idnits 2.17.1 draft-ietf-rmcat-sbd-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 : ---------------------------------------------------------------------------- No issues found here. Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year -- The document date (July 1, 2015) is 3194 days in the past. Is this intentional? Checking references for intended status: Experimental ---------------------------------------------------------------------------- == Outdated reference: A later version (-05) exists of draft-welzl-rmcat-coupled-cc-04 Summary: 0 errors (**), 0 flaws (~~), 2 warnings (==), 1 comment (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 RTP Media Congestion Avoidance D. Hayes, Ed. 3 Techniques University of Oslo 4 Internet-Draft S. Ferlin 5 Intended status: Experimental Simula Research Laboratory 6 Expires: January 2, 2016 M. Welzl 7 University of Oslo 8 July 1, 2015 10 Shared Bottleneck Detection for Coupled Congestion Control for RTP 11 Media. 12 draft-ietf-rmcat-sbd-01 14 Abstract 16 This document describes a mechanism to detect whether end-to-end data 17 flows share a common bottleneck. It relies on summary statistics 18 that are calculated by a data receiver based on continuous 19 measurements and regularly fed to a grouping algorithm that runs 20 wherever the knowledge is needed. This mechanism complements the 21 coupled congestion control mechanism in draft-welzl-rmcat-coupled-cc. 23 Status of this Memo 25 This Internet-Draft is submitted in full conformance with the 26 provisions of BCP 78 and BCP 79. 28 Internet-Drafts are working documents of the Internet Engineering 29 Task Force (IETF). Note that other groups may also distribute 30 working documents as Internet-Drafts. The list of current Internet- 31 Drafts is at http://datatracker.ietf.org/drafts/current/. 33 Internet-Drafts are draft documents valid for a maximum of six months 34 and may be updated, replaced, or obsoleted by other documents at any 35 time. It is inappropriate to use Internet-Drafts as reference 36 material or to cite them other than as "work in progress." 38 This Internet-Draft will expire on January 2, 2016. 40 Copyright Notice 42 Copyright (c) 2015 IETF Trust and the persons identified as the 43 document authors. All rights reserved. 45 This document is subject to BCP 78 and the IETF Trust's Legal 46 Provisions Relating to IETF Documents 47 (http://trustee.ietf.org/license-info) in effect on the date of 48 publication of this document. Please review these documents 49 carefully, as they describe your rights and restrictions with respect 50 to this document. Code Components extracted from this document must 51 include Simplified BSD License text as described in Section 4.e of 52 the Trust Legal Provisions and are provided without warranty as 53 described in the Simplified BSD License. 55 Table of Contents 57 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 3 58 1.1. The signals . . . . . . . . . . . . . . . . . . . . . . . 3 59 1.1.1. Packet Loss . . . . . . . . . . . . . . . . . . . . . 3 60 1.1.2. Packet Delay . . . . . . . . . . . . . . . . . . . . . 3 61 1.1.3. Path Lag . . . . . . . . . . . . . . . . . . . . . . . 4 62 2. Definitions . . . . . . . . . . . . . . . . . . . . . . . . . 4 63 2.1. Parameters and their Effect . . . . . . . . . . . . . . . 5 64 2.2. Recommended Parameter Values . . . . . . . . . . . . . . . 7 65 3. Mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . 7 66 3.1. Key metrics and their calculation . . . . . . . . . . . . 9 67 3.1.1. Mean delay . . . . . . . . . . . . . . . . . . . . . . 9 68 3.1.2. Skewness Estimate . . . . . . . . . . . . . . . . . . 9 69 3.1.3. Variability Estimate . . . . . . . . . . . . . . . . . 10 70 3.1.4. Oscillation Estimate . . . . . . . . . . . . . . . . . 11 71 3.1.5. Packet loss . . . . . . . . . . . . . . . . . . . . . 11 72 3.2. Flow Grouping . . . . . . . . . . . . . . . . . . . . . . 12 73 3.2.1. Flow Grouping Algorithm . . . . . . . . . . . . . . . 12 74 3.2.2. Using the flow group signal . . . . . . . . . . . . . 13 75 3.3. Removing Noise from the Estimates . . . . . . . . . . . . 13 76 3.3.1. PDV noise . . . . . . . . . . . . . . . . . . . . . . 14 77 3.3.2. Oscillation noise . . . . . . . . . . . . . . . . . . 14 78 3.3.3. Clock skew . . . . . . . . . . . . . . . . . . . . . . 15 79 3.4. Reducing lag and Improving Responsiveness . . . . . . . . 15 80 3.4.1. Improving the response of the skewness estimate . . . 16 81 3.4.2. Improving the response of the variability estimate . . 16 82 4. Measuring OWD . . . . . . . . . . . . . . . . . . . . . . . . 