idnits 2.17.1 draft-ietf-rmcat-sbd-08.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 3, 2017) is 2461 days in the past. Is this intentional? Checking references for intended status: Experimental ---------------------------------------------------------------------------- == Outdated reference: A later version (-04) exists of draft-dt-rmcat-feedback-message-02 == Outdated reference: A later version (-09) exists of draft-ietf-rmcat-coupled-cc-06 Summary: 0 errors (**), 0 flaws (~~), 3 warnings (==), 1 comment (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 RTP Media Congestion Avoidance Techniques D. Hayes, Ed. 3 Internet-Draft S. Ferlin 4 Intended status: Experimental Simula Research Laboratory 5 Expires: January 4, 2018 M. Welzl 6 K. Hiorth 7 University of Oslo 8 July 3, 2017 10 Shared Bottleneck Detection for Coupled Congestion Control for RTP 11 Media. 12 draft-ietf-rmcat-sbd-08 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 based on continuous measurements and used as 19 input to a grouping algorithm that runs wherever the knowledge is 20 needed. This mechanism complements the coupled congestion control 21 mechanism in draft-ietf-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 4, 2018. 40 Copyright Notice 42 Copyright (c) 2017 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 basic mechanism . . . . . . . . . . . . . . . . . . . 3 59 1.2. The signals . . . . . . . . . . . . . . . . . . . . . . . 3 60 1.2.1. Packet loss . . . . . . . . . . . . . . . . . . . . . 3 61 1.2.2. Packet delay . . . . . . . . . . . . . . . . . . . . 4 62 1.2.3. Path lag . . . . . . . . . . . . . . . . . . . . . . 4 63 2. Definitions . . . . . . . . . . . . . . . . . . . . . . . . . 4 64 2.1. Parameters and their effect . . . . . . . . . . . . . . . 6 65 2.2. Recommended parameter values . . . . . . . . . . . . . . 7 66 3. Mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . 7 67 3.1. SBD feedback requirements . . . . . . . . . . . . . . . . 8 68 3.1.1. Feedback when all the logic is placed at the sender . 9 69 3.1.2. Feedback when the statistics are calculated at the 70 receiver and SBD performed at the sender . . . . . . 9 71 3.1.3. Feedback when bottlenecks can be determined at both 72 senders and receivers . . . . . . . . . . . . . . . . 10 73 3.2. Key metrics and their calculation . . . . . . . . . . . . 10 74 3.2.1. Mean delay . . . . . . . . . . . . . . . . . . . . . 10 75 3.2.2. Skewness estimate . . . . . . . . . . . . . . . . . . 11 76 3.2.3. Variability estimate . . . . . . . . . . . . . . . . 12 77 3.2.4. Oscillation estimate . . . . . . . . . . . . . . . . 12 78 3.2.5. Packet loss . . . . . . . . . . . . . . . . . . . . . 13 79 3.3. Flow Grouping . . . . . . . . . . . . . . . . . . . . . . 13 80 3.3.1. Flow grouping algorithm . . . . . . . . . . . . . . . 13 81 3.3.2. Using the flow group signal . . . . . . . . . . . . . 16 82 4. Enhancements to the basic SBD algorithm . . . . . . . . . . . 16 83 4.1. Reducing lag and improving responsiveness . . . . . . . . 16 84 4.1.1. Improving the response of the skewness estimate . . . 17 85 4.1.2. Improving the response of the variability estimate . 19 86 4.2. Removing oscillation noise . . . . . . . . . . . . . . . 19 87 5. Measuring OWD . . . . . . . . . . . . . . . . . . . . . . . . 20 88 5.1. Time stamp resolution . . . . . . . . . . . . . . . . . . 20 89 5.2. Clock skew . . . . . . . . . . . . . . . . . . . . . . . 20 90 6. Expected feedback from experiments . . . . . . . . . . . . . 20 91 7. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . 21 92 8. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 21 93 9. Security Considerations . . . . . . . . . . . . . . . . . . . 21 94 10. Change history . . . . . . . . . . . . . . . . . . . . . . . 21 95 11. References . . . . . . . . . . . . . . . . . . . . . . . . . 22 96 11.1. Normative References . . . . . . . . . . . . . . . . . . 22 97 11.2. Informative References . . . . . . . . . . . . . . . . . 23 98 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 24 100 1. Introduction 102 In the Internet, it is not normally known if flows (e.g., TCP 103 connections or UDP data streams) traverse the same bottlenecks. Even 104 flows that have the same sender and receiver may take different paths 105 and may or may not share a bottleneck. Flows that share a bottleneck 106 link usually compete with one another for their share of the 107 capacity. This competition has the potential to increase packet loss 108 and delays. This is especially relevant for interactive applications 109 that communicate simultaneously with multiple peers (such as multi- 110 party video). For RTP media applications such as RTCWEB, 111 [I-D.ietf-rmcat-coupled-cc] describes a scheme that combines the 112 congestion controllers of flows in order to honor their priorities 113 and avoid unnecessary packet loss as well as delay. This mechanism 114 relies on some form of Shared Bottleneck Detection (SBD); here, a 115 measurement-based SBD approach is described. 