idnits 2.17.1 draft-ietf-rmcat-cc-requirements-01.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 abstract seems to contain references ([I-D.jesup-rtp-congestion-reqs]), which it shouldn't. Please replace those with straight textual mentions of the documents in question. Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year -- The document date (December 21, 2013) is 3771 days in the past. Is this intentional? Checking references for intended status: Informational ---------------------------------------------------------------------------- == Unused Reference: 'RFC5506' is defined on line 371, but no explicit reference was found in the text == Outdated reference: A later version (-19) exists of draft-ietf-rtcweb-overview-08 == Outdated reference: A later version (-05) exists of draft-welzl-rmcat-coupled-cc-02 Summary: 1 error (**), 0 flaws (~~), 4 warnings (==), 1 comment (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 Network Working Group R. Jesup 3 Internet-Draft Mozilla 4 Intended status: Informational December 21, 2013 5 Expires: June 24, 2014 7 Congestion Control Requirements For RMCAT 8 draft-ietf-rmcat-cc-requirements-01 10 Abstract 12 Congestion control is needed for all data transported across the 13 Internet, in order to promote fair usage and prevent congestion 14 collapse. The requirements for interactive, point-to-point real time 15 multimedia, which needs low-delay, semi-reliable data delivery, are 16 different from the requirements for bulk transfer like FTP or bursty 17 transfers like Web pages. 19 This document attempts to describe a set of requirements that can be 20 used to evaluate other congestion control mechanisms in order to 21 figure out their fitness for this purpose, and in particular to 22 provide a set of possible requirements for proposals coming out of 23 the RMCAT Working Group. 25 This document is derived from draft-jesup-rtp-congestion-reqs 26 [I-D.jesup-rtp-congestion-reqs]. 28 Requirements Language 30 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 31 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 32 document are to be interpreted as described in RFC 2119 [RFC2119]. 34 Status of This Memo 36 This Internet-Draft is submitted in full conformance with the 37 provisions of BCP 78 and BCP 79. 39 Internet-Drafts are working documents of the Internet Engineering 40 Task Force (IETF). Note that other groups may also distribute 41 working documents as Internet-Drafts. The list of current Internet- 42 Drafts is at http://datatracker.ietf.org/drafts/current/. 44 Internet-Drafts are draft documents valid for a maximum of six months 45 and may be updated, replaced, or obsoleted by other documents at any 46 time. It is inappropriate to use Internet-Drafts as reference 47 material or to cite them other than as "work in progress." 48 This Internet-Draft will expire on June 24, 2014. 50 Copyright Notice 52 Copyright (c) 2013 IETF Trust and the persons identified as the 53 document authors. All rights reserved. 55 This document is subject to BCP 78 and the IETF Trust's Legal 56 Provisions Relating to IETF Documents 57 (http://trustee.ietf.org/license-info) in effect on the date of 58 publication of this document. Please review these documents 59 carefully, as they describe your rights and restrictions with respect 60 to this document. Code Components extracted from this document must 61 include Simplified BSD License text as described in Section 4.e of 62 the Trust Legal Provisions and are provided without warranty as 63 described in the Simplified BSD License. 65 Table of Contents 67 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 2 68 2. Requirements . . . . . . . . . . . . . . . . . . . . . . . . 3 69 3. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 7 70 4. Security Considerations . . . . . . . . . . . . . . . . . . . 7 71 5. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 7 72 6. References . . . . . . . . . . . . . . . . . . . . . . . . . 8 73 6.1. Normative References . . . . . . . . . . . . . . . . . . 8 74 6.2. Informative References . . . . . . . . . . . . . . . . . 8 75 Author's Address . . . . . . . . . . . . . . . . . . . . . . . . 8 77 1. Introduction 79 The traditional TCP congestion control requirements were developed in 80 order to promote efficient use of the Internet for reliable bulk 81 transfer of non-time-critical data, such as transfer of large files. 82 They have also been used successfully to govern the reliable transfer 83 of smaller chunks of data in as short a time as possible, such as 84 when fetching Web pages. 