17 83 4.1. Time stamp resolution . . . . . . . . . . . . . . . . . . 17 84 5. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . 17 85 6. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 17 86 7. Security Considerations . . . . . . . . . . . . . . . . . . . 17 87 8. Change history . . . . . . . . . . . . . . . . . . . . . . . . 18 88 9. References . . . . . . . . . . . . . . . . . . . . . . . . . . 18 89 9.1. Normative References . . . . . . . . . . . . . . . . . . . 18 90 9.2. Informative References . . . . . . . . . . . . . . . . . . 18 91 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . . 19 93 1. Introduction 95 In the Internet, it is not normally known if flows (e.g., TCP 96 connections or UDP data streams) traverse the same bottlenecks. Even 97 flows that have the same sender and receiver may take different paths 98 and share a bottleneck or not. Flows that share a bottleneck link 99 usually compete with one another for their share of the capacity. 100 This competition has the potential to increase packet loss and 101 delays. This is especially relevant for interactive applications 102 that communicate simultaneously with multiple peers (such as multi- 103 party video). For RTP media applications such as RTCWEB, 104 [I-D.welzl-rmcat-coupled-cc] describes a scheme that combines the 105 congestion controllers of flows in order to honor their priorities 106 and avoid unnecessary packet loss as well as delay. This mechanism 107 relies on some form of Shared Bottleneck Detection (SBD); here, a 108 measurement-based SBD approach is described. 110 1.1. The signals 112 The current Internet is unable to explicitly inform endpoints as to 113 which flows share bottlenecks, so endpoints need to infer this from 114 whatever information is available to them. The mechanism described 115 here currently utilises packet loss and packet delay, but is not 116 restricted to these. 118 1.1.1. Packet Loss 120 Packet loss is often a relatively rare signal. Therefore, on its own 121 it is of limited use for SBD, however, it is a valuable supplementary 122 measure when it is more prevalent. 124 1.1.2. Packet Delay 126 End-to-end delay measurements include noise from every device along 127 the path in addition to the delay perturbation at the bottleneck 128 device. The noise is often significantly increased if the round-trip 129 time is used. The cleanest signal is obtained by using One-Way-Delay 130 (OWD). 132 Measuring absolute OWD is difficult since it requires both the sender 133 and receiver clocks to be synchronised. However, since the 134 statistics being collected are relative to the mean OWD, a relative 135 OWD measurement is sufficient. Clock skew is not usually significant 136 over the time intervals used by this SBD mechanism (see [RFC6817] A.2 137 for a discussion on clock skew and OWD measurements). However, in 138 circumstances where it is significant, Section 3.3.3 outlines a way 139 of adjusting the calculations to cater for it. 141 Each packet arriving at the bottleneck buffer may experience very 142 different queue lengths, and therefore different waiting times. A 143 single OWD sample does not, therefore, characterize the path well. 144 However, multiple OWD measurements do reflect the distribution of 145 delays experienced at the bottleneck. 147 1.1.3. Path Lag 149 Flows that share a common bottleneck may traverse different paths, 150 and these paths will often have different base delays. This makes it 151 difficult to correlate changes in delay or loss. This technique uses 152 the long term shape of the delay distribution as a base for 153 comparison to counter this. 155 2. Definitions 157 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 158 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 159 document are to be interpreted as described in RFC 2119 [RFC2119]. 161 Acronyms used in this document: 163 OWD -- One Way Delay 165 PDV -- Packet Delay Variation 167 MAD -- Mean Absolute Deviation 169 RTT -- Round Trip Time 171 SBD -- Shared Bottleneck Detection 173 Conventions used in this document: 175 T -- the base time interval over which measurements are 176 made. 178 N -- the number of base time, T, intervals used in some 179 calculations. 