117 1.1. The basic mechanism 119 The mechanism groups flows that have similar statistical 120 characteristics together. Section 3.3.1 describes a simple method 121 for achieving this, however, a major part of this draft is concerned 122 with collecting suitable statistics for this purpose. 124 1.2. The signals 126 The current Internet is unable to explicitly inform endpoints as to 127 which flows share bottlenecks, so endpoints need to infer this from 128 whatever information is available to them. The mechanism described 129 here currently utilizes packet loss and packet delay, but is not 130 restricted to these. As ECN becomes more prevalent it too will 131 become a valuable base signal. 133 1.2.1. Packet loss 135 Packet loss is often a relatively rare signal. Therefore, on its own 136 it is of limited use for SBD, however, it is a valuable supplementary 137 measure when it is more prevalent. 139 1.2.2. Packet delay 141 End-to-end delay measurements include noise from every device along 142 the path in addition to the delay perturbation at the bottleneck 143 device. The noise is often significantly increased if the round-trip 144 time is used. The cleanest signal is obtained by using One-Way-Delay 145 (OWD). 147 Measuring absolute OWD is difficult since it requires both the sender 148 and receiver clocks to be synchronized. However, since the 149 statistics being collected are relative to the mean OWD, a relative 150 OWD measurement is sufficient. Clock skew is not usually significant 151 over the time intervals used by this SBD mechanism (see [RFC6817] A.2 152 for a discussion on clock skew and OWD measurements). However, in 153 circumstances where it is significant, Section 5.2 outlines a way of 154 adjusting the calculations to cater for it. 156 Each packet arriving at the bottleneck buffer may experience very 157 different queue lengths, and therefore different waiting times. A 158 single OWD sample does not, therefore, characterize the path well. 159 However, multiple OWD measurements do reflect the distribution of 160 delays experienced at the bottleneck. 162 1.2.3. Path lag 164 Flows that share a common bottleneck may traverse different paths, 165 and these paths will often have different base delays. This makes it 166 difficult to correlate changes in delay or loss. This technique uses 167 the long term shape of the delay distribution as a base for 168 comparison to counter this. 170 2. Definitions 172 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 173 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 174 document are to be interpreted as described in RFC 2119 [RFC2119]. 176 Acronyms used in this document: 178 OWD -- One Way Delay 180 MAD -- Mean Absolute Deviation 182 RTT -- Round Trip Time 184 SBD -- Shared Bottleneck Detection 186 Conventions used in this document: 188 T -- the base time interval over which measurements are 189 made. 191 N -- the number of base time, T, intervals used in some 192 calculations. 194 M -- the number of base time, T, intervals used in some 195 calculations. 197 sum_T(...) -- summation of all the measurements of the variable 198 in parentheses taken over the interval T 200 sum(...) -- summation of terms of the variable in parentheses 202 sum_N(...) -- summation of N terms of the variable in parentheses 204 sum_NT(...) -- summation of all measurements taken over the 205 interval N*T 207 sum_MT(...) -- summation of all measurements taken over the 208 interval M*T 210 E_T(...) -- the expectation or mean of the measurements of the 211 variable in parentheses over T 213 E_N(...) -- the expectation or mean of the last N values of the 214 variable in parentheses 216 E_M(...) -- the expectation or mean of the last M values of the 217 variable in parentheses, where M <= N. 219 num_T(...) -- the count of measurements of the variable in 220 parentheses taken in the interval T 222 num_VM(...) -- the count of valid values of the variable in 223 parentheses given M records 225 PB -- a boolean variable indicating the particular flow 226 was identified transiting a bottleneck in the 227 previous interval T (i.e. Previously Bottleneck) 229 skew_est -- a measure of skewness in a OWD distribution. 