86 These algorithms have also been used for transfer of media streams 87 that are viewed in a non-interactive manner, such as "streaming" 88 video, where having the data ready when the viewer wants it is 89 important, but the exact timing of the delivery is not. 91 When doing real time interactive media, the requirements are 92 different; one needs to provide the data continuously, within a very 93 limited time window (no more than 100s of milliseconds end-to-end 94 delay), the sources of data may be able to adapt the amount of data 95 that needs sending within fairly wide margins, and may tolerate some 96 amount of packet loss, but since the data is generated in real time, 97 sending "future" data is impossible, and since it's consumed in real 98 time, data delivered late is useless. 100 While the requirements for RMCAT differ from the requirements for the 101 other flow types, these other flow types will be present in the 102 network. The RMCAT congestion control algorithm must work properly 103 when these other flow types are present as cross traffic on the 104 network. 106 One particular protocol portofolio being developed for this use case 107 is WebRTC [I-D.ietf-rtcweb-overview], where one envisions sending 108 multiple RTP-based flows between two peers, in conjunction with data 109 flows, all at the same time, without having special arrangements with 110 the intervening service providers. 112 Given that this use case is the focus of this document, use cases 113 involving noninteractive media such as YouTube-like video streaming, 114 and use cases using multicast/broadcast-type technologies, are out of 115 scope. 117 The terminology defined in [I-D.ietf-rtcweb-overview] is used in this 118 memo. 120 2. Requirements 122 1. The congestion control algorithm must attempt to provide as-low- 123 as-possible-delay transit for real-time traffic while still 124 providing a useful amount of bandwidth, even when faced with 125 intermediate bottlenecks and competing flows. There may be 126 lower limits on the amount of bandwidth that is useful, but this 127 is largely application-specific and the application may be able 128 to modify or remove flows in order allow some useful flows to 129 get enough bandwidth. (Example: not enough bandwidth for low- 130 latency video+audio, but enough for audio-only.) 132 A. It should also handle routing changes and interface changes 133 (WiFi to 3G data, etc) which may radically change the 134 bandwidth available, and react quickly, especially if there 135 is a reduction in available bandwidth. 137 B. The offered load may be less than the available bandwidth at 138 any given moment, and may vary dramatically over time, 139 including dropping to no load and then resuming a high load, 140 such as in a mute operation. The reaction time between a 141 change in the bandwidth available from the algorithm and a 142 change in the offered load is variable, and may be different 143 when increasing versus decreasing. 145 C. The algorithm must not overreact to short-term bursts (such 146 as web-browsing) which can quickly saturate a local- 147 bottleneck router or link, but also clear quickly, and 148 should recover quickly when the burst ends. This is 149 inherently at odds with the need to react quickly-enough to 150 avoid queue buildup. 152 D. Similarly periodic bursty flows such as DASH or proprietary 153 media streaming algorithms may compete in bursts with the 154 algorithm, and may not be adaptive within a burst. They are 155 often are layered on top of TCP. The algorithm must avoid 156 too much delay buildup during those bursts, and quickly 157 recover. Note that this traffic may on an access link, or 158 may cause a shift in the location of the bottleneck fir the 159 duration of the burst. 161 2. The algorithm must be fair to other flows, both realtime flows 162 (such as other instances of itself), and TCP flows, both long- 163 lived and bursts such as the traffic generated by a typical web 164 browsing session. Note that 'fair' is a rather hard-to-define 165 term. 167 3. The algorithm should where possible merge information across 168 multiple RTP streams between the same endpoints, whether or not 169 they're multiplexed on the same ports, in order to allow 170 congestion control of the set of streams together instead of as 171 multiple independent streams. This allows better overall 172 bandwidth management, faster response to changing conditions, 173 and fairer sharing of bandwidth with other network users. 