idnits 2.17.1 draft-ietf-avtcore-rtp-circuit-breakers-12.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 (February 9, 2016) is 2996 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(See RFCs 3967 and 4897 for information about using normative references to lower-maturity documents in RFCs) == Outdated reference: A later version (-15) exists of draft-ietf-tsvwg-circuit-breaker-11 -- Obsolete informational reference (is this intentional?): RFC 5405 (Obsoleted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: August 12, 2016 February 9, 2016 8 Multimedia Congestion Control: Circuit Breakers for Unicast RTP Sessions 9 draft-ietf-avtcore-rtp-circuit-breakers-12 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 the applications, then network congestion will 17 deteriorate the user's multimedia experience. This acts as a safety 18 measure to prevent starvation of network resources denying other 19 flows from access to the Internet, such measures are essential for an 20 Internet that is heterogeneous and for traffic that is hard to 21 predict in advance. This document does not propose a congestion 22 control algorithm; instead, it defines a minimal set of RTP circuit- 23 breakers. Circuit-breakers are conditions under which an RTP sender 24 needs to stop transmitting media data in order to protect the network 25 from excessive congestion. It is expected that, in the absence of 26 severe congestion, all RTP applications running on best-effort IP 27 networks will be able to run without triggering these circuit 28 breakers. Any future RTP congestion control specification will be 29 expected to operate within the constraints defined by these circuit 30 breakers. 32 Status of This Memo 34 This Internet-Draft is submitted in full conformance with the 35 provisions of BCP 78 and BCP 79. 37 Internet-Drafts are working documents of the Internet Engineering 38 Task Force (IETF). Note that other groups may also distribute 39 working documents as Internet-Drafts. The list of current Internet- 40 Drafts is at http://datatracker.ietf.org/drafts/current/. 42 Internet-Drafts are draft documents valid for a maximum of six months 43 and may be updated, replaced, or obsoleted by other documents at any 44 time. It is inappropriate to use Internet-Drafts as reference 45 material or to cite them other than as "work in progress." 47 This Internet-Draft will expire on August 12, 2016. 49 Copyright Notice 51 Copyright (c) 2016 IETF Trust and the persons identified as the 52 document authors. All rights reserved. 54 This document is subject to BCP 78 and the IETF Trust's Legal 55 Provisions Relating to IETF Documents 56 (http://trustee.ietf.org/license-info) in effect on the date of 57 publication of this document. Please review these documents 58 carefully, as they describe your rights and restrictions with respect 59 to this document. Code Components extracted from this document must 60 include Simplified BSD License text as described in Section 4.e of 61 the Trust Legal Provisions and are provided without warranty as 62 described in the Simplified BSD License. 64 Table of Contents 66 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 2 67 2. Background . . . . . . . . . . . . . . . . . . . . . . . . . 3 68 3. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . 6 69 4. RTP Circuit Breakers for Systems Using the RTP/AVP Profile . 7 70 4.1. RTP/AVP Circuit Breaker #1: RTCP Timeout . . . . . . . . 9 71 4.2. RTP/AVP Circuit Breaker #2: Media Timeout . . . . . . . . 10 72 4.3. RTP/AVP Circuit Breaker #3: Congestion . . . . . . . . . 12 73 4.4. RTP/AVP Circuit Breaker #4: Media Usability . . . . . . . 15 74 4.5. Ceasing Transmission . . . . . . . . . . . . . . . . . . 16 75 5. RTP Circuit Breakers and the RTP/AVPF and RTP/SAVPF Profiles 17 76 6. Impact of RTCP Extended Reports (XR) . . . . . . . . . . . . 18 77 7. Impact of Explicit Congestion Notification (ECN) . . . . . . 18 78 8. Impact of Bundled Media and Layered Coding . . . . . . . . . 19 79 9. Security Considerations . . . . . . . . . . . . . . . . . . . 19 80 10. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 20 81 11. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 20 82 12. References . . . . . . . . . . . . . . . . . . . . . . . . . 20 83 12.1. Normative References . . . . . . . . . . . . . . . . . . 20 84 12.2. Informative References . . . . . . . . . . . . . . . . . 21 85 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 23 87 1. Introduction 89 The Real-time Transport Protocol (RTP) [RFC3550] is widely used in 90 voice-over-IP, video teleconferencing, and telepresence systems. 91 Many of these systems run over best-effort UDP/IP networks, and can 92 suffer from packet loss and increased latency if network congestion 93 occurs. Designing effective RTP congestion control algorithms, to 94 adapt the transmission of RTP-based media to match the available 95 network capacity, while also maintaining the user experience, is a 96 difficult but important problem. Many such congestion control and 97 media adaptation algorithms have been proposed, but to date there is 98 no consensus on the correct approach, or even that a single standard 99 algorithm is desirable. 101 This memo does not attempt to propose a new RTP congestion control 102 algorithm. Instead, we propose a small set of RTP circuit breakers. 103 These are conditions under which there is general agreement that an 104 RTP flow is causing serious congestion, and hence ought to cease 105 transmission. The RTP circuit breakers proposed in this memo are a 106 specific instance of the general class of network transport circuit 107 breakers [I-D.ietf-tsvwg-circuit-breaker], designed to act as a 108 protection mechanism of last resort to avoid persistent excessive 109 congestion. It is expected that future standards-track congestion 110 control algorithms for RTP will operate within the envelope defined 111 by this memo. 113 2. Background 115 We consider congestion control for unicast RTP traffic flows. This 116 is the problem of adapting the transmission of an audio/visual data 117 flow, encapsulated within an RTP transport session, from one sender 118 to one receiver, so that it does not use more capacity than is 119 available along the network path. Such adaptation needs to be done 120 in a way that limits the disruption to the user experience caused by 121 both packet loss and excessive rate changes. Congestion control for 122 multicast flows is outside the scope of this memo. Multicast traffic 123 needs different solutions, since the available capacity estimator for 124 a group of receivers will differ from that for a single receiver, and 125 because multicast congestion control has to consider issues of 126 fairness across groups of receivers that do not apply to unicast 127 flows. 129 Congestion control for unicast RTP traffic can be implemented in one 130 of two places in the protocol stack. One approach is to run the RTP 131 traffic over a congestion controlled transport protocol, for example 132 over TCP, and to adapt the media encoding to match the dictates of 133 the transport-layer congestion control algorithm. This is safe for 134 the network, but can be suboptimal for the media quality unless the 135 transport protocol is designed to support real-time media flows. We 136 do not consider this class of applications further in this memo, as 137 their network safety is guaranteed by the underlying transport. 