231 skew_base_T -- a variable used as an intermediate step in 232 calculating skew_est. 234 var_est -- a measure of variability in OWD measurements. 236 var_base_T -- a variable used as an intermediate step in 237 calculating var_est. 239 freq_est -- a measure of low frequency oscillation in the OWD 240 measurements. 242 p_l, p_f, p_mad, c_s, c_h, p_s, p_d, p_v -- various thresholds 243 used in the mechanism 245 M and F -- number of values related to N 247 2.1. Parameters and their effect 249 T T should be long enough so that there are enough packets 250 received during T for a useful estimate of short term mean 251 OWD and variation statistics. Making T too large can limit 252 the efficacy of freq_est. It will also increase the response 253 time of the mechanism. Making T too small will make the 254 metrics noisier. 256 N & M N should be large enough to provide a stable estimate of 257 oscillations in OWD. Usually M=N, though having M mean_delay) skew_base_T-- 492 The mean_delay does not include the mean of the current T interval to 493 enable it to be calculated iteratively. 495 skew_est = sum_MT(skew_base_T)/num_MT(OWD) 497 where skew_est is a number between -1 and 1 499 Note: Care must be taken when implementing the comparisons to ensure 500 that rounding does not bias skew_est. It is important that the mean 501 is calculated with a higher precision than the samples. 503 3.2.3. Variability estimate 505 Mean Absolute Deviation (MAD) delay is a robust variability measure 506 that copes well with different send rates. It can be implemented in 507 an online manner as follows: 509 var_base_T = sum_T(|OWD - E_T(OWD)|) 511 where 513 |x| is the absolute value of x 515 E_T(OWD) is the mean OWD calculated in the previous T 517 var_est = MAD_MT = sum_MT(var_base_T)/num_MT(OWD) 519 For calculation of freq_est p_v=0.7 521 For the grouping threshold p_mad=0.1 523 3.2.4. Oscillation estimate 525 An estimate of the low frequency oscillation of the delay signal is 526 calculated by counting and normalizing the significant mean, 527 E_T(OWD), crossings of mean_delay: 529 freq_est = number_of_crossings / N 531 where we define a significant mean crossing as a crossing that 532 extends p_v * var_est from mean_delay. In our experiments we 533 have found that p_v = 0.7 is a good value. 535 Freq_est is a number between 0 and 1. Freq_est can be approximated 536 incrementally as follows: 538 With each new calculation of E_T(OWD) a decision is made as to 539 whether this value of E_T(OWD) significantly crosses the current 540 long term mean, mean_delay, with respect to the previous 541 significant mean crossing. 543 A cyclic buffer, last_N_crossings, records a 1 if there is a 544 significant mean crossing, otherwise a 0. 546 The counter, number_of_crossings, is incremented when there is a 547 significant mean crossing and decremented when a non-zero value is 548 removed from the last_N_crossings. 550 This approximation of freq_est was not used in [Hayes-LCN14], which 551 calculated freq_est every T using the current E_N(E_T(OWD)). Our 552 tests show that this approximation of freq_est yields results that 553 are almost identical to when the full calculation is performed every 554 T. 556 3.2.5. Packet loss 558 The proportion of packets lost over the period NT is used as a 559 supplementary measure: 561 pkt_loss = sum_NT(lost packets) / sum_NT(total packets) 563 Note: When pkt_loss is small it is very variable, however, when 564 pkt_loss is high it becomes a stable measure for making grouping 565 decisions. 567 3.3. Flow Grouping 569 3.3.1. Flow grouping algorithm 571 The following grouping algorithm is RECOMMENDED for SBD in the RMCAT 572 context and is sufficient and efficient for small to moderate numbers 573 of flows. For very large numbers of flows (e.g. hundreds), a more 574 complex clustering algorithm may be substituted. 576 Since no single metric is precise enough to group flows (due to 577 noise), the algorithm uses multiple metrics. Each metric offers a 578 different "view" of the bottleneck link characteristics, and used 579 together they enable a more precise grouping of flows than would 580 otherwise be possible. 