181 sum_T(...) -- summation of all the measurements of the variable 182 in parentheses taken over the interval T 184 sum(...) -- summation of terms of the variable in parentheses 185 sum_N(...) -- summation of N terms of the variable in parentheses 187 sum_NT(...) -- summation of all measurements taken over the 188 interval N*T 190 E_T(...) -- the expectation or mean of the measurements of the 191 variable in parentheses over T 193 E_N(...) -- the expectation or mean of the last N values of the 194 variable in parentheses 196 E_M(...) -- the expectation or mean of the last M values of the 197 variable in parentheses, where M <= N. 199 max_T(...) -- the maximum recorded measurement of the variable in 200 parentheses taken over the interval T 202 min_T(...) -- the minimum recorded measurement of the variable in 203 parentheses taken over the interval T 205 num_T(...) -- the count of measurements of the variable in 206 parentheses taken in the interval T 208 num_VM(...) -- the count of valid values of the variable in 209 parentheses given M records 211 PC -- a boolean variable indicating the particular flow 212 was identified as experiencing congestion in the 213 previous interval T (i.e. Previously Congested) 215 skew_est -- a measure of skewness in a OWD distribution. 217 var_est -- a measure of variability in OWD measurements. 219 freq_est -- a measure of low frequency oscillation in the OWD 220 measurements. 222 p_l, p_f, p_pdv, p_mad, c_s, c_h, p_s, p_d, p_v -- various 223 thresholds used in the mechanism 225 M and F -- number of values related to N 227 2.1. Parameters and their Effect 228 T T should be long enough so that there are enough packets 229 received during T for a useful estimate of short term mean 230 OWD and variation statistics. Making T too large can limit 231 the efficacy of PDV and freq_est. It will also increase the 232 response time of the mechanism. Making T too small will make 233 the metrics noisier. 235 N & M N should be large enough provide a stable estimate of 236 oscillations in OWD and average PDV. Usually M=N, though 237 having M mean_delay) 396 where 398 if (OWD < mean_delay) 1 else 0 400 if (OWD > mean_delay) 1 else 0 402 and mean_delay does not include the mean of the current T 403 interval. 405 skew_est = sum_MT(skew_base_T)/num_MT(OWD) 407 where skew_est is a number between -1 and 1 409 Note: Care must be taken when implementing the comparisons to ensure 410 that rounding does not bias skew_est. It is important that the mean 411 is calculated with a higher precision than the samples. 413 3.1.3. Variability Estimate 415 Packet Delay Variation (PDV) ([RFC5481] and [ITU-Y1540]) is used as 416 an estimator of the variability of the delay signal. We define PDV 417 as follows: 419 PDV = PDV_max = max_T(OWD) - E_T(OWD) 421 var_est = E_M(PDV) = sum_M(PDV) / M 423 This modifies PDV as outlined in [RFC5481] to provide a summary 424 statistic version that best aids the grouping decisions of the 425 algorithm (see [Hayes-LCN14] section IVB). 427 Generally the maximum is sampled well during congestion, though it is 428 more sensitive to path and operating system noise. The use of PDV = 429 PDV_min = E_T(OWD) - min_T(OWD) would be less sensitive to this 430 noise, but is not well sampled during congestion at the bottleneck 431 and therefore not recommended. 433 3.1.4. Oscillation Estimate 435 An estimate of the low frequency oscillation of the delay signal is 436 calculated by counting and normalising the significant mean, 437 E_T(OWD), crossings of mean_delay: 439 freq_est = number_of_crossings / N 441 where we define a significant mean crossing as a crossing that 442 extends p_v * var_est from mean_delay. In our experiments we 443 have found that p_v = 0.2 is a good value. 445 Freq_est is a number between 0 and 1. Freq_est can be approximated 446 incrementally as follows: 448 With each new calculation of E_T(OWD) a decision is made as to 449 whether this value of E_T(OWD) significantly crosses the current 450 long term mean, mean_delay, with respect to the previous 451 significant mean crossing. 453 A cyclic buffer, last_N_crossings, records a 1 if there is a 454 significant mean crossing, otherwise a 0. 456 The counter, number_of_crossings, is incremented when there is a 457 significant mean crossing and decremented when a non-zero value is 458 removed from the last_N_crossings. 460 This approximation of freq_est was not used in [Hayes-LCN14], which 461 calculated freq_est every T using the current E_N(E_T(OWD)). Our 462 tests show that this approximation of freq_est yields results that 463 are almost identical to when the full calculation is performed every 464 T. 466 3.1.5. Packet loss 468 The proportion of packets lost is used as a supplementary measure: 470 pkt_loss = sum_NT(lost packets) / sum_NT(total packets) 472 Note: When pkt_loss is small it is very variable, however, when 473 pkt_loss is high it becomes a stable measure for making grouping 474 decisions.. 