idnits 2.17.1 draft-singh-rmcat-cc-eval-04.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 document seems to lack a both a reference to RFC 2119 and the recommended RFC 2119 boilerplate, even if it appears to use RFC 2119 keywords. RFC 2119 keyword, line 704: '... 1. It MUST emit a packet at least ...' Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year -- The document date (October 20, 2013) is 3833 days in the past. Is this intentional? Checking references for intended status: Informational ---------------------------------------------------------------------------- == Unused Reference: 'RFC5681' is defined on line 596, but no explicit reference was found in the text == Unused Reference: 'SA4-EVAL' is defined on line 599, but no explicit reference was found in the text == Unused Reference: 'SA4-LR' is defined on line 603, but no explicit reference was found in the text == Unused Reference: 'TCP-eval-suite' is defined on line 606, but no explicit reference was found in the text == Outdated reference: A later version (-18) exists of draft-ietf-avtcore-rtp-circuit-breakers-01 Summary: 1 error (**), 0 flaws (~~), 6 warnings (==), 1 comment (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 RMCAT WG V. Singh 3 Internet-Draft J. Ott 4 Intended status: Informational Aalto University 5 Expires: April 23, 2014 October 20, 2013 7 Evaluating Congestion Control for Interactive Real-time Media 8 draft-singh-rmcat-cc-eval-04 10 Abstract 12 The Real-time Transport Protocol (RTP) is used to transmit media in 13 telephony and video conferencing applications. This document 14 describes the guidelines to evaluate new congestion control 15 algorithms for interactive point-to-point real-time media. 17 Status of This Memo 19 This Internet-Draft is submitted in full conformance with the 20 provisions of BCP 78 and BCP 79. 22 Internet-Drafts are working documents of the Internet Engineering 23 Task Force (IETF). Note that other groups may also distribute 24 working documents as Internet-Drafts. The list of current Internet- 25 Drafts is at http://datatracker.ietf.org/drafts/current/. 27 Internet-Drafts are draft documents valid for a maximum of six months 28 and may be updated, replaced, or obsoleted by other documents at any 29 time. It is inappropriate to use Internet-Drafts as reference 30 material or to cite them other than as "work in progress." 32 This Internet-Draft will expire on April 23, 2014. 34 Copyright Notice 36 Copyright (c) 2013 IETF Trust and the persons identified as the 37 document authors. All rights reserved. 39 This document is subject to BCP 78 and the IETF Trust's Legal 40 Provisions Relating to IETF Documents 41 (http://trustee.ietf.org/license-info) in effect on the date of 42 publication of this document. Please review these documents 43 carefully, as they describe your rights and restrictions with respect 44 to this document. Code Components extracted from this document must 45 include Simplified BSD License text as described in Section 4.e of 46 the Trust Legal Provisions and are provided without warranty as 47 described in the Simplified BSD License. 49 Table of Contents 51 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 3 52 2. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . 3 53 3. Metrics . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 54 3.1. RTP Log Format . . . . . . . . . . . . . . . . . . . . . 5 55 4. Guidelines . . . . . . . . . . . . . . . . . . . . . . . . . 5 56 4.1. Avoiding Congestion Collapse . . . . . . . . . . . . . . 5 57 4.2. Stability . . . . . . . . . . . . . . . . . . . . . . . . 5 58 4.3. Media Traffic . . . . . . . . . . . . . . . . . . . . . . 6 59 4.4. Start-up Behaviour . . . . . . . . . . . . . . . . . . . 6 60 4.5. Diverse Environments . . . . . . . . . . . . . . . . . . 6 61 4.6. Varying Path Characteristics . . . . . . . . . . . . . . 7 62 4.7. Reacting to Transient Events or Interruptions . . . . . . 7 63 4.8. Fairness With Similar Cross-Traffic . . . . . . . . . . . 7 64 4.9. Impact on Cross-Traffic . . . . . . . . . . . . . . . . . 7 65 4.10. Extensions to RTP/RTCP . . . . . . . . . . . . . . . . . 8 66 5. Minimum Requirements for Evaluation . . . . . . . . . . . . . 8 67 6. Evaluation Parameters . . . . . . . . . . . . . . . . . . . . 8 68 6.1. Bottleneck Traffic Flows . . . . . . . . . . . . . . . . 8 69 6.2. Access Links . . . . . . . . . . . . . . . . . . . . . . 9 70 6.3. Example Bottleneck Link Parameters . . . . . . . . . . . 9 71 6.4. DropTail Router Queue Parameters . . . . . . . . . . . . 