582 Flows determined to be transiting a bottleneck are successively 583 divided into groups based on freq_est, var_est, skew_est and 584 pkt_loss. 586 The first step is to determine which flows are transiting a 587 bottleneck. This is important, since if a flow is not transiting a 588 bottleneck its delay based metrics will not describe the bottleneck, 589 but the "noise" from the rest of the path. Skewness, with proportion 590 of packet loss as a supplementary measure, is used to do this: 592 1. Grouping will be performed on flows that are inferred to be 593 traversing a bottleneck by: 595 skew_est < c_s 597 || ( skew_est < c_h & PB ) || pkt_loss > p_l 599 The parameter c_s controls how sensitive the mechanism is in 600 detecting a bottleneck. C_s = 0.0 was used in [Hayes-LCN14]. A 601 value of c_s = 0.05 is a little more sensitive, and c_s = -0.05 is a 602 little less sensitive. C_h controls the hysteresis on flows that 603 were grouped as transiting a bottleneck last time. If the test 604 result is TRUE, PB=TRUE, otherwise PB=FALSE. 606 These flows, flows transiting a bottleneck, are then progressively 607 divided into groups based on the freq_est, var_est, and skew_est 608 summary statistics. The process proceeds according to the following 609 steps: 611 2. Group flows whose difference in sorted freq_est is less than a 612 threshold: 614 diff(freq_est) < p_f 616 3. Subdivide the groups obtained in 2. by grouping flows whose 617 difference in sorted E_M(var_est) (highest to lowest) is less 618 than a threshold: 620 diff(var_est) < (p_mad * var_est) 622 The threshold, (p_mad * var_est), is with respect to the highest 623 value in the difference. 625 4. Subdivide the groups obtained in 3. by grouping flows whose 626 difference in sorted skew_est is less than a threshold: 628 diff(skew_est) < p_s 630 5. When packet loss is high enough to be reliable (pkt_loss > p_l), 631 Subdivide the groups obtained in 4. by grouping flows whose 632 difference is less than a threshold 634 diff(pkt_loss) < (p_d * pkt_loss) 636 The threshold, (p_d * pkt_loss), is with respect to the highest 637 value in the difference. 639 This procedure involves sorting estimates from highest to lowest. It 640 is simple to implement, and efficient for small numbers of flows (up 641 to 10-20). Figure 2 illustrates this algorithm. 643 ********* 644 * Flows * 645 ***.**.** 646 / ' 647 / '--. 648 / \ 649 .---v--. .----v---. 650 1. Flows traversing | Cong | | UnCong | 651 a bottleneck '-.--.-' '--------' 652 / \ 653 / \ 654 / \ 655 .--v--. v-----. 656 2. Divide by | g_1 | ... | g_n | 657 freq_est '---.-. '----.. 658 / \ / \ 659 / '--. v '------. 660 / \ \ 661 .----v-. .-v----. .---v--. 662 3. Divide by | g_1a | ... | g_1z | ... | g_nz | 663 var_est '---.-.' '-----.. '-.-.--' 664 / \ / \ / | 665 / '-----. v v v | 666 / \ | 667 .-v-----. .-v-----. .---v---. 668 4. Divide by | g_1ai | ... | g_1ax | ... | g_nzx | 669 skew_est '----.-.' '------.. '-.-.---' 670 / \ / \ / | 671 / '--. v v v | 672 / \ | 673 .-----v--. .-v------. .----v---. 674 5. Divide by | g_1aiA | ... | g_1aiZ | ... | g_nzxZ | 675 pkt_loss '--------' '--------' '--------' 676 (when applicable) 678 Simple grouping algorithm. 680 Figure 2 682 3.3.2. Using the flow group signal 684 Grouping decisions can be made every T from the second T, however 685 they will not attain their full design accuracy until after the 686 2*N'th T interval. We recommend that grouping decisions are not made 687 until 2*M T intervals. 689 Network conditions, and even the congestion controllers, can cause 690 bottlenecks to fluctuate. A coupled congestion controller MAY decide 691 only to couple groups that remain stable, say grouped together 90% of 692 the time, depending on its objectives. Recommendations concerning 693 this are beyond the scope of this document and will be specific to 694 the coupled congestion controllers objectives. 696 4. Enhancements to the basic SBD algorithm 698 The SBD algorithm as specified in Section 3 was found to work well 699 for a broad variety of conditions. The following enhancements to the 700 basic mechanisms have been found to significantly improve the 701 algorithm's performance under some circumstances and SHOULD be 702 implemented. These "tweaks" are described separately to keep the 703 main description succinct. 705 4.1. Reducing lag and improving responsiveness 707 This section describes how to improve the responsiveness of the basic 708 algorithm. 710 Measurement based shared bottleneck detection makes decisions in the 711 present based on what has been measured in the past. This means that 712 there is always a lag in responding to changing conditions. This 713 mechanism is based on summary statistics taken over (N*T) seconds. 714 This mechanism can be made more responsive to changing conditions by: