Farrell, Nazila Fough, Kevin Gross, Cullen Jennings, Randell Jesup, 1000 Mirja Kuehlewind, Jonathan Lennox, Matt Mathis, Stephen McQuistin, 1001 Simon Perreault, Eric Rescorla, Abheek Saha, Meral Shirazipour, Fabio 1002 Verdicchio, and Magnus Westerlund for their valuable feedback. 1004 12. References 1006 12.1. Normative References 1008 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 1009 Requirement Levels", BCP 14, RFC 2119, 1010 DOI 10.17487/RFC2119, March 1997, 1011 . 1013 [RFC3550] Schulzrinne, H., Casner, S., Frederick, R., and V. 1014 Jacobson, "RTP: A Transport Protocol for Real-Time 1015 Applications", STD 64, RFC 3550, DOI 10.17487/RFC3550, 1016 July 2003, . 1018 [RFC3551] Schulzrinne, H. and S. Casner, "RTP Profile for Audio and 1019 Video Conferences with Minimal Control", STD 65, RFC 3551, 1020 DOI 10.17487/RFC3551, July 2003, 1021 . 1023 [RFC3611] Friedman, T., Ed., Caceres, R., Ed., and A. Clark, Ed., 1024 "RTP Control Protocol Extended Reports (RTCP XR)", 1025 RFC 3611, DOI 10.17487/RFC3611, November 2003, 1026 . 1028 [RFC4585] Ott, J., Wenger, S., Sato, N., Burmeister, C., and J. Rey, 1029 "Extended RTP Profile for Real-time Transport Control 1030 Protocol (RTCP)-Based Feedback (RTP/AVPF)", RFC 4585, 1031 DOI 10.17487/RFC4585, July 2006, 1032 . 1034 [RFC5348] Floyd, S., Handley, M., Padhye, J., and J. Widmer, "TCP 1035 Friendly Rate Control (TFRC): Protocol Specification", 1036 RFC 5348, DOI 10.17487/RFC5348, September 2008, 1037 . 1039 [RFC6679] Westerlund, M., Johansson, I., Perkins, C., O'Hanlon, P., 1040 and K. Carlberg, "Explicit Congestion Notification (ECN) 1041 for RTP over UDP", RFC 6679, DOI 10.17487/RFC6679, August 1042 2012, . 1044 12.2. Informative References 1046 [Floyd] Floyd, S., Handley, M., Padhye, J., and J. Widmer, 1047 "Equation-Based Congestion Control for Unicast 1048 Applications", Proceedings of the ACM SIGCOMM 1049 conference, 2000, DOI 10.1145/347059.347397, August 2000. 1051 [I-D.briscoe-aqm-dualq-coupled] 1052 Schepper, K., Briscoe, B., Bondarenko, O., and I. Tsang, 1053 "DualQ Coupled AQM for Low Latency, Low Loss and Scalable 1054 Throughput", draft-briscoe-aqm-dualq-coupled-01 (work in 1055 progress), March 2016. 1057 [I-D.ietf-avtcore-rtp-multi-stream] 1058 Lennox, J., Westerlund, M., Wu, Q., and C. Perkins, 1059 "Sending Multiple RTP Streams in a Single RTP Session", 1060 draft-ietf-avtcore-rtp-multi-stream-11 (work in progress), 1061 December 2015. 1063 [I-D.ietf-rmcat-nada] 1064 Zhu, X., Pan, R., Ramalho, M., Cruz, S., Jones, P., Fu, 1065 J., D'Aronco, S., and C. Ganzhorn, "NADA: A Unified 1066 Congestion Control Scheme for Real-Time Media", draft- 1067 ietf-rmcat-nada-02 (work in progress), March 2016. 1069 [I-D.ietf-rmcat-scream-cc] 1070 Johansson, I. and Z. Sarker, "Self-Clocked Rate Adaptation 1071 for Multimedia", draft-ietf-rmcat-scream-cc-04 (work in 1072 progress), June 2016. 1074 [I-D.ietf-tsvwg-circuit-breaker] 1075 Fairhurst, G., "Network Transport Circuit Breakers", 1076 draft-ietf-tsvwg-circuit-breaker-15 (work in progress), 1077 April 2016. 1079 [I-D.ietf-tsvwg-rtcweb-qos] 1080 Jones, P., Dhesikan, S., Jennings, C., and D. Druta, "DSCP 1081 Packet Markings for WebRTC QoS", draft-ietf-tsvwg-rtcweb- 1082 qos-17 (work in progress), May 2016. 1084 [I-D.khademi-tcpm-alternativebackoff-ecn] 1085 Khademi, N., Welzl, M., Armitage, G., and G. Fairhurst, 1086 "TCP Alternative Backoff with ECN (ABE)", draft-khademi- 1087 tcpm-alternativebackoff-ecn-00 (work in progress), May 1088 2016. 1090 [Mathis] Mathis, M., Semke, J., Mahdavi, J., and T. Ott, "The 1091 macroscopic behavior of the TCP congestion avoidance 1092 algorithm", ACM SIGCOMM Computer Communication 1093 Review 27(3), DOI 10.1145/263932.264023, July 1997. 1095 [Padhye] Padhye, J., Firoiu, V., Towsley, D., and J. Kurose, 1096 "Modeling TCP Throughput: A Simple Model and its Empirical 1097 Validation", Proceedings of the ACM SIGCOMM 1098 conference, 1998, DOI 10.1145/285237.285291, August 1998. 1100 [RFC2862] Civanlar, M. and G. Cash, "RTP Payload Format for Real- 1101 Time Pointers", RFC 2862, DOI 10.17487/RFC2862, June 2000, 1102 . 