10 72 6.5. Media Flow Parameters . . . . . . . . . . . . . . . . . . 11 73 6.6. Cross-traffic Parameters . . . . . . . . . . . . . . . . 11 74 7. Status of Proposals . . . . . . . . . . . . . . . . . . . . . 11 75 8. Security Considerations . . . . . . . . . . . . . . . . . . . 12 76 9. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 12 77 10. Contributors . . . . . . . . . . . . . . . . . . . . . . . . 12 78 11. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 12 79 12. References . . . . . . . . . . . . . . . . . . . . . . . . . 12 80 12.1. Normative References . . . . . . . . . . . . . . . . . . 12 81 12.2. Informative References . . . . . . . . . . . . . . . . . 13 82 Appendix A. Application Trade-off . . . . . . . . . . . . . . . 14 83 A.1. Measuring Quality . . . . . . . . . . . . . . . . . . . . 14 84 Appendix B. Proposal to evaluate Self-fairness of RMCAT 85 congestion control algorithm . . . . . . . . . . . . 14 86 B.1. Evaluation Parameters . . . . . . . . . . . . . . . . . . 15 87 B.1.1. Media Traffic Generator . . . . . . . . . . . . . . . 15 88 B.1.2. Bottleneck Link Bandwidth . . . . . . . . . . . . . . 16 89 B.1.3. Bottleneck Link Queue Type and Length . . . . . . . . 16 90 B.1.4. RMCAT flows and delay legs . . . . . . . . . . . . . 16 91 B.1.5. Impairment Generator . . . . . . . . . . . . . . . . 17 92 B.2. Proposed Passing Criteria . . . . . . . . . . . . . . . . 17 93 B.3. Extensibility of the Experiment . . . . . . . . . . . . . 17 94 Appendix C. Change Log . . . . . . . . . . . . . . . . . . . . . 18 95 C.1. Changes in draft-singh-rmcat-cc-eval-04 . . . . . . . . . 18 96 C.2. Changes in draft-singh-rmcat-cc-eval-03 . . . . . . . . . 18 97 C.3. Changes in draft-singh-rmcat-cc-eval-02 . . . . . . . . . 18 98 C.4. Changes in draft-singh-rmcat-cc-eval-01 . . . . . . . . . 18 99 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 19 101 1. Introduction 103 This memo describes the guidelines to help with evaluating new 104 congestion control algorithms for interactive point-to-point real 105 time media. The requirements for the congestion control algorithm 106 are outlined in [I-D.jesup-rmcat-reqs]). This document builds upon 107 previous work at the IETF: Specifying New Congestion Control 108 Algorithms [RFC5033] and Metrics for the Evaluation of Congestion 109 Control Algorithms [RFC5166]. 111 The guidelines proposed in the document are intended to help prevent 112 a congestion collapse, promote fair capacity usage and optimize the 113 media flow's throughput. Furthermore, the proposed algorithms are 114 expected to operate within the envelope of the circuit breakers 115 defined in [I-D.ietf-avtcore-rtp-circuit-breakers]. 117 This document only provides broad-level criteria for evaluating a new 118 congestion control algorithm and the working group should expect a 119 thorough scientific study to make its decision. The results of the 120 evaluation are not expected to be included within the internet-draft 121 but should be cited in the document. 123 2. Terminology 125 The terminology defined in RTP [RFC3550], RTP Profile for Audio and 126 Video Conferences with Minimal Control [RFC3551], RTCP Extended 127 Report (XR) [RFC3611], Extended RTP Profile for RTCP-based Feedback 128 (RTP/AVPF) [RFC4585] and Support for Reduced-Size RTCP [RFC5506] 129 apply. 131 3. Metrics 133 [RFC5166] describes the basic metrics for congestion control. 134 Metrics that are of interest for interactive multimedia are: 136 o Throughput. 138 o Minimizing oscillations in the transmission rate (stability) when 139 the end-to-end capacity varies slowly. 141 o Delay. 143 o Reactivity to transient events. 145 o Packet losses and discards. 147 o Section 2.1 of [RFC5166] discusses the tradeoff between 148 throughput, delay and loss. 150 Each experiment is expected to log every incoming and outgoing packet 151 (the RTP logging format is described in Section 3.1). The logging 152 can be done inside the application or at the endpoints using pcap 153 (packet capture, e.g., tcpdump, wireshark). The following are 154 calculated based on the information in the packet logs: 156 1. Sending rate, Receiver rate, Goodput 158 2. Packet delay 160 3. Packet loss 162 4. If using, retransmission or FEC: residual loss 164 5. Packets discarded from the playout or de-jitter buffer 166 [Open issue (1): The "unfairness" test is (measured at 1s intervals): 167 1. Do not trigger the circuit breaker. 