174 Alternatively, it should work with an external bandwidth control 175 framework to coordinate bandwidth usage across a bottleneck, 176 such as draft-welzl-rmcat-coupled-cc 177 [I-D.welzl-rmcat-coupled-cc]. 179 A. If possible, it should also share information and adaptation 180 with other non-RTP flows between the same endpoints, such as 181 a WebRTC data channel 183 B. The most correlated bandwidth usage would be with other 184 flows on the same 5-tuple, but there may be use in 185 coordinating measurement and control of the local link(s). 187 C. Use of information about previous flows, especially on the 188 same 5-tuple, may be useful input to the algorithm, 189 especially to startup performance of a new flow. 191 4. The algorithm should not require any special support from 192 network elements (ECN, etc). As much as possible, it should 193 leverage available information about the incoming flow to 194 provide feedback to the sender. Examples of this information 195 are the ECN, packet arrival times, acknowledgments and feedback, 196 packet timestamps, and packet losses; all of these can provide 197 information about the state of the path and any bottlenecks. 199 A. Extra information could be added to the packets to provide 200 more detailed information on actual send times (as opposed 201 to sampling times), but should not be required. 203 B. When additional input signals such as ECN are available, 204 they should be utilized if possible. 206 5. Since the assumption here is a set of RTP streams, the 207 backchannel typically should be done via RTCP; one alternative 208 would be to include it instead in a reverse RTP channel using 209 header extensions. 211 A. In order to react sufficiently quickly when using RTCP for a 212 backchannel, an RTP profile such as AVPF/SAVPF that allows 213 sufficiently frequent feedback [RFC4585] MUST be used. 215 B. Note that in some cases, backchannel messages may be delayed 216 until the RTCP channel can be allocated enough bandwidth, 217 even under AVPF rules. This may also imply negotiating a 218 higher maximum percentage for RTCP data or allowing RMCAT 219 solutions to violate or modify the rules specified for AVPF. 221 C. Bandwidth for the feedback messages should be minimized 222 (such as via RFC 5506 [RFC5506]to allow RTCP without SR/RR) 224 D. Header extensions would avoid the RTCP timing rules issues, 225 and allow the application to allocate bandwidth as needed 226 for the congestion algorithm. 228 E. Backchannel data should be minimized to avoid taking too 229 much reverse-channel bandwidth (since this will often be 230 used in a bidirectional set of flows). In areas of 231 stability, backchannel data may be sent more infrequently so 232 long as algorithm stability and fairness are maintained. 233 When the channel is unstable or has not yet reached 234 equilibrium after a change, backchannel feedback may be more 235 frequent and use more reverse-channel bandwidth. This is an 236 area with considerable flexibility of design, and different 237 approaches to backchannel messages and frequency are 238 expected to be evaluated. 240 6. Flows managed by this algorithm and flows competed against at a 241 bottleneck may have different DSCP markings depending on the 242 type of traffic. A particular bottleneck or section of the 243 network path may or may not honor these markings. 245 A. In WebRTC, a division of packets into 4 classes is 246 envisioned in order of priority: faster-than-audio, audio, 247 video, best-effort, and bulk-transfer. Typically the flows 248 managed by this algorithm would be audio or video in that 249 heirarchy, and feedback flows would be faster-than-audio. 251 7. The algorithm should sense the unexpected lack of backchannel 252 information as a possible indication of a channel overuse 253 problem and react accordingly to avoid burst events causing a 254 congestion collapse. 256 8. It should attempt to avoid bandwidth 'collapse' when facing a 257 long-lived saturating TCP flow or flows. (I.e. a classic delay- 258 sensitive algorithm will reduce bandwidth to keep delay down 259 until the TCP flow has all the bandwidth). See the Cx-TCP 260 algorithm discussed in a recent Transactions On Networking 261 [cx-tcp] for an example of a delay-sensitive congestion-control 262 algorithm that transitions to a loss-based mode when competing 263 with TCP flows - at the cost of increased delay. 