139 Alternatively, RTP flows can be run over a non-congestion controlled 140 transport protocol, for example UDP, performing rate adaptation at 141 the application layer based on RTP Control Protocol (RTCP) feedback. 142 With a well-designed, network-aware, application, this allows highly 143 effective media quality adaptation, but there is potential to cause 144 persistent congestion in the network if the application does not 145 adapt its sending rate in a timely and effective manner. We consider 146 this class of applications in this memo. 148 Congestion control relies on monitoring the delivery of a media flow, 149 and responding to adapt the transmission of that flow when there are 150 signs that the network path is congested. Network congestion can be 151 detected in one of three ways: 1) a receiver can infer the onset of 152 congestion by observing an increase in one-way delay caused by queue 153 build-up within the network; 2) if Explicit Congestion Notification 154 (ECN) [RFC3168] is supported, the network can signal the presence of 155 congestion by marking packets using ECN Congestion Experienced (CE) 156 marks; or 3) in the extreme case, congestion will cause packet loss 157 that can be detected by observing a gap in the received RTP sequence 158 numbers. 160 Once the onset of congestion is observed, the receiver has to send 161 feedback to the sender to indicate that the transmission rate needs 162 to be reduced. How the sender reduces the transmission rate is 163 highly dependent on the media codec being used, and is outside the 164 scope of this memo. 166 There are several ways in which a receiver can send feedback to a 167 media sender within the RTP framework: 169 o The base RTP specification [RFC3550] defines RTCP Reception Report 170 (RR) packets to convey reception quality feedback information, and 171 Sender Report (SR) packets to convey information about the media 172 transmission. RTCP SR packets contain data that can be used to 173 reconstruct media timing at a receiver, along with a count of the 174 total number of octets and packets sent. RTCP RR packets report 175 on the fraction of packets lost in the last reporting interval, 176 the cumulative number of packets lost, the highest sequence number 177 received, and the inter-arrival jitter. The RTCP RR packets also 178 contain timing information that allows the sender to estimate the 179 network round trip time (RTT) to the receivers. RTCP reports are 180 sent periodically, with the reporting interval being determined by 181 the number of SSRCs used in the session and a configured session 182 bandwidth estimate (the number of synchronisation sources (SSRCs) 183 used is usually two in a unicast session, one for each 184 participant, but can be greater if the participants send multiple 185 media streams). The interval between reports sent from each 186 receiver tends to be on the order of a few seconds on average, 187 although it varies with the session bandwidth, and sub-second 188 reporting intervals are possible in high bandwidth sessions, and 189 it is randomised to avoid synchronisation of reports from multiple 190 receivers. RTCP RR packets allow a receiver to report ongoing 191 network congestion to the sender. However, if a receiver detects 192 the onset of congestion part way through a reporting interval, the 193 base RTP specification contains no provision for sending the RTCP 194 RR packet early, and the receiver has to wait until the next 195 scheduled reporting interval. 197 o The RTCP Extended Reports (XR) [RFC3611] allow reporting of more 198 complex and sophisticated reception quality metrics, but do not 199 change the RTCP timing rules. RTCP extended reports of potential 200 interest for congestion control purposes are the extended packet 201 loss, discard, and burst metrics [RFC3611], [RFC7002], [RFC7097], 202 [RFC7003], [RFC6958]; and the extended delay metrics [RFC6843], 203 [RFC6798]. Other RTCP Extended Reports that could be helpful for 204 congestion control purposes might be developed in future. 206 o Rapid feedback about the occurrence of congestion events can be 207 achieved using the Extended RTP Profile for RTCP-Based Feedback 208 (RTP/AVPF) [RFC4585] (or its secure variant, RTP/SAVPF [RFC5124]) 209 in place of the RTP/AVP profile [RFC3551]. This modifies the RTCP 210 timing rules to allow RTCP reports to be sent early, in some cases 211 immediately, provided the RTCP transmission rate keeps within its 212 bandwidth allocation. It also defines transport-layer feedback 213 messages, including negative acknowledgements (NACKs), that can be 214 used to report on specific congestion events. RTP Codec Control 215 Messages [RFC5104] extend the RTP/AVPF profile with additional 216 feedback messages that can be used to influence that way in which 217 rate adaptation occurs, but do not further change the dynamics of 218 how rapidly feedback can be sent. Use of the RTP/AVPF profile is 219 dependent on signalling. 221 o Finally, Explicit Congestion Notification (ECN) for RTP over UDP 222 [RFC6679] can be used to provide feedback on the number of packets 223 that received an ECN Congestion Experienced (CE) mark. This RTCP 224 extension builds on the RTP/AVPF profile to allow rapid congestion 225 feedback when ECN is supported. 227 In addition to these mechanisms for providing feedback, the sender 228 can include an RTP header extension in each packet to record packet 229 transmission times [RFC5450]. Accurate transmission timestamps can 230 be helpful for estimating queuing delays, to get an early indication 231 of the onset of congestion. 233 Taken together, these various mechanisms allow receivers to provide 234 feedback on the senders when congestion events occur, with varying 235 degrees of timeliness and accuracy. The key distinction is between 236 systems that use only the basic RTCP mechanisms, without RTP/AVPF 237 rapid feedback, and those that use the RTP/AVPF extensions to respond 238 to congestion more rapidly. 240 3. Terminology 242 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 243 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 244 document are to be interpreted as described in RFC 2119 [RFC2119]. 245 This interpretation of these key words applies only when written in 246 ALL CAPS. Mixed- or lower-case uses of these key words are not to be 247 interpreted as carrying special significance in this memo. 249 The definition of the RTP circuit breaker is specified in terms of 250 the following variables: 252 o Td is the deterministic RTCP reporting interval, as defined in 253 Section 6.3.1 of [RFC3550]. 255 o Tdr is the sender's estimate of the deterministic RTCP reporting 256 interval, Td, calculated by a receiver of the data it is sending. 257 Tdr is not known at the sender, but can be estimated by executing 258 the algorithm in Section 6.2 of [RFC3550] using the average RTCP 259 packet size seen at the sender, the number of members reported in 260 the receiver's SR/RR report blocks, and whether the receiver is 261 sending SR or RR packets. Tdr is recalculated when each new RTCP 262 SR/RR report is received, but the media timeout circuit breaker 263 (see Section 4.2) is only reconsidered when Tdr increases. 