168 2. Over 3 times or less than 1/3 times the throughput for an RMCAT 169 media stream compared to identical RMCAT streams competing on a 170 bottleneck, for a case when the competing streams have similar RTTs. 171 3. Over 3 times delay compared to RTT measurements performed before 172 starting the RMCAT flow or for the case when competing with identical 173 RMCAT streams having similar RTTs. 174 ] 176 [Open issue (2): Possibly using Jain-fairness index.] 178 Convergence time: the time taken to reach a stable rate at startup, 179 after the available link capacity changes, or when new flows get 180 added to the bottleneck link. 182 Bandwidth Utilization, defined as ratio of the instantaneous sending 183 rate to the instantaneous bottleneck capacity. This metric is useful 184 when an RMCAT flow is by itself or competing with similar cross- 185 traffic. 187 From the logs the statistical measures (min, max, mean, standard 188 deviation and variance) for the whole duration or any specific part 189 of the session can be calculated. Also the metrics (sending rate, 190 receiver rate, goodput, latency) can be visualized in graphs as 191 variation over time, the measurements in the plot are at 1 second 192 intervals. Additionally, from the logs it is possible to plot the 193 histogram or CDF of packet delay. 195 3.1. RTP Log Format 197 The log file is tab or comma separated containing the following 198 details: 200 Send or receive timestamp (unix) 201 RTP payload type 202 SSRC 203 RTP sequence no 204 RTP timestamp 205 marker bit 206 payload size 208 If the congestion control implements, retransmissions or FEC, the 209 evaluation should report both packet loss (before applying error- 210 resilience) and residual packet loss (after applying error- 211 resilience). 213 4. Guidelines 215 A congestion control algorithm should be tested in simulation or a 216 testbed environment, and the experiments should be repeated multiple 217 times to infer statistical significance. The following guidelines 218 are considered for evaluation: 220 4.1. Avoiding Congestion Collapse 222 The congestion control algorithm is expected to take an action, such 223 as reducing the sending rate, when it detects congestion. Typically, 224 it should intervene before the circuit breaker 225 [I-D.ietf-avtcore-rtp-circuit-breakers] is engaged. 227 Does the congestion control propose any changes to (or diverge from) 228 the circuit breaker conditions defined in 229 [I-D.ietf-avtcore-rtp-circuit-breakers]. 231 4.2. Stability 232 The congestion control should be assessed for its stability when the 233 path characteristics do not change over time. Changing the media 234 encoding rate estimate too often or by too much may adversely affect 235 the application layer performance. 237 4.3. Media Traffic 239 The congestion control algorithm should be assessed with different 240 types of media behavior, i.e., the media should contain idle and 241 data-limited periods. For example, periods of silence for audio, 242 varying amount of motion for video, or bursty nature of I-frames. 244 The evaluation may be done in two stages. In the first stage, the 245 endpoint generates traffic at the rate calculated by the congestion 246 controller. In the second stage, real codecs or models of video 247 codecs are used to mimic application-limited data periods and varying 248 video frame sizes. 250 4.4. Start-up Behaviour 252 The congestion control algorithm should be assessed with different 253 start-rates. The main reason is to observe the behavior of the 254 congestion control in different evaluation scenarios, such as when 255 competing with varying amount of cross-traffic or how quickly does 256 the congestion control algorithm achieve a stable sending rate. 258 [Editor's note: requires a robust definition for unfriendliness and 259 convergence time.] 261 4.5. Diverse Environments 263 The congestion control algorithm should be assessed in heterogeneous 264 environments, containing both wired and wireless paths. Examples of 265 wireless access technologies are: 802.11, GPRS, HSPA, or LTE. One of 266 the main challenges of the wireless environments for the congestion 267 control algorithm is to distinguish between congestion induced loss 268 and transmission (bit-error corruption) loss. Congestion control 269 algorithms may incorrectly identify transmission loss as congestion 270 loss and reduce the media encoding rate by too much, which may cause 271 oscillatory behavior and deteriorate the users' quality of 272 experience. Furthermore, packet loss may induce additional delay in 273 networks with wireless paths due to link-layer retransmissions. 