476 3.2. Flow Grouping 478 3.2.1. Flow Grouping Algorithm 480 The following grouping algorithm is RECOMMENDED for SBD in the RMCAT 481 context and is sufficient and efficient for small to moderate numbers 482 of flows. For very large numbers of flows (e.g. hundreds), a more 483 complex clustering algorithm may be substituted. 485 Since no single metric is precise enough to group flows (due to 486 noise), the algorithm uses multiple metrics. Each metric offers a 487 different "view" of the bottleneck link characteristics, and used 488 together they enable a more precise grouping of flows than would 489 otherwise be possible. 491 Flows determined to be experiencing congestion are successively 492 divided into groups based on freq_est, var_est, and skew_est. 494 The first step is to determine which flows are experiencing 495 congestion. This is important, since if a flow is not experiencing 496 congestion its delay based metrics will not describe the bottleneck, 497 but the "noise" from the rest of the path. Skewness, with proportion 498 of packets loss as a supplementary measure, is used to do this: 500 1. Grouping will be performed on flows where: 502 skew_est < c_s 504 || ( skew_est < c_h && PC ) 506 || pkt_loss > p_l 508 The parameter c_s controls how sensitive the mechanism is in 509 detecting congestion. C_s = 0.0 was used in [Hayes-LCN14]. A value 510 of c_s = 0.05 is a little more sensitive, and c_s = -0.05 is a little 511 less sensitive. C_h controls the hysteresis on flows that were 512 grouped as experiencing congestion last time. 514 These flows, flows experiencing congestion, are then progressively 515 divided into groups based on the freq_est, PDV, and skew_est summary 516 statistics. The process proceeds according to the following steps: 518 2. Group flows whose difference in sorted freq_est is less than a 519 threshold: 521 diff(freq_est) < p_f 523 3. Group flows whose difference in sorted E_N(PDV) (highest to 524 lowest) is less than a threshold: 526 diff(var_est) < (p_pdv * var_est) 528 The threshold, (p_pdv * var_est), is with respect to the highest 529 value in the difference. 531 4. Group flows whose difference in sorted skew_est or pkt_loss is 532 less than a threshold: 534 if pkt_loss < p_l 536 diff(skew_est) < p_s 538 otherwise 540 diff(pkt_loss) < (p_d * pkt_loss) 542 The threshold, (p_d * pkt_loss), is with respect to the 543 highest value in the difference. 545 This procedure involves sorting estimates from highest to lowest. It 546 is simple to implement, and efficient for small numbers of flows (up 547 to 10-20). 549 3.2.2. Using the flow group signal 551 A grouping decisions is made every T from the second T, though they 552 will not attain their full design accuracy until after the N'th T 553 interval. 555 Network conditions, and even the congestion controllers, can cause 556 bottlenecks to fluctuate. A coupled congestion controller MAY decide 557 only to couple groups that remain stable, say grouped together 90% of 558 the time, depending on its objectives. Recommendations concerning 559 this are beyond the scope of this draft and will be specific to the 560 coupled congestion controllers objectives. 562 3.3. Removing Noise from the Estimates 564 The following describe small changes to the calculation of the key 565 metrics that help remove noise from them. Currently these "tweaks" 566 are described separately to keep the main description succinct. In 567 future revisions of the draft these enhancements may replace the 568 original key metric calculations. 570 3.3.1. PDV noise 572 Usually during congestion the max_T(OWD) is quite well sampled as the 573 delay distribution is skewed toward the maximum. However max_T(OWD) 574 is subject to delay noise from other queues along the path as well as 575 the host operating system. Min_T(OWD) is less prone to noise along 576 the path and from the host operating system, but is not well sampled 577 during congestion (i.e. when there is a bottleneck). Flows with very 578 different packet send rates exacerbate the problem. 