265 o Tr is the network round-trip time, calculated by the sender using 266 the algorithm in Section 6.4.1 of [RFC3550] and smoothed using an 267 exponentially weighted moving average as Tr = (0.8 * Tr) + (0.2 * 268 Tr_new) where Tr_new is the latest RTT estimate obtained from an 269 RTCP report. The weight is chosen so old estimates decay over k 270 intervals. 272 o k is the non-reporting threshold (see Section 4.2). 274 o Tf is the media framing interval at the sender. For applications 275 sending at a constant frame rate, Tf is the inter-frame interval. 276 For applications that switch between a small set of possible frame 277 rates, for example when sending speech with comfort noise, where 278 comfort noise frames are sent less often than speech frames, Tf is 279 set to the longest of the inter-frame intervals of the different 280 frame rates. For applications that send periodic frames but 281 dynamically vary their frame rate, Tf is set to the largest inter- 282 frame interval used in the last 10 seconds. For applications that 283 send less than one frame every 10 seconds, or that have no concept 284 of periodic frames (e.g., text conversation [RFC4103], or pointer 285 events [RFC2862]), Tf is set to the time interval since the 286 previous frame when each frame is sent. 288 o G is the frame group size. That is, the number of frames that are 289 coded together based on a particular sending rate setting. If the 290 codec used by the sender can change its rate on each frame, G = 1; 291 otherwise G is set to the number of frames before the codec can 292 adjust to the new rate. For codecs that have the concept of a 293 group-of-pictures (GoP), G is likely the GoP length. 295 o T_rr_interval is the minimal interval between RTCP reports, as 296 defined in Section 3.4 of [RFC4585]; it is only meaningful for 297 implementations of RTP/AVPF profile [RFC4585] or the RTP/SAVPF 298 profile [RFC5124]. 300 o X is the estimated throughput a TCP connection would achieve over 301 a path, in bytes per second. 303 o s is the size of RTP packets being sent, in bytes. If the RTP 304 packets being sent vary in size, then the average size over the 305 packet comprising the last 4 * G frames MUST be used (this is 306 intended to be comparable to the four loss intervals used in 307 [RFC5348]). 309 o p is the loss event rate, between 0.0 and 1.0, that would be seen 310 by a TCP connection over a particular path. When used in the RTP 311 congestion circuit breaker, this is approximated as described in 312 Section 4.3. 314 o t_RTO is the retransmission timeout value that would be used by a 315 TCP connection over a particular path, in seconds. This MUST be 316 approximated using t_RTO = 4 * Tr when used as part of the RTP 317 congestion circuit breaker. 319 o b is the number of packets that are acknowledged by a single TCP 320 acknowledgement. Following [RFC5348], it is RECOMMENDED that the 321 value b = 1 is used as part of the RTP congestion circuit breaker. 323 4. RTP Circuit Breakers for Systems Using the RTP/AVP Profile 325 The feedback mechanisms defined in [RFC3550] and available under the 326 RTP/AVP profile [RFC3551] are the minimum that can be assumed for a 327 baseline circuit breaker mechanism that is suitable for all unicast 328 applications of RTP. Accordingly, for an RTP circuit breaker to be 329 useful, it needs to be able to detect that an RTP flow is causing 330 excessive congestion using only basic RTCP features, without needing 331 RTCP XR feedback or the RTP/AVPF profile for rapid RTCP reports. 333 RTCP is a fundamental part of the RTP protocol, and the mechanisms 334 described here rely on the implementation of RTCP. Implementations 335 that claim to support RTP, but that do not implement RTCP, will be 336 unable to use the circuit breaker mechanisms described in this memo. 337 Such implementations SHOULD NOT be used on networks that might be 338 subject to congestion unless equivalent mechanisms are defined using 339 some non-RTCP feedback channel to report congestion and signal 340 circuit breaker conditions. 342 The RTCP timeout circuit breaker (Section 4.1) will trigger if an 343 implementation of this memo attempts to interwork with an endpoint 344 that does not support RTCP. Implementations that sometimes need to 345 interwork with endpoints that do not support RTCP need to disable the 346 RTP circuit breakers if they don't receive some confirmation via 347 signalling that the remote endpoint implements RTCP (the presence of 348 an SDP "a=rtcp:" attribute in an answer might be such an indication). 349 This approach SHOULD NOT be used on networks that might be subject to 350 congestion unless equivalent mechanisms are defined using some non- 351 RTCP feedback channel to report congestion and signal circuit breaker 352 conditions [I-D.ietf-tsvwg-circuit-breaker]. 354 Three potential congestion signals are available from the basic RTCP 355 SR/RR packets and are reported for each SSRC in the RTP session: 357 1. The sender can estimate the network round-trip time once per RTCP 358 reporting interval, based on the contents and timing of RTCP SR 359 and RR packets. 361 2. Receivers report a jitter estimate (the statistical variance of 362 the RTP data packet inter-arrival time) calculated over the RTCP 363 reporting interval. Due to the nature of the jitter calculation 364 ([RFC3550], section 6.4.4), the jitter is only meaningful for RTP 365 flows that send a single data packet for each RTP timestamp value 366 (i.e., audio flows, or video flows where each packet comprises 367 one video frame). 369 3. Receivers report the fraction of RTP data packets lost during the 370 RTCP reporting interval, and the cumulative number of RTP packets 371 lost over the entire RTP session. 373 These congestion signals limit the possible circuit breakers, since 374 they give only limited visibility into the behaviour of the network. 376 RTT estimates are widely used in congestion control algorithms, as a 377 proxy for queuing delay measures in delay-based congestion control or 378 to determine connection timeouts. RTT estimates derived from RTCP SR 379 and RR packets sent according to the RTP/AVP timing rules are too 380 infrequent to be useful for congestion control, and don't give enough 381 information to distinguish a delay change due to routing updates from 382 queuing delay caused by congestion. Accordingly, we cannot use the 383 RTT estimate alone as an RTP circuit breaker. 385 Increased jitter can be a signal of transient network congestion, but 386 in the highly aggregated form reported in RTCP RR packets, it offers 387 insufficient information to estimate the extent or persistence of 388 congestion. Jitter reports are a useful early warning of potential 389 network congestion, but provide an insufficiently strong signal to be 390 used as a circuit breaker. 392 The remaining congestion signals are the packet loss fraction and the 393 cumulative number of packets lost. If considered carefully, and over 394 an appropriate time frame to distinguish transient problems from long 395 term issues [I-D.ietf-tsvwg-circuit-breaker], these can be effective 396 indicators that persistent excessive congestion is occurring in 397 networks where packet loss is primarily due to queue overflows, 398 although loss caused by non-congestive packet corruption can distort 399 the result in some networks. TCP congestion control [RFC5681] 400 intentionally tries to fill the router queues, and uses the resulting 401 packet loss as congestion feedback. An RTP flow competing with TCP 402 traffic will therefore expect to see a non-zero packet loss fraction, 403 and some variation in queuing latency, in normal operation when 404 sharing a path with other flows, that needs to be accounted for when 405 determining the circuit breaker threshold 406 [I-D.ietf-tsvwg-circuit-breaker]. This behaviour of TCP is reflected 407 in the congestion circuit breaker below, and will affect the design 408 of any RTP congestion control protocol. 