275 4.6. Varying Path Characteristics 277 The congestion control algorithm should be evaluated for a range of 278 path characteristics such as, different end-to-end capacity and 279 latency, varying amount of cross traffic on a bottleneck link and a 280 router's queue length. For the moment, only DropTail queues are 281 used. However, if new Active Queue Management (AQM) schemes become 282 available, the performance of the congestion control algorithm should 283 be again evaluated. 285 In an experiment, if the media only flows in a single direction, the 286 feedback path should also be tested with varying amounts of 287 impairments. 289 The main motivation for the previous and current criteria is to 290 identify situations in which the proposed congestion control is less 291 performant. 293 4.7. Reacting to Transient Events or Interruptions 295 The congestion control algorithm should be able to handle changes in 296 end-to-end capacity and latency. Latency may change due to route 297 updates, link failures, handovers etc. In mobile environment the 298 end-to-end capacity may vary due to the interference, fading, 299 handovers, etc. In wired networks the end-to-end capacity may vary 300 due to changes in resource reservation. 302 4.8. Fairness With Similar Cross-Traffic 304 The congestion control algorithm should be evaluated when competing 305 with other RTP flows using the same or another candidate congestion 306 control algorithm. The proposal should highlight the bottleneck 307 capacity share of each RTP flow. 309 [Editor's note: If we define Unfriendliness then that criteria should 310 be applied here.] 312 4.9. Impact on Cross-Traffic 314 The congestion control algorithm should be evaluated when competing 315 with standard TCP. Short TCP flows may be considered as transient 316 events and the RTP flow may give way to the short TCP flow to 317 complete quickly. However, long-lived TCP flows may starve out the 318 RTP flow depending on router queue length. 320 The proposal should also measure the impact on varied number of 321 cross-traffic sources, i.e., few and many competing flows, or mixing 322 various amounts of TCP and similar cross-traffic. 324 4.10. Extensions to RTP/RTCP 326 The congestion control algorithm should indicate if any protocol 327 extensions are required to implement it and should carefully describe 328 the impact of the extension. 330 5. Minimum Requirements for Evaluation 332 [Editor's Note: If needed, a minimum evaluation criteria can be based 333 on the above guidelines or defined tests/scenarios.] 335 6. Evaluation Parameters 337 An evaluation scenario is created from a list of network, link and 338 flow characteristics. The example parameters discussed in the 339 following subsections are meant to aid in creating evaluation 340 scenarios and do not describe an evaluation scenario. The scenario 341 discussed in Appendix B takes into account all these parameters. 343 6.1. Bottleneck Traffic Flows 345 The network scenario describes the types of flows sharing the common 346 bottleneck with a single RMCAT flow, they are: 348 1. A single RMCAT flow by itself. 350 2. Competing with similar RMCAT flows. These competing flows may 351 use the same algorithm or another candidate RMCAT algorithm. 353 3. Compete with long-lived TCP. 355 4. Compete with bursty TCP. 357 5. Compete with LEDBAT flows. 359 6. Compete with unresponsive interactive media flows (i.e., not only 360 CBR). 362 Figure 1 shows an example evaluation topology, where S1..Sn are 363 traffic sources, these sources are either RMCAT or a mixture of 364 traffic flows listed above. R1..Rn are the corresponding receivers. 365 A and B are routers that can be configured to introduce impairments. 366 Access links are in between the sender/receiver and the router, while 367 the bottleneck link is between the Routers A and B. 369 +---+ Access Access +---+ 370 |S1 |======= \ / =======|R1 | 371 +---+ link \\ // link +---+ 372 \\ // 373 +---+ +-----+ Bottleneck +-----+ +---+ 374 |S2 |=======| A |------------------------------>| B |=======|R2 | 375 +---+ | |<------------------------------| | +---+ 376 +-----+ Link +-----+ 377 (...) // \\ (...) 378 // \\ 379 +---+ // \\ +---+ 380 |Sn |====== / \ ======|Rn | 381 +---+ +---+ 383 Figure 1: Simple Topology 385 [Open Issue: Discuss more complex topologies] 387 6.2. Access Links 389 The media senders and receivers are typically connected to the 390 bottleneck link, common access links are: 392 1. Ethernet (LAN) 394 2. Wireless LAN (WLAN) 396 3. 