1104 [RFC3168] Ramakrishnan, K., Floyd, S., and D. Black, "The Addition 1105 of Explicit Congestion Notification (ECN) to IP", 1106 RFC 3168, DOI 10.17487/RFC3168, September 2001, 1107 . 1109 [RFC4103] Hellstrom, G. and P. Jones, "RTP Payload for Text 1110 Conversation", RFC 4103, DOI 10.17487/RFC4103, June 2005, 1111 . 1113 [RFC5104] Wenger, S., Chandra, U., Westerlund, M., and B. Burman, 1114 "Codec Control Messages in the RTP Audio-Visual Profile 1115 with Feedback (AVPF)", RFC 5104, DOI 10.17487/RFC5104, 1116 February 2008, . 1118 [RFC5124] Ott, J. and E. Carrara, "Extended Secure RTP Profile for 1119 Real-time Transport Control Protocol (RTCP)-Based Feedback 1120 (RTP/SAVPF)", RFC 5124, DOI 10.17487/RFC5124, February 1121 2008, . 1123 [RFC5405] Eggert, L. and G. Fairhurst, "Unicast UDP Usage Guidelines 1124 for Application Designers", BCP 145, RFC 5405, 1125 DOI 10.17487/RFC5405, November 2008, 1126 . 1128 [RFC5450] Singer, D. and H. Desineni, "Transmission Time Offsets in 1129 RTP Streams", RFC 5450, DOI 10.17487/RFC5450, March 2009, 1130 . 1132 [RFC5506] Johansson, I. and M. Westerlund, "Support for Reduced-Size 1133 Real-Time Transport Control Protocol (RTCP): Opportunities 1134 and Consequences", RFC 5506, DOI 10.17487/RFC5506, April 1135 2009, . 1137 [RFC5681] Allman, M., Paxson, V., and E. Blanton, "TCP Congestion 1138 Control", RFC 5681, DOI 10.17487/RFC5681, September 2009, 1139 . 1141 [RFC6798] Clark, A. and Q. Wu, "RTP Control Protocol (RTCP) Extended 1142 Report (XR) Block for Packet Delay Variation Metric 1143 Reporting", RFC 6798, DOI 10.17487/RFC6798, November 2012, 1144 . 1146 [RFC6843] Clark, A., Gross, K., and Q. Wu, "RTP Control Protocol 1147 (RTCP) Extended Report (XR) Block for Delay Metric 1148 Reporting", RFC 6843, DOI 10.17487/RFC6843, January 2013, 1149 . 1151 [RFC6958] Clark, A., Zhang, S., Zhao, J., and Q. Wu, Ed., "RTP 1152 Control Protocol (RTCP) Extended Report (XR) Block for 1153 Burst/Gap Loss Metric Reporting", RFC 6958, 1154 DOI 10.17487/RFC6958, May 2013, 1155 . 1157 [RFC7002] Clark, A., Zorn, G., and Q. Wu, "RTP Control Protocol 1158 (RTCP) Extended Report (XR) Block for Discard Count Metric 1159 Reporting", RFC 7002, DOI 10.17487/RFC7002, September 1160 2013, . 1162 [RFC7003] Clark, A., Huang, R., and Q. Wu, Ed., "RTP Control 1163 Protocol (RTCP) Extended Report (XR) Block for Burst/Gap 1164 Discard Metric Reporting", RFC 7003, DOI 10.17487/RFC7003, 1165 September 2013, . 1167 [RFC7097] Ott, J., Singh, V., Ed., and I. Curcio, "RTP Control 1168 Protocol (RTCP) Extended Report (XR) for RLE of Discarded 1169 Packets", RFC 7097, DOI 10.17487/RFC7097, January 2014, 1170 . 1172 [RFC7201] Westerlund, M. and C. Perkins, "Options for Securing RTP 1173 Sessions", RFC 7201, DOI 10.17487/RFC7201, April 2014, 1174 . 1176 [RFC7657] Black, D., Ed. and P. Jones, "Differentiated Services 1177 (Diffserv) and Real-Time Communication", RFC 7657, 1178 DOI 10.17487/RFC7657, November 2015, 1179 . 1181 [RFC7713] Mathis, M. and B. Briscoe, "Congestion Exposure (ConEx) 1182 Concepts, Abstract Mechanism, and Requirements", RFC 7713, 1183 DOI 10.17487/RFC7713, December 2015, 1184 . 1186 [Sarker] Sarker, Z., Singh, V., and C. Perkins, "An Evaluation of 1187 RTP Circuit Breaker Performance on LTE Networks", 1188 Proceedings of the IEEE Infocom workshop on Communication 1189 and Networking Techniques for Contemporary Video, 2014, 1190 April 2014. 1192 [Singh] Singh, V., McQuistin, S., Ellis, M., and C. Perkins, 1193 "Circuit Breakers for Multimedia Congestion Control", 1194 Proceedings of the International Packet Video 1195 Workshop, 2013, DOI 10.1109/PV.2013.6691439, December 1196 2013. 1198 Authors' Addresses 1200 Colin Perkins 1201 University of Glasgow 1202 School of Computing Science 1203 Glasgow G12 8QQ 1204 United Kingdom 1206 Email: csp@csperkins.org 1208 Varun Singh 1209 Nemu Dialogue Systems Oy 1210 Runeberginkatu 4c A 4 1211 Helsinki 00100 1212 Finland 1214 Email: varun.singh@iki.fi 1215 URI: http://www.callstats.io/