410 Two packet loss regimes can be observed: 1) RTCP RR packets show a 411 non-zero packet loss fraction, while the extended highest sequence 412 number received continues to increment; and 2) RR packets show a loss 413 fraction of zero, but the extended highest sequence number received 414 does not increment even though the sender has been transmitting RTP 415 data packets. The former corresponds to the TCP congestion avoidance 416 state, and indicates a congested path that is still delivering data; 417 the latter corresponds to a TCP timeout, and is most likely due to a 418 path failure. A third condition is that data is being sent but no 419 RTCP feedback is received at all, corresponding to a failure of the 420 reverse path. We derive circuit breaker conditions for these loss 421 regimes in the following. 423 4.1. RTP/AVP Circuit Breaker #1: RTCP Timeout 425 An RTCP timeout can occur when RTP data packets are being sent, but 426 there are no RTCP reports returned from the receiver. This is either 427 due to a failure of the receiver to send RTCP reports, or a failure 428 of the return path that is preventing those RTCP reporting from being 429 delivered. In either case, it is not safe to continue transmission, 430 since the sender has no way of knowing if it is causing congestion. 432 An RTP sender that has not received any RTCP SR or RTCP RR packets 433 reporting on the SSRC it is using, for a time period of at least 434 three times its deterministic RTCP reporting interval, Td, without 435 the randomization factor, and using the fixed minimum interval of 436 Tmin=5 seconds, SHOULD cease transmission (see Section 4.5). The 437 rationale for this choice of timeout is as described in Section 6.2 438 of [RFC3550] ("so that implementations which do not use the reduced 439 value for transmitting RTCP packets are not timed out by other 440 participants prematurely"), as updated by Section 6.1.4 of 441 [I-D.ietf-avtcore-rtp-multi-stream] to account for the use of the 442 RTP/AVPF profile [RFC4585] or the RTP/SAVPF profile [RFC5124]. 444 To reduce the risk of premature timeout, implementations SHOULD NOT 445 configure the RTCP bandwidth such that Td is larger than 5 seconds. 446 Similarly, implementations that use the RTP/AVPF profile [RFC4585] or 447 the RTP/SAVPF profile [RFC5124] SHOULD NOT configure T_rr_interval to 448 values larger than 4 seconds (the reduced limit for T_rr_interval 449 follows Section 6.1.3 of [I-D.ietf-avtcore-rtp-multi-stream]). 451 The choice of three RTCP reporting intervals as the timeout is made 452 following Section 6.3.5 of RFC 3550 [RFC3550]. This specifies that 453 participants in an RTP session will timeout and remove an RTP sender 454 from the list of active RTP senders if no RTP data packets have been 455 received from that RTP sender within the last two RTCP reporting 456 intervals. Using a timeout of three RTCP reporting intervals is 457 therefore large enough that the other participants will have timed 458 out the sender if a network problem stops the data packets it is 459 sending from reaching the receivers, even allowing for loss of some 460 RTCP packets. 462 If a sender is transmitting a large number of RTP media streams, such 463 that the corresponding RTCP SR or RR packets are too large to fit 464 into the network MTU, the receiver will generate RTCP SR or RR 465 packets in a round-robin manner. In this case, the sender SHOULD 466 treat receipt of an RTCP SR or RR packet corresponding to any SSRC it 467 sent on the same 5-tuple of source and destination IP address, port, 468 and protocol, as an indication that the receiver and return path are 469 working, preventing the RTCP timeout circuit breaker from triggering. 471 4.2. RTP/AVP Circuit Breaker #2: Media Timeout 473 If RTP data packets are being sent, but the RTCP SR or RR packets 474 reporting on that SSRC indicate a non-increasing extended highest 475 sequence number received, this is an indication that those RTP data 476 packets are not reaching the receiver. This could be a short-term 477 issue affecting only a few RTP packets, perhaps caused by a slow to 478 open firewall or a transient connectivity problem, but if the issue 479 persists, it is a sign of a more ongoing and significant problem (a 480 "media timeout"). 482 The time needed to declare a media timeout depends on the parameters 483 Tdr, Tr, Tf, and on the non-reporting threshold k. The value of k is 484 chosen so that when Tdr is large compared to Tr and Tf, receipt of at 485 least k RTCP reports with non-increasing extended highest sequence 486 number received gives reasonable assurance that the forward path has 487 failed, and that the RTP data packets have not been lost by chance. 488 The RECOMMENDED value for k is 5 reports. 490 When Tdr < Tf, then RTP data packets are being sent at a rate less 491 than one per RTCP reporting interval of the receiver, so the extended 492 highest sequence number received can be expected to be non-increasing 493 for some receiver RTCP reporting intervals. Similarly, when Tdr < 494 Tr, some receiver RTCP reporting intervals might pass before the RTP 495 data packets arrive at the receiver, also leading to reports where 496 the extended highest sequence number received is non-increasing. 497 Both issues require the media timeout interval to be scaled relative 498 to the threshold, k. 500 The media timeout RTP circuit breaker is therefore as follows. When 501 starting sending, calculate MEDIA_TIMEOUT using: 503 MEDIA_TIMEOUT = ceil(k * max(Tf, Tr, Tdr) / Tdr) 505 When a sender receives an RTCP packet that indicates reception of the 506 media it has been sending, then it cancels the media timeout circuit 507 breaker. If it is still sending, then it MUST calculate a new value 508 for MEDIA_TIMEOUT, and set a new media timeout circuit breaker. 510 If a sender receives an RTCP packet indicating that its media was not 511 received, it MUST calculate a new value for MEDIA_TIMEOUT. If the 512 new value is larger than the previous, it replaces MEDIA_TIMEOUT with 513 the new value, extending the media timeout circuit breaker; otherwise 514 it keeps the original value of MEDIA_TIMEOUT. This process is known 515 as reconsidering the media timeout circuit breaker. 517 If MEDIA_TIMEOUT consecutive RTCP packets are received indicating 518 that the media being sent was not received, and the media timeout 519 circuit breaker has not been cancelled, then the media timeout 520 circuit breaker triggers. When the media timeout circuit breaker 521 triggers, the sender SHOULD cease transmission (see Section 4.5). 523 When stopping sending an RTP stream, a sender MUST cancel the 524 corresponding media timeout circuit breaker. 