3G/LTE 398 [Open issue: point to a reference containing parameters or traces to 399 model WLAN and 3G/LTE.] 401 A real-world network typically consists of a mixture of links, the 402 most important aspect is to identify the location of the bottleneck 403 link. The bottleneck link can move from one node to another 404 depending on the amount of cross-traffic or due to the varying link 405 capacity. The design of the experiments should take this into 406 account. In the simplest case the access link may not be the 407 bottleneck link but an intermediate node. 409 6.3. Example Bottleneck Link Parameters 411 The bottleneck link carries multiple flows, these flows may be other 412 RMCAT flows or other types of cross-traffic. The experiments should 413 dimension the bottleneck link based on the number of flows and the 414 expected behavior. For example, if 5 media flows are expected to 415 share the bottleneck link equally, the bottleneck link is set to 5 416 times the desired transmission rate. 418 If the experiment carries only media in one direction, then the 419 upstream (sender to receiver) bottleneck link carries media packets 420 while the downstream (receiver to sender) bottleneck carries the 421 feedback packets. The bottleneck link parameters discussed in this 422 section apply only to a single direction, hence the bottleneck link 423 in the reverse direction can choose the same or have different 424 parameters. 426 The link latency corresponds to the propagation delay of the link, 427 i.e., the time it takes for a packet to traverse the bottleneck link, 428 it does not include queuing delay. In an experiment with several 429 links the experiment should describe if the links add latency or not. 430 It is possible for experiments to have multiple hops with different 431 link latencies. Experiments are expected to verify that the 432 congestion control is able to work in challenging situations, for 433 example over trans-continental and/or satellite links. The 434 experiment should pick link latency values from the following: 436 1. Very low latency: 0-1ms 438 2. Low latency: 50ms 440 3. High latency: 150ms 442 4. Extreme latency: 300ms 444 Similarly, to model lossy links, the experiments can choose one of 445 the following loss rates, the fractional loss is the ratio of packets 446 lost and packets sent. 448 1. no loss: 0% 450 2. 1% 452 3. 5% 454 4. 10% 456 5. 20% 458 These fractional losses can be generated using traces, Gilbert-Elliot 459 model, randomly (uncorrelated) loss. 461 6.4. DropTail Router Queue Parameters 463 The router queue length is measured as the time taken to drain the 464 FIFO queue, they are: 466 1. QoS-aware (or short): 70ms 467 2. Nominal: 500ms 469 3. Buffer-bloated: 2000ms 471 However, the size of the queue is typically measured in bytes or 472 packets and to convert the queue length measured in seconds to queue 473 length in bytes: 475 QueueSize (in bytes) = QueueSize (in sec) x Throughput (in bps)/8 477 6.5. Media Flow Parameters 479 The media sources can be modeled in two ways. In the first, the 480 sources always have data to send, i.e., have no data limited 481 intervals and are able to generate the media rate requested by the 482 RMCAT congestion control algorithm. In the second, the traffic 483 generator models the behavior of a media codec, mainly the burstiness 484 (time-varying data produced by a video GOP). 486 At the beginning of the session, the media sources are configured to 487 start at a given start rate, they are: 489 1. 200 kbps 491 2. 800 kbps 493 3. 1300 kbps 495 4. 4000 kbps 497 6.6. Cross-traffic Parameters 499 Long-lived TCP flows will download data throughout the session and 500 are expected to have infinite amount of data to send or receive.] 502 [Open issue: short-lived/bursty TCP cross-traffic parameters are 503 still TBD. 505 7. Status of Proposals 507 Congestion control algorithms are expected to be published as 508 "Experimental" documents until they are shown to be safe to deploy. 509 An algorithm published as a draft should be experimented in 510 simulation, or a controlled environment (testbed) to show its 511 applicability. Every congestion control algorithm should include a 512 note describing the environments in which the algorithm is tested and 513 safe to deploy. It is possible that an algorithm is not recommended 514 for certain environments or perform sub-optimally for the user. 516 [Editor's Note: Should there be a distinction between "Informational" 517 and "Experimental" drafts for congestion control algorithms in RMCAT. 