580 An alternative delay variation measure that is less sensitive to 581 extreme values and different send rates is Mean Absolute Deviation 582 (MAD). It can be implemented in an online manner as follows: 584 var_base_T = sum_T(|OWD - E_T(OWD)|) 586 where 588 |x| is the absolute value of x 590 E_T(OWD) is the mean OWD calculated in the previous T 592 var_est = MAD_MT = sum_MT(var_base_T)/num_MT(OWD) 594 For calculation of freq_est p_v=0.7 (MAD is a smaller number than 595 PDV) 597 For the grouping threshold p_mad=0.1 instead of p_pdv (MAD is less 598 noisy so the test can be tighter) 600 Note that the method for improving responsiveness of MAD_MT is the 601 same as that described in Section 3.4.1 for skew_est. 603 3.3.2. Oscillation noise 605 When a path has no congestion, var_est will be very small and the 606 recorded significant mean crossings will be the result of path noise. 607 Thus up to N-1 meaningless mean crossings can be a source of error at 608 the point a link becomes a bottleneck and flows traversing it begin 609 to be grouped. 611 To remove this source of noise from freq_est: 613 1. Set the current PDV to PDV = NaN (a value representing an invalid 614 record, i.e. Not a Number) for flows that are deemed to not be 615 experiencing congestion by the first skew_est based grouping test 616 (see Section 3.2.1). 618 2. Then var_est = sum_M(PDV != NaN) / num_VM(PDV) 620 3. For freq_est, only record a significant mean crossing if flow is 621 experiencing congestion. 623 These three changes will remove the non-congestion noise from 624 freq_est. A similar adjustment can be made for MAD based var_est. 626 3.3.3. Clock skew 628 Generally sender and receiver clock skew will be too small to cause 629 significant errors in the estimators. Skew_est is most sensitive to 630 this type of noise. In circumstances where clock skew is high, 631 making M < N can reduce this error. 633 A better method is to estimate the effect the clock skew is having on 634 the summary statistics, and then adjust statistics accordingly. A 635 simple online method of doing this based on min_T(OWD) will be 636 described here in a subsequent version of the draft. 638 3.4. Reducing lag and Improving Responsiveness 640 Measurement based shared bottleneck detection makes decisions in the 641 present based on what has been measured in the past. This means that 642 there is always a lag in responding to changing conditions. This 643 mechanism is based on summary statistics taken over (N*T) seconds. 644 This mechanism can be made more responsive to changing conditions by: 646 1. Reducing N and/or M -- but at the expense of having less accurate 647 metrics, and/or 649 2. Exploiting the fact that more recent measurements are more 650 valuable than older measurements and weighting them accordingly. 652 Although more recent measurements are more valuable, older 653 measurements are still needed to gain an accurate estimate of the 654 distribution descriptor we are measuring. Unfortunately, the simple 655 exponentially weighted moving average weights drop off too quickly 656 for our requirements and have an infinite tail. A simple linearly 657 declining weighted moving average also does not provide enough weight 658 to the most recent measurements. We propose a piecewise linear 659 distribution of weights, such that the first section (samples 1:F) is 660 flat as in a simple moving average, and the second section (samples 661 F+1:M) is linearly declining weights to the end of the averaging 662 window. We choose integer weights, which allows incremental 663 calculation without introducing rounding errors. 665 3.4.1. Improving the response of the skewness estimate 667 The weighted moving average for skew_est, based on skew_est in 668 Section 3.1.2, can be calculated as follows: 670 skew_est = ((M-F+1)*sum(skew_base_T(1:F)) 672 + sum([(M-F):1].*skew_base_T(F+1:M))) 674 / ((M-F+1)*sum(numsampT(1:F)) 676 + sum([(M-F):1].*numsampT(F+1:M))) 678 where numsampT is an array of the number of OWD samples in each T 679 (i.e. num_T(OWD)), and numsampT(1) is the most recent; skew_base_T(1) 680 is the most recent calculation of skew_base_T; 1:F refers to the 681 integer values 1 through to F, and [(M-F):1] refers to an array of 682 the integer values (M-F) declining through to 1; and ".*" is the 683 array scalar dot product operator. 685 3.4.2. Improving the response of the variability estimate 687 The weighted moving average for var_est can be calculated as follows: 689 var_est = ((M-F+1)*sum(PDV(1:F)) + sum([(M-F):1].*PDV(F+1:M))) 691 / (F*(M-F+1) + sum([(M-F):1]) 693 where 1:F refers to the integer values 1 through to F, and [(M-F):1] 694 refers to an array of the integer values (M-F) declining through to 695 1; and ".*" is the array scalar dot product operator. When removing 696 oscillation noise (see Section 3.3.2) this calculation must be 697 adjusted to allow for invalid PDV records. 699 4. Measuring OWD 701 This section discusses the OWD measurements required for this 702 algorithm to detect shared bottlenecks. 