526 4.3. RTP/AVP Circuit Breaker #3: Congestion 528 If RTP data packets are being sent, and the corresponding RTCP SR or 529 RR packets show non-zero packet loss fraction and increasing extended 530 highest sequence number received, then those RTP data packets are 531 arriving at the receiver, but some degree of congestion is occurring. 532 The RTP/AVP profile [RFC3551] states that: 534 If best-effort service is being used, RTP receivers SHOULD monitor 535 packet loss to ensure that the packet loss rate is within 536 acceptable parameters. Packet loss is considered acceptable if a 537 TCP flow across the same network path and experiencing the same 538 network conditions would achieve an average throughput, measured 539 on a reasonable time scale, that is not less than the throughput 540 the RTP flow is achieving. This condition can be satisfied by 541 implementing congestion control mechanisms to adapt the 542 transmission rate (or the number of layers subscribed for a 543 layered multicast session), or by arranging for a receiver to 544 leave the session if the loss rate is unacceptably high. 546 The comparison to TCP cannot be specified exactly, but is intended 547 as an "order-of-magnitude" comparison in time scale and 548 throughput. The time scale on which TCP throughput is measured is 549 the round-trip time of the connection. In essence, this 550 requirement states that it is not acceptable to deploy an 551 application (using RTP or any other transport protocol) on the 552 best-effort Internet which consumes bandwidth arbitrarily and does 553 not compete fairly with TCP within an order of magnitude. 555 The phase "order of magnitude" in the above means within a factor of 556 ten, approximately. In order to implement this, it is necessary to 557 estimate the throughput a bulk TCP connection would achieve over the 558 path. For a long-lived TCP Reno connection, it has been shown that 559 the TCP throughput, X, in bytes per second, can be estimated using 560 [Padhye]: 562 s 563 X = ------------------------------------------------------------- 564 Tr*sqrt(2*b*p/3)+(t_RTO * (3*sqrt(3*b*p/8) * p * (1+32*p*p))) 566 This is the same approach to estimated TCP throughput that is used in 567 [RFC5348]. Under conditions of low packet loss the second term on 568 the denominator is small, so this formula can be approximated with 569 reasonable accuracy as follows [Mathis]: 571 s 572 X = ---------------- 573 Tr*sqrt(2*b*p/3) 575 It is RECOMMENDED that this simplified throughout equation be used, 576 since the reduction in accuracy is small, and it is much simpler to 577 calculate than the full equation. Measurements have shown that the 578 simplified TCP throughput equation is effective as an RTP circuit 579 breaker for multimedia flows sent to hosts on residential networks 580 using ADSL and cable modem links [Singh]. The data shows that the 581 full TCP throughput equation tends to be more sensitive to packet 582 loss and triggers the RTP circuit breaker earlier than the simplified 583 equation. Implementations that desire this extra sensitivity MAY use 584 the full TCP throughput equation in the RTP circuit breaker. Initial 585 measurements in LTE networks have shown that the extra sensitivity is 586 helpful in that environment, with the full TCP throughput equation 587 giving a more balanced circuit breaker response than the simplified 588 TCP equation [Sarker]; other networks might see similar behaviour. 590 No matter what TCP throughput equation is chosen, two parameters need 591 to be estimated and reported to the sender in order to calculate the 592 throughput: the round trip time, Tr, and the loss event rate, p (the 593 packet size, s, is known to the sender). The round trip time can be 594 estimated from RTCP SR and RR packets. This is done too infrequently 595 for accurate statistics, but is the best that can be done with the 596 standard RTCP mechanisms. 598 Report blocks in RTCP SR or RR packets contain the packet loss 599 fraction, rather than the loss event rate, so p cannot be reported 600 (TCP typically treats the loss of multiple packets within a single 601 RTT as one loss event, but RTCP RR packets report the overall 602 fraction of packets lost, and does not report when the packet losses 603 occurred). Using the loss fraction in place of the loss event rate 604 can overestimate the loss. We believe that this overestimate will 605 not be significant, given that we are only interested in order of 606 magnitude comparison ([Floyd] section 3.2.1 shows that the difference 607 is small for steady-state conditions and random loss, but using the 608 loss fraction is more conservative in the case of bursty loss). 610 The congestion circuit breaker is therefore: when a sender that is 611 transmitting at least one RTP packet every max(Tdr, Tr) seconds 612 receives an RTCP SR or RR packet that contains a report block for an 613 SSRC it is using, the sender MUST record the value of the fraction 614 lost field in the report block and the time since the last report 615 block was received for that SSRC. If more than CB_INTERVAL (see 616 below) report blocks have been received for that SSRC, the sender 617 MUST calculate the average fraction lost over the last CB_INTERVAL 618 reporting intervals, and then estimate the TCP throughput that would 619 be achieved over the path using the chosen TCP throughput equation 620 and the measured values of the round-trip time, Tr, the loss event 621 rate, p (approximated by the average fraction lost, as is described 622 below), and the packet size, s. The estimate of the TCP throughput, 623 X, is then compared with the actual sending rate of the RTP stream. 624 If the actual sending rate of the RTP stream is more than 10 * X, 625 then the congestion circuit breaker is triggered. 627 The average fraction lost is calculated based on the sum, over the 628 last CB_INTERVAL reporting intervals, of the fraction lost in each 629 reporting interval multiplied by the duration of the corresponding 630 reporting interval, divided by the total duration of the last 631 CB_INTERVAL reporting intervals. The CB_INTERVAL parameter is set 632 to: 634 CB_INTERVAL = 635 ceil(3*min(max(10*G*Tf, 10*Tr, 3*Tdr), max(15, 3*Td))/(3*Tdr)) 637 The parameters that feed into CB_INTERVAL are chosen to give the 638 congestion control algorithm time to react to congestion. They give 639 at least three RTCP reports, ten round trip times, and ten groups of 640 frames to adjust the rate to reduce the congestion to a reasonable 641 level. It is expected that a responsive congestion control algorithm 642 will begin to respond with the next group of frames after it receives 643 indication of congestion, so CB_INTERVAL ought to be a much longer 644 interval than the congestion response. 646 If the RTP/AVPF profile [RFC4585] or the RTP/SAVPF [RFC5124] is used, 647 and the T_rr_interval parameter is used to reduce the frequency of 648 regular RTCP reports, then the value Tdr in the above expression for 649 the CB_INTERVAL parameter MUST be replaced by max(T_rr_interval, 650 Tdr). 652 The CB_INTERVAL parameter is calculated on joining the session, and 653 recalculated on receipt of each RTCP packet, after checking whether 654 the media timeout circuit breaker or the congestion circuit breaker 655 has been triggered. 657 To ensure a timely response to persistent congestion, implementations 658 SHOULD NOT configure the RTCP bandwidth such that Tdr is larger than 659 5 seconds. Similarly, implementations that use the RTP/AVPF profile 660 [RFC4585] or the RTP/SAVPF profile [RFC5124] SHOULD NOT configure 661 T_rr_interval to values larger than 4 seconds (the reduced limit for 662 T_rr_interval follows Section 6.1.3 of 663 [I-D.ietf-avtcore-rtp-multi-stream]). 665 The rationale for enforcing a minimum sending rate below which the 666 congestion circuit breaker will not trigger is to avoid spurious 667 circuit breaker triggers when the number of packets sent per RTCP 668 reporting interval is small, and hence the fraction lost samples are 669 subject to measurement artefacts. The bound of at least one packet 670 every max(Tdr, Tr) seconds is derived from the one packet per RTT 671 minimum sending rate of TCP [RFC5405], adapted for use with RTP where 672 the RTCP reporting interval is decoupled from the network RTT. 674 When the congestion circuit breaker is triggered, the sender SHOULD 675 cease transmission (see Section 4.5). However, if the sender is able 676 to reduce its sending rate by a factor of (approximately) ten, then 677 it MAY first reduce its sending rate by this factor (or some larger 678 amount) to see if that resolves the congestion. If the sending rate 679 is reduced in this way and the congestion circuit breaker triggers 680 again after the next CB_INTERVAL RTCP reporting intervals, the sender 681 MUST then cease transmission. An example of such a rate reduction 682 might be a video conferencing system that backs off to sending audio 683 only, before completely dropping the call. If such a reduction in 684 sending rate resolves the congestion problem, the sender MAY 685 gradually increase the rate at which it sends data after a reasonable 686 amount of time has passed, provided it takes care not to cause the 687 problem to recur ("reasonable" is intentionally not defined here). 689 The RTCP reporting interval of the media sender does not affect how 690 quickly congestion circuit breaker can trigger. The timing is based 691 on the RTCP reporting interval of the receiver that generates the SR/ 692 RR packets from which the loss rate and RTT estimate are derived 693 (note that RTCP requires all participants in a session to have 694 similar reporting intervals, else the participant timeout rules in 695 [RFC3550] will not work, so this interval is likely similar to that 696 of the sender). If the incoming RTCP SR or RR packets are using a 697 reduced minimum RTCP reporting interval (as specified in Section 6.2 698 of RFC 3550 [RFC3550] or the RTP/AVPF profile [RFC4585]), then that 699 reduced RTCP reporting interval is used when determining if the 700 circuit breaker is triggered. 702 If there are more media streams that can be reported in a single RTCP 703 SR or RR packet, or if the size of a complete RTCP SR or RR packet 704 exceeds the network MTU, then the receiver will report on a subset of 705 sources in each reporting interval, with the subsets selected round- 706 robin across multiple intervals so that all sources are eventually 707 reported [RFC3550]. When generating such round-robin RTCP reports, 708 priority SHOULD be given to reports on sources that have high packet 709 loss rates, to ensure that senders are aware of network congestion 710 they are causing (this is an update to [RFC3550]). 712 4.4. RTP/AVP Circuit Breaker #4: Media Usability 714 Applications that use RTP are generally tolerant to some amount of 715 packet loss. How much packet loss can be tolerated will depend on 716 the application, media codec, and the amount of error correction and 717 packet loss concealment that is applied. There is an upper bound on 718 the amount of loss that can be corrected, however, beyond which the 719 media becomes unusable. Similarly, many applications have some upper 720 bound on the media capture to play-out latency that can be tolerated 721 before the application becomes unusable. The latency bound will 722 depend on the application, but typical values can range from the 723 order of a few hundred milliseconds for voice telephony and 724 interactive conferencing applications, up to several seconds for some 725 video-on-demand systems. 727 As a final circuit breaker, RTP senders SHOULD monitor the reported 728 packet loss and delay to estimate whether the media is likely to be 729 suitable for the intended purpose. If the packet loss rate and/or 730 latency is such that the media has become unusable, and has remained 731 unusable for a significant time period, then the application SHOULD 732 cease transmission. Similarly, receivers SHOULD monitor the quality 733 of the media they receive, and if the quality is unusable for a 734 significant time period, they SHOULD terminate the session. This 735 memo intentionally does not define a bound on the packet loss rate or 736 latency that will result in unusable media, nor does it specify what 737 time period is deemed significant, as these are highly application 738 dependent. 740 Sending media that suffers from such high packet loss or latency that 741 it is unusable at the receiver is both wasteful of resources, and of 742 no benefit to the user of the application. It also is highly likely 743 to be congesting the network, and disrupting other applications. As 744 such, the congestion circuit breaker will almost certainly trigger to 745 stop flows where the media would be unusable due to high packet loss 746 or latency. However, in pathological scenarios where the congestion 747 circuit breaker does not stop the flow, it is desirable to prevent 748 the application sending unnecessary traffic that might disrupt other 749 uses of the network. The role of the media usability circuit breaker 750 is to protect the network in such cases. 752 4.5. Ceasing Transmission 754 What it means to cease transmission depends on the application. The 755 intention is that the application will stop sending RTP data packets 756 to a particular destination 3-tuple (transport protocol, destination 757 port, IP address), until the user makes an explicit attempt to 758 restart the call. It is important that a human user is involved in 759 the decision to try to restart the call, since that user will 760 eventually give up if the calls repeatedly trigger the circuit 761 breaker. This will help avoid problems with automatic redial systems 762 from congesting the network. Accordingly, RTP flows halted by the 763 circuit breaker SHOULD NOT be restarted automatically unless the 764 sender has received information that the congestion has dissipated, 765 or can reasonably be expected to have dissipated. This is 766 necessarily application dependent, but could be, for example, an 767 indication that a competing flow has finished, freeing up some 768 capacity, or that the device has moved so the flow would traverse a 769 different path if restarted. 771 It is recognised that the RTP implementation in some systems might 772 not be able to determine if a call set-up request was initiated by a 773 human user, or automatically by some scripted higher-level component 774 of the system. These implementations MUST rate limit attempts to 775 restart a call to the same destination 3-tuple as used by a call that 776 triggered the circuit breaker, so that the reaction to a triggered 777 circuit breaker lasts for at least the triggering interval 778 [I-D.ietf-tsvwg-circuit-breaker]. The chosen rate limit ought to not 779 exceed the rate at which an annoyed human caller might redial a 780 misbehaving phone. 