518 [RFC5033] describes Informational proposals as algorithms that are 519 not safe for deployment but are proposals to experiment with in 520 simulation/testbeds. While Experimental algorithms are ones that are 521 deemed safe in some environments but require a more thorough 522 evaluation (from the community).] 524 8. Security Considerations 526 Security issues have not been discussed in this memo. 528 9. IANA Considerations 530 There are no IANA impacts in this memo. 532 10. Contributors 534 The content and concepts within this document are a product of the 535 discussion carried out in the Design Team. 537 Michael Ramalho provided the text for the scenario discussed in 538 Appendix B. 540 11. Acknowledgements 542 Much of this document is derived from previous work on congestion 543 control at the IETF. 545 The authors would like to thank Harald Alvestrand, Luca De Cicco, 546 Wesley Eddy, Lars Eggert, Kevin Gross, Vinayak Hegde, Stefan Holmer, 547 Randell Jesup, Piers O'Hanlon, Colin Perkins, Michael Ramalho, 548 Zaheduzzaman Sarker, Timothy B. Terriberry, Michael Welzl, and Mo 549 Zanaty for providing valuable feedback on earlier versions of this 550 draft. Additionally, also thank the participants of the design team 551 for their comments and discussion related to the evaluation criteria. 553 12. References 555 12.1. Normative References 557 [RFC3550] Schulzrinne, H., Casner, S., Frederick, R., and V. 558 Jacobson, "RTP: A Transport Protocol for Real-Time 559 Applications", STD 64, RFC 3550, July 2003. 561 [RFC3551] Schulzrinne, H. and S. Casner, "RTP Profile for Audio and 562 Video Conferences with Minimal Control", STD 65, RFC 3551, 563 July 2003. 565 [RFC3611] Friedman, T., Caceres, R., and A. Clark, "RTP Control 566 Protocol Extended Reports (RTCP XR)", RFC 3611, November 567 2003. 569 [RFC4585] Ott, J., Wenger, S., Sato, N., Burmeister, C., and J. Rey, 570 "Extended RTP Profile for Real-time Transport Control 571 Protocol (RTCP)-Based Feedback (RTP/AVPF)", RFC 4585, July 572 2006. 574 [RFC5506] Johansson, I. and M. Westerlund, "Support for Reduced-Size 575 Real-Time Transport Control Protocol (RTCP): Opportunities 576 and Consequences", RFC 5506, April 2009. 578 [I-D.jesup-rmcat-reqs] 579 Jesup, R., "Congestion Control Requirements For RMCAT", 580 draft-jesup-rmcat-reqs-01 (work in progress), February 581 2013. 583 [I-D.ietf-avtcore-rtp-circuit-breakers] 584 Perkins, C. and V. Singh, "RTP Congestion Control: Circuit 585 Breakers for Unicast Sessions", draft-ietf-avtcore-rtp- 586 circuit-breakers-01 (work in progress), October 2012. 588 12.2. Informative References 590 [RFC5033] Floyd, S. and M. Allman, "Specifying New Congestion 591 Control Algorithms", BCP 133, RFC 5033, August 2007. 593 [RFC5166] Floyd, S., "Metrics for the Evaluation of Congestion 594 Control Mechanisms", RFC 5166, March 2008. 596 [RFC5681] Allman, M., Paxson, V., and E. Blanton, "TCP Congestion 597 Control", RFC 5681, September 2009. 599 [SA4-EVAL] 600 R1-081955, 3GPP., "LTE Link Level Throughput Data for SA4 601 Evaluation Framework", 3GPP R1-081955, 5 2008. 603 [SA4-LR] S4-050560, 3GPP., "Error Patterns for MBMS Streaming over 604 UTRAN and GERAN", 3GPP S4-050560, 5 2008. 606 [TCP-eval-suite] 607 Lachlan, A., Marcondes, C., Floyd, S., Dunn, L., Guillier, 608 R., Gang, W., Eggert, L., Ha, S., and I. Rhee, "Towards a 609 Common TCP Evaluation Suite", Proc. PFLDnet. 2008, August 610 2008. 612 Appendix A. Application Trade-off 614 Application trade-off is yet to be defined. see RMCAT requirements 615 [I-D.jesup-rmcat-reqs] document. Perhaps each experiment should 616 define the application's expectation or trade-off. 618 A.1. Measuring Quality 620 No quality metric is defined for performance evaluation, it is 621 currently an open issue. However, there is consensus that congestion 622 control algorithm should be able to show that it is useful for 623 interactive video by performing analysis using a real codec and video 624 sequences. 626 Appendix B. Proposal to evaluate Self-fairness of RMCAT congestion 627 control algorithm 629 The goal of the experiment discussed in this section is to initially 630 take out as many unknowns from the scenario. Later experiments can 631 define more complex environments, topologies and media behavior. 632 This experiment evaluates the performance of the RMCAT sender 633 competing with other similar RMCAT flows (running the same algorithm 634 or other RMCAT proposals) on the bottleneck link. There are up to 20 635 RMCAT flows competing for capacity, but the media only flows in one 636 direction, from senders (S1..S20) to receivers (R1..R20) and the 637 feedback packets flow in the reverse direction. 639 Figure 2 shows the experiment setup and it has subtle differences 640 compared to the simple topology in Figure 1. Groups of 10 receivers 641 are connected to the bottleneck link through two different routers 642 (Router C and D). The rationale for adding these additional routers 643 is to create two delay legs, i.e., two groups of endpoints with 644 different network latencies and measure the performance of the RMCAT 645 congestion control algorithm. If fewer than 10 sources are 646 initialized, all traffic flows experience the same delay because they 647 share the same delay leg. 649 Router A has a single forward direction bottleneck link (i.e., the 650 bottleneck capacity and delay constraints applies only to the media 651 packets going from the sender to the receiver, the feedback packets 652 are unaffected). Hence, the Round-Trip Time (RTT) is primarily 653 composed of the bottleneck queue delay and any forward path 654 (propagation) latency. The main reason for not applying any 655 constraints on the return path is to provide the best-case 656 performance scenario for the congestion control algorithm. In later 657 experiments, it is possible to add similar capacity and delay 658 constraints on the return path. 660 +---+ 661 / === |R1 | 662 +---+ +-----+ // +---+ 663 |S1 |======= \ / =| C | // 664 +---+ \\ // +-----+ \\ (...) 665 \\ // \\ 666 +---+ +-----+ Bottleneck +-----+ \\ +---+ 667 |S2 |=======| A |-------------------->| B | \ ===|R10| 668 +---+ | |<--------------------| | +---+ 669 +-----+ Link +-----+ 670 (...) // \\ +---+ 671 // \\ / === |R11| 672 +---+ // \\ +-----+ // +---+ 673 |S20|====== / \ =| D |// 674 +---+ +-----+\\ (...) 675 \\ 676 \\ +---+ 677 \ ===|R20| 678 +---+ 680 Figure 2: Self-fairness Evaluation Setup 682 Loss impairments are applied at Router C and Router D, but only to 683 the feedback flows. If the losses are set to 0%, it represents a 684 case where the return path is over-provisioned for all traffic. In 685 later experiments the loss impairments can be added to the media path 686 as well. 688 The media sources are configured to send infinite amount of data, 689 i.e., the sources always have data to send and have no data limited 690 intervals. Additionally, the media sources are always successful in 691 generating the media rate requested by the RMCAT congestion control 692 algorithm. In this experiment, we avoid the potentially complicated 693 scenario of using media traffic generators that try to model the 694 behavior of media codecs (mainly the burstiness). 696 B.1. Evaluation Parameters 698 B.1.1. Media Traffic Generator 700 The media source always generates at the rate requested by the 701 congestion control and has infinite data to send. Furthermore, the 702 media packet generator is subject to the following constraints: 704 1. It MUST emit a packet at least once per 100 ms time interval. 706 2. For low media rate source: when generating data at a rate less 707 than a maximum length MTU every 100 ms would allow (e.g., 120 708 kbps = 1500 bytes/packet * 10 packets/sec * 8 bits/byte), the 709 RMCAT source must modulate the packet size (RTP payload size) of 710 RTP packets that are sent every 100 ms to attain the desired 711 rate. 713 3. For high media rate sources: when generating data at a rate 714 greater than a maximum length MTU every 100 ms would allow, the 715 source must do so by sending (approximately) maximum MTU sized 716 packets and adjusting the inter-departure interval to be 717 approximately equal. The intent of this to ensure the data is 718 sent relatively smoothly independent of the bit rate, subject to 719 the first constraint. 721 B.1.2. Bottleneck Link Bandwidth 723 The bottleneck link capacity is dimensioned such that each RMCAT flow 724 in an ideal situation with perfectly equal capacity sharing for all 725 the flows on the bottleneck obtains the following throughputs: 200 726 kbps, 800 kbps, 1.3 Mbps and 4 Mbps. 727 For example, experiments with five RMCAT flows with an 800 kbps/flow 728 target rate should set the bottleneck link capacity to 4 Mbps. 730 B.1.3. Bottleneck Link Queue Type and Length 732 The bottleneck link queue (Router A) is a simple FIFO queue having a 733 buffer length corresponding to 70 ms, 500 ms or 2000 ms (defined in 734 Section 6.4) of delay at the bottleneck link rate (i.e., actual 735 buffer lengths in bytes are dependent on bottleneck link bandwidth). 737 B.1.4. RMCAT flows and delay legs 739 Experiments run with 1, 3, 5, 10 and 20 RMCAT sources, they are 740 outlined as follows: 742 1. Experiments with 1, 3, and 5 RMCAT flows, all RMCAT flows 743 commence simultaneously. A single delay leg is used and the link 744 latency is set to one of the following : 0 ms, 50 ms and 150 ms. 746 2. For 10 and 20 source experiments where all RMCAT flows begin 747 simultaneously the sources are split evenly into two different 748 bulk delay legs. One leg is set to 0 ms bulk delay leg and the 749 other is set to 150 ms. 751 3. For 10 and 20 source experiments where the first set will use 0 752 ms of bulk delay and the second set will use 150 ms bulk delay. 754 1. Random starts within interval [0 ms, 500 ms]. 756 2. One "early-coming" flow (i.e., the 1st flow starting and 757 achieving steady-state before the next N-1 simultaneously 758 begin). 760 3. One "late-coming" flow (i.e., the Nth flow starting after 761 steady-state has occurred for the existing N-1 flows). 763 These cases assess if there are any early or late-comer 764 advantages or disadvantages for a particular algorithm and to see 765 if any unfairness is reproducible or unpredictable. 767 [Open issue (A.1): which group does the early and late flow belong 768 to?] 770 [Open issue (A.2): Start rate for the media flows] 772 B.1.5. Impairment Generator 774 Packet loss is created in the reverse path (affects only feedback 775 packets). Cases of 0%, 1%, 5% and 10% are studied for the 1, 3, and 776 5 RMCAT flow experiments, losses are not applied to flows with 10 or 777 20 RMCAT flows. 779 B.2. Proposed Passing Criteria 781 [Editor's note: there has been little or no discussion on the below 782 criteria, however, they are listed here for the sake of completeness. 784 No unfairness is observed, i.e., at steady state each flow attains a 785 throughput between [ B/(3*N), (3*B)/N ], where B is the link 786 bandwidth and N is the number of flows. 788 No flow experiences packet loss when queue length is set to 500 ms or 789 greater. 791 All individual sources must be in their steady state within twenty 792 LRTTs (where LRTT is defined as the RTT associated with the flow with 793 the Largest RTT in the experiment). ] 795 B.3. Extensibility of the Experiment 797 The above scenario describes only RMCAT sources competing for 798 capacity on the bottleneck link, however, future experiments can use 799 different types of cross-traffic (as described in Section 6.1). 801 Currently, the forward path (carrying media packets) is characterized 802 to add delay and a fixed bottleneck link capacity, in the future 803 packet losses and capacity changes can be applied to mimic a wireless 804 link layer (for e.g., WiFi, 3G, LTE). Additionally, only losses are 805 applied to the reverse path (carrying feedback packets), later 806 experiments can apply the same forward path (carrying media packets) 807 impairments to the reverse path. 809 Appendix C. Change Log 811 Note to the RFC-Editor: please remove this section prior to 812 publication as an RFC. 814 C.1. Changes in draft-singh-rmcat-cc-eval-04 816 o Incorporate feedback from IETF 87, Berlin. 818 o Clarified metrics: convergence time, bandwidth utilization. 820 o Changed fairness criteria to fairness test. 822 o Added measuring pre- and post-repair loss. 824 o Added open issue of measuring video quality to appendix. 826 o clarified use of DropTail and AQM. 828 o Updated text in "Minimum Requirements for Evaluation" 830 C.2. Changes in draft-singh-rmcat-cc-eval-03 832 o Incorporate the discussion within the design team. 834 o Added a section on evaluation parameters, it describes the flow 835 and network characteristics. 837 o Added Appendix with self-fairness experiment. 839 o Changed bottleneck parameters from a proposal to an example set. 841 C.3. Changes in draft-singh-rmcat-cc-eval-02 843 o Added scenario descriptions. 845 C.4. Changes in draft-singh-rmcat-cc-eval-01 847 o Removed QoE metrics. 849 o Changed stability to steady-state. 851 o Added measuring impact against few and many flows. 853 o Added guideline for idle and data-limited periods. 855 o Added reference to TCP evaluation suite in example evaluation 856 scenarios. 858 Authors' Addresses 860 Varun Singh 861 Aalto University 862 School of Electrical Engineering 863 Otakaari 5 A 864 Espoo, FIN 02150 865 Finland 867 Email: varun@comnet.tkk.fi 868 URI: http://www.netlab.tkk.fi/~varun/ 870 Joerg Ott 871 Aalto University 872 School of Electrical Engineering 873 Otakaari 5 A 874 Espoo, FIN 02150 875 Finland 877 Email: jo@comnet.tkk.fi