704 The SBD mechanism described in this draft relies on differences 705 between OWD measurements to avoid the practical problems with 706 measuring absolute OWD (see [Hayes-LCN14] section IIIC). Since all 707 summary statistics are relative to the mean OWD and sender/receiver 708 clock offsets should be approximately constant over the measurement 709 periods, the offset is subtracted out in the calculation. 711 4.1. Time stamp resolution 713 The SBD mechanism requires timing information precise enough to be 714 able to make comparisons. As a rule of thumb, the time resolution 715 should be less than one hundredth of a typical path's range of 716 delays. In general, the lower the time resolution, the more care 717 that needs to be taken to ensure rounding errors do not bias the 718 skewness calculation. 720 Typical RTP media flows use sub-millisecond timers, which should be 721 adequate in most situations. 723 5. Acknowledgements 725 This work was part-funded by the European Community under its Seventh 726 Framework Programme through the Reducing Internet Transport Latency 727 (RITE) project (ICT-317700). The views expressed are solely those of 728 the authors. 730 6. IANA Considerations 732 This memo includes no request to IANA. 734 7. Security Considerations 736 The security considerations of RFC 3550 [RFC3550], RFC 4585 737 [RFC4585], and RFC 5124 [RFC5124] are expected to apply. 739 Non-authenticated RTCP packets carrying shared bottleneck indications 740 and summary statistics could allow attackers to alter the bottleneck 741 sharing characteristics for private gain or disruption of other 742 parties communication. 744 8. Change history 746 Changes made to this document: 748 WG-00->WG-01 : Moved unbiased skew section to replace skew 749 estimate, more robust variability estimator, the 750 term variance replaced with variability, clock 751 drift term corrected to clock skew, revision to 752 clock skew section with a place holder, description 753 of parameters. 755 02->WG-00 : Fixed missing 0.5 in 3.3.2 and missing brace in 756 3.3.3 758 01->02 : New section describing improvements to the key 759 metric calculations that help to remove noise, 760 bias, and reduce lag. Some revisions to the 761 notation to make it clearer. Some tightening of 762 the thresholds. 764 00->01 : Revisions to terminology for clarity 766 9. References 768 9.1. Normative References 770 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 771 Requirement Levels", BCP 14, RFC 2119, March 1997. 773 9.2. Informative References 775 [Hayes-LCN14] 776 Hayes, D., Ferlin, S., and M. Welzl, "Practical Passive 777 Shared Bottleneck Detection using Shape Summary 778 Statistics", Proc. the IEEE Local Computer Networks 779 (LCN) p150-158, September 2014, . 783 [I-D.welzl-rmcat-coupled-cc] 784 Welzl, M., Islam, S., and S. Gjessing, "Coupled congestion 785 control for RTP media", draft-welzl-rmcat-coupled-cc-04 786 (work in progress), October 2014. 788 [ITU-Y1540] 789 ITU-T, "Internet Protocol Data Communication Service - IP 790 Packet Transfer and Availability Performance Parameters", 791 Series Y: Global Information Infrastructure, Internet 792 Protocol Aspects and Next-Generation Networks , 793 March 2011, 794 . 796 [RFC3550] Schulzrinne, H., Casner, S., Frederick, R., and V. 797 Jacobson, "RTP: A Transport Protocol for Real-Time 798 Applications", STD 64, RFC 3550, July 2003. 800 [RFC4585] Ott, J., Wenger, S., Sato, N., Burmeister, C., and J. Rey, 801 "Extended RTP Profile for Real-time Transport Control 802 Protocol (RTCP)-Based Feedback (RTP/AVPF)", RFC 4585, 803 July 2006. 805 [RFC5124] Ott, J. and E. Carrara, "Extended Secure RTP Profile for 806 Real-time Transport Control Protocol (RTCP)-Based Feedback 807 (RTP/SAVPF)", RFC 5124, February 2008. 809 [RFC5481] Morton, A. and B. Claise, "Packet Delay Variation 810 Applicability Statement", RFC 5481, March 2009. 812 [RFC6817] Shalunov, S., Hazel, G., Iyengar, J., and M. Kuehlewind, 813 "Low Extra Delay Background Transport (LEDBAT)", RFC 6817, 814 December 2012. 816 Authors' Addresses 818 David Hayes (editor) 819 University of Oslo 820 PO Box 1080 Blindern 821 Oslo, N-0316 822 Norway 824 Phone: +47 2284 5566 825 Email: davihay@ifi.uio.no 827 Simone Ferlin 828 Simula Research Laboratory 829 P.O.Box 134 830 Lysaker, 1325 831 Norway 833 Phone: +47 4072 0702 834 Email: ferlin@simula.no 835 Michael Welzl 836 University of Oslo 837 PO Box 1080 Blindern 838 Oslo, N-0316 839 Norway 841 Phone: +47 2285 2420 842 Email: michawe@ifi.uio.no