782 5. RTP Circuit Breakers and the RTP/AVPF and RTP/SAVPF Profiles 784 Use of the Extended RTP Profile for RTCP-based Feedback (RTP/AVPF) 785 [RFC4585] allows receivers to send early RTCP reports in some cases, 786 to inform the sender about particular events in the media stream. 787 There are several use cases for such early RTCP reports, including 788 providing rapid feedback to a sender about the onset of congestion. 789 The RTP/SAVPF Profile [RFC5124] is a secure variant of the RTP/AVPF 790 profile, that is treated the same in the context of the RTP circuit 791 breaker. These feedback profiles are often used with non-compound 792 RTCP reports [RFC5506] to reduce the reporting overhead. 794 Receiving rapid feedback about congestion events potentially allows 795 congestion control algorithms to be more responsive, and to better 796 adapt the media transmission to the limitations of the network. It 797 is expected that many RTP congestion control algorithms will adopt 798 the RTP/AVPF profile or the RTP/SAVPF profile for this reason, 799 defining new transport layer feedback reports that suit their 800 requirements. Since these reports are not yet defined, and likely 801 very specific to the details of the congestion control algorithm 802 chosen, they cannot be used as part of the generic RTP circuit 803 breaker. 805 Reduced-size RTCP reports sent under the RTP/AVPF early feedback 806 rules that do not contain an RTCP SR or RR packet MUST be ignored by 807 the congestion circuit breaker (they do not contain the information 808 needed by the congestion circuit breaker algorithm), but MUST be 809 counted as received packets for the RTCP timeout circuit breaker. 810 Reduced-size RTCP reports sent under the RTP/AVPF early feedback 811 rules that contain RTCP SR or RR packets MUST be processed by the 812 congestion circuit breaker as if they were sent as regular RTCP 813 reports, and counted towards the circuit breaker conditions specified 814 in Section 4 of this memo. This will potentially make the RTP 815 circuit breaker trigger earlier than it would if the RTP/AVPF profile 816 was not used. 818 When using ECN with RTP (see Section 7), early RTCP feedback packets 819 can contain ECN feedback reports. The count of ECN-CE marked packets 820 contained in those ECN feedback reports is counted towards the number 821 of lost packets reported if the ECN Feedback Report is sent in a 822 compound RTCP packet along with an RTCP SR/RR report packet. Reports 823 of ECN-CE packets sent as reduced-size RTCP ECN feedback packets 824 without an RTCP SR/RR packet MUST be ignored. 826 These rules are intended to allow the use of low-overhead RTP/AVPF 827 feedback for generic NACK messages without triggering the RTP circuit 828 breaker. This is expected to make such feedback suitable for RTP 829 congestion control algorithms that need to quickly report loss events 830 in between regular RTCP reports. The reaction to reduced-size RTCP 831 SR/RR packets is to allow such algorithms to send feedback that can 832 trigger the circuit breaker, when desired. 834 The RTP/AVPF and RTP/SAVPF profiles include the T_rr_interval 835 parameter that can be used to adjust the regular RTCP reporting 836 interval. The use of the T_rr_interval parameter changes the 837 behaviour of the RTP circuit breaker, as described in Section 4. 839 6. Impact of RTCP Extended Reports (XR) 841 RTCP Extended Report (XR) blocks provide additional reception quality 842 metrics, but do not change the RTCP timing rules. Some of the RTCP 843 XR blocks provide information that might be useful for congestion 844 control purposes, others provide non-congestion-related metrics. 845 With the exception of RTCP XR ECN Summary Reports (see Section 7), 846 the presence of RTCP XR blocks in a compound RTCP packet does not 847 affect the RTP circuit breaker algorithm. For consistency and ease 848 of implementation, only the reception report blocks contained in RTCP 849 SR packets, RTCP RR packets, or RTCP XR ECN Summary Report packets, 850 are used by the RTP circuit breaker algorithm. 852 7. Impact of Explicit Congestion Notification (ECN) 854 The use of ECN for RTP flows does not affect the media timeout RTP 855 circuit breaker (Section 4.2) or the RTCP timeout circuit breaker 856 (Section 4.1), since these are both connectivity checks that simply 857 determinate if any packets are being received. 859 ECN-CE marked packets SHOULD be treated as if it were lost for the 860 purposes of congestion control, when determining the optimal media 861 sending rate for an RTP flow. If an RTP sender has negotiated ECN 862 support for an RTP session, and has successfully initiated ECN use on 863 the path to the receiver [RFC6679], then ECN-CE marked packets SHOULD 864 be treated as if they were lost when calculating if the congestion- 865 based RTP circuit breaker (Section 4.3) has been met. The count of 866 ECN-CE marked RTP packets is returned in RTCP XR ECN summary report 867 packets if support for ECN has been initiated for an RTP session. 869 8. Impact of Bundled Media and Layered Coding 871 The RTP circuit breaker operates on a per-RTP session basis. An RTP 872 sender that participates in several RTP sessions MUST treat each RTP 873 session independently with regards to the RTP circuit breaker. 875 An RTP sender can generate several media streams within a single RTP 876 session, with each stream using a different SSRC. This can happen if 877 bundled media are in use, when using simulcast, or when using layered 878 media coding. By default, each SSRC will be treated independently by 879 the RTP circuit breaker. However, the sender MAY choose to treat the 880 flows (or a subset thereof) as a group, such that a circuit breaker 881 trigger for one flow applies to the group of flows as a whole, and 882 either causes the entire group to cease transmission, or the sending 883 rate of the group to reduce by a factor of ten, depending on the RTP 884 circuit breaker triggered. Grouping flows in this way is expected to 885 be especially useful for layered flows sent using multiple SSRCs, as 886 it allows the layered flow to react as a whole, ceasing transmission 887 on the enhancement layers first to reduce sending rate if necessary, 888 rather than treating each layer independently. 890 9. Security Considerations 892 The security considerations of [RFC3550] apply. 894 If the RTP/AVPF profile is used to provide rapid RTCP feedback, the 895 security considerations of [RFC4585] apply. If ECN feedback for RTP 896 over UDP/IP is used, the security considerations of [RFC6679] apply. 898 If non-authenticated RTCP reports are used, an on-path attacker can 899 trivially generate fake RTCP packets that indicate high packet loss 900 rates, causing the circuit breaker to trigger and disrupt an RTP 901 session. This is somewhat more difficult for an off-path attacker, 902 due to the need to guess the randomly chosen RTP SSRC value and the 903 RTP sequence number. This attack can be avoided if RTCP packets are 904 authenticated; authentication options are discussed in [RFC7201]. 