265 9. The algorithm should be stable and low-delay when faced with 266 active queue management (AQM) algorithms. Also note that these 267 algorithms may apply across multiple queues in the bottleneck, 268 or to a single queue 270 10. The algorithm should quickly adapt to initial network conditions 271 at the start of a flow. This should occur both if the initial 272 bandwidth is above or below the bottleneck bandwidth. 274 A. The startup adaptation may be faster than adaptation later 275 in a flow. It should allow for both slow-start operation 276 (adapt up) and history-based startup (start at a point 277 expected to be at or below channel bandwidth from historical 278 information, which may need to adapt down quickly if the 279 initial guess is wrong). Starting too low and/or adapting 280 up too slowly can cause a critical point in a personal 281 communication to be poor ("Hello!"). Starting over- 282 bandwidth causes other problems for user experience, so 283 there's a tension here. 285 B. Alternative methods to help startup like probing during 286 setup with dummy data may be useful in some applications; in 287 some cases there will be a considerable gap in time between 288 flow creation and the initial flow of data. 290 C. A flow may need to change adaptation rates due to network 291 conditions or changes in the provided flows (such as un- 292 muting or sending data after a gap). 294 11. It should be evaluated in how it works both with backbone-router 295 bottlenecks, (asymmetric) local-loop bottlenecks, and local-lan 296 (WiFi/etc) bottlenecks, and in competition with varying numbers 297 and types of streams (TCP, TCP variants in use, LEDBAT 298 [I-D.ietf-ledbat-congestion], inflexible VoIP UDP flows). 300 12. It should be stable if the RTP streams are halted or 301 discontinuous (VAD/DTX). 303 A. After a resumption of RTP data it may adapt more quickly 304 (similar to the start of a flow), and previous bandwidth 305 estimates may need to be aged or thrown away. 307 3. IANA Considerations 309 This document makes no request of IANA. 311 Note to RFC Editor: this section may be removed on publication as an 312 RFC. 314 4. Security Considerations 316 An attacker with the ability to delete, delay or insert messages in 317 the flow can fake congestion signals, unless they are passed on a 318 tamper-proof path. Since some possible algorithms depend on the 319 timing of packet arrival, even a traditional protected channel does 320 not fully mitigate such attacks. 322 An attack that reduces bandwidth is not necessarily significant, 323 since an on-path attacker could break the connection by discarding 324 all packets. Attacks that increase the percieved available bandwidth 325 are concievable, and need to be evaluated. 327 Algorithm designers SHOULD consider the possibility of malicious on- 328 path attackers. 330 5. Acknowledgements 332 This document is the result of discussions in various fora of the 333 WebRTC effort, in particular on the rtp-congestion@alvestrand.no 334 mailing list. Many people contributed their thoughts to this. 336 6. References 338 6.1. Normative References 340 [I-D.ietf-rtcweb-overview] 341 Alvestrand, H., "Overview: Real Time Protocols for Brower- 342 based Applications", draft-ietf-rtcweb-overview-08 (work 343 in progress), September 2013. 345 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 346 Requirement Levels", BCP 14, RFC 2119, March 1997. 348 [RFC4585] Ott, J., Wenger, S., Sato, N., Burmeister, C., and J. Rey, 349 "Extended RTP Profile for Real-time Transport Control 350 Protocol (RTCP)-Based Feedback (RTP/AVPF)", RFC 4585, July 351 2006. 353 6.2. Informative References 355 [I-D.ietf-ledbat-congestion] 356 Shalunov, S., Hazel, G., Iyengar, J., and M. Kuehlewind, 357 "Low Extra Delay Background Transport (LEDBAT)", draft- 358 ietf-ledbat-congestion-10 (work in progress), September 359 2012. 361 [I-D.jesup-rtp-congestion-reqs] 362 Jesup, R. and H. Alvestrand, "Congestion Control 363 Requirements For Real Time Media", draft-jesup-rtp- 364 congestion-reqs-00 (work in progress), March 2012. 366 [I-D.welzl-rmcat-coupled-cc] 367 Welzl, M., Islam, S., and S. Gjessing, "Coupled congestion 368 control for RTP media", draft-welzl-rmcat-coupled-cc-02 369 (work in progress), October 2013. 371 [RFC5506] Johansson, I. and M. Westerlund, "Support for Reduced-Size 372 Real-Time Transport Control Protocol (RTCP): Opportunities 373 and Consequences", RFC 5506, April 2009. 375 [cx-tcp] Budzisz, L., Stanojevic, R., Schlote, A., Baker, F., and 376 R. Shorten, "On the Fair Coexistence of Loss- and Delay- 377 Based TCP", December 2011. 379 Author's Address 380 Randell Jesup 381 Mozilla 382 USA 384 Email: randell-ietf@jesup.org