906 Timely operation of the RTP circuit breaker depends on the choice of 907 RTCP reporting interval. If the receiver has a reporting interval 908 that is overly long, then the responsiveness of the circuit breaker 909 decreases. In the limit, the RTP circuit breaker can be disabled for 910 all practical purposes by configuring an RTCP reporting interval that 911 is many minutes duration. This issue is not specific to the circuit 912 breaker: long RTCP reporting intervals also prevent reception quality 913 reports, feedback messages, codec control messages, etc., from being 914 used. Implementations are expected to impose an upper limit on the 915 RTCP reporting interval they are willing to negotiate (based on the 916 session bandwidth and RTCP bandwidth fraction) when using the RTP 917 circuit breaker, as discussed in Section 4.3. 919 10. IANA Considerations 921 There are no actions for IANA. 923 11. Acknowledgements 925 The authors would like to thank Bernard Aboba, Harald Alvestrand, 926 Gorry Fairhurst, Nazila Fough, Kevin Gross, Cullen Jennings, Randell 927 Jesup, Jonathan Lennox, Matt Mathis, Stephen McQuistin, Simon 928 Perreault, Eric Rescorla, Abheek Saha, Meral Shirazipour, Fabio 929 Verdicchio, and Magnus Westerlund for their valuable feedback. 931 12. References 933 12.1. Normative References 935 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 936 Requirement Levels", BCP 14, RFC 2119, 937 DOI 10.17487/RFC2119, March 1997, 938 . 940 [RFC3550] Schulzrinne, H., Casner, S., Frederick, R., and V. 941 Jacobson, "RTP: A Transport Protocol for Real-Time 942 Applications", STD 64, RFC 3550, DOI 10.17487/RFC3550, 943 July 2003, . 945 [RFC3551] Schulzrinne, H. and S. Casner, "RTP Profile for Audio and 946 Video Conferences with Minimal Control", STD 65, RFC 3551, 947 DOI 10.17487/RFC3551, July 2003, 948 . 950 [RFC3611] Friedman, T., Ed., Caceres, R., Ed., and A. Clark, Ed., 951 "RTP Control Protocol Extended Reports (RTCP XR)", 952 RFC 3611, DOI 10.17487/RFC3611, November 2003, 953 . 955 [RFC4585] Ott, J., Wenger, S., Sato, N., Burmeister, C., and J. Rey, 956 "Extended RTP Profile for Real-time Transport Control 957 Protocol (RTCP)-Based Feedback (RTP/AVPF)", RFC 4585, 958 DOI 10.17487/RFC4585, July 2006, 959 . 961 [RFC5348] Floyd, S., Handley, M., Padhye, J., and J. Widmer, "TCP 962 Friendly Rate Control (TFRC): Protocol Specification", 963 RFC 5348, DOI 10.17487/RFC5348, September 2008, 964 . 966 12.2. Informative References 968 [Floyd] Floyd, S., Handley, M., Padhye, J., and J. Widmer, 969 "Equation-Based Congestion Control for Unicast 970 Applications", Proceedings of the ACM SIGCOMM 971 conference, 2000, DOI 10.1145/347059.347397, August 2000. 973 [I-D.ietf-avtcore-rtp-multi-stream] 974 Lennox, J., Westerlund, M., Wu, Q., and C. Perkins, 975 "Sending Multiple RTP Streams in a Single RTP Session", 976 draft-ietf-avtcore-rtp-multi-stream-11 (work in progress), 977 December 2015. 979 [I-D.ietf-tsvwg-circuit-breaker] 980 Fairhurst, G., "Network Transport Circuit Breakers", 981 draft-ietf-tsvwg-circuit-breaker-11 (work in progress), 982 December 2015. 984 [Mathis] Mathis, M., Semke, J., Mahdavi, J., and T. Ott, "The 985 macroscopic behavior of the TCP congestion avoidance 986 algorithm", ACM SIGCOMM Computer Communication 987 Review 27(3), DOI 10.1145/263932.264023, July 1997. 989 [Padhye] Padhye, J., Firoiu, V., Towsley, D., and J. Kurose, 990 "Modeling TCP Throughput: A Simple Model and its Empirical 991 Validation", Proceedings of the ACM SIGCOMM 992 conference, 1998, DOI 10.1145/285237.285291, August 1998. 994 [RFC2862] Civanlar, M. and G. Cash, "RTP Payload Format for Real- 995 Time Pointers", RFC 2862, DOI 10.17487/RFC2862, June 2000, 996 . 998 [RFC3168] Ramakrishnan, K., Floyd, S., and D. Black, "The Addition 999 of Explicit Congestion Notification (ECN) to IP", 1000 RFC 3168, DOI 10.17487/RFC3168, September 2001, 1001 . 1003 [RFC4103] Hellstrom, G. and P. Jones, "RTP Payload for Text 1004 Conversation", RFC 4103, DOI 10.17487/RFC4103, June 2005, 1005 . 1007 [RFC5104] Wenger, S., Chandra, U., Westerlund, M., and B. Burman, 1008 "Codec Control Messages in the RTP Audio-Visual Profile 1009 with Feedback (AVPF)", RFC 5104, DOI 10.17487/RFC5104, 1010 February 2008, . 1012 [RFC5124] Ott, J. and E. Carrara, "Extended Secure RTP Profile for 1013 Real-time Transport Control Protocol (RTCP)-Based Feedback 1014 (RTP/SAVPF)", RFC 5124, DOI 10.17487/RFC5124, February 1015 2008, . 1017 [RFC5405] Eggert, L. and G. Fairhurst, "Unicast UDP Usage Guidelines 1018 for Application Designers", BCP 145, RFC 5405, 1019 DOI 10.17487/RFC5405, November 2008, 1020 . 1022 [RFC5450] Singer, D. and H. Desineni, "Transmission Time Offsets in 1023 RTP Streams", RFC 5450, DOI 10.17487/RFC5450, March 2009, 1024 . 1026 [RFC5506] Johansson, I. and M. Westerlund, "Support for Reduced-Size 1027 Real-Time Transport Control Protocol (RTCP): Opportunities 1028 and Consequences", RFC 5506, DOI 10.17487/RFC5506, April 1029 2009, . 1031 [RFC5681] Allman, M., Paxson, V., and E. Blanton, "TCP Congestion 1032 Control", RFC 5681, DOI 10.17487/RFC5681, September 2009, 1033 . 1035 [RFC6679] Westerlund, M., Johansson, I., Perkins, C., O'Hanlon, P., 1036 and K. Carlberg, "Explicit Congestion Notification (ECN) 1037 for RTP over UDP", RFC 6679, DOI 10.17487/RFC6679, August 1038 2012, . 1040 [RFC6798] Clark, A. and Q. Wu, "RTP Control Protocol (RTCP) Extended 1041 Report (XR) Block for Packet Delay Variation Metric 1042 Reporting", RFC 6798, DOI 10.17487/RFC6798, November 2012, 1043 . 1045 [RFC6843] Clark, A., Gross, K., and Q. Wu, "RTP Control Protocol 1046 (RTCP) Extended Report (XR) Block for Delay Metric 1047 Reporting", RFC 6843, DOI 10.17487/RFC6843, January 2013, 1048 . 1050 [RFC6958] Clark, A., Zhang, S., Zhao, J., and Q. Wu, Ed., "RTP 1051 Control Protocol (RTCP) Extended Report (XR) Block for 1052 Burst/Gap Loss Metric Reporting", RFC 6958, 1053 DOI 10.17487/RFC6958, May 2013, 1054 . 1056 [RFC7002] Clark, A., Zorn, G., and Q. Wu, "RTP Control Protocol 1057 (RTCP) Extended Report (XR) Block for Discard Count Metric 1058 Reporting", RFC 7002, DOI 10.17487/RFC7002, September 1059 2013, . 1061 [RFC7003] Clark, A., Huang, R., and Q. Wu, Ed., "RTP Control 1062 Protocol (RTCP) Extended Report (XR) Block for Burst/Gap 1063 Discard Metric Reporting", RFC 7003, DOI 10.17487/RFC7003, 1064 September 2013, . 1066 [RFC7097] Ott, J., Singh, V., Ed., and I. Curcio, "RTP Control 1067 Protocol (RTCP) Extended Report (XR) for RLE of Discarded 1068 Packets", RFC 7097, DOI 10.17487/RFC7097, January 2014, 1069 . 1071 [RFC7201] Westerlund, M. and C. Perkins, "Options for Securing RTP 1072 Sessions", RFC 7201, DOI 10.17487/RFC7201, April 2014, 1073 . 1075 [Sarker] Sarker, Z., Singh, V., and C. Perkins, "An Evaluation of 1076 RTP Circuit Breaker Performance on LTE Networks", 1077 Proceedings of the IEEE Infocom workshop on Communication 1078 and Networking Techniques for Contemporary Video, 2014, 1079 April 2014. 1081 [Singh] Singh, V., McQuistin, S., Ellis, M., and C. Perkins, 1082 "Circuit Breakers for Multimedia Congestion Control", 1083 Proceedings of the International Packet Video 1084 Workshop, 2013, DOI 10.1109/PV.2013.6691439, December 1085 2013. 1087 Authors' Addresses 1089 Colin Perkins 1090 University of Glasgow 1091 School of Computing Science 1092 Glasgow G12 8QQ 1093 United Kingdom 1095 Email: csp@csperkins.org 1096 Varun Singh 1097 Nemu Dialogue Systems Oy 1098 Runeberginkatu 4c A 4 1099 Helsinki 00100 1100 Finland 1102 Email: varun.singh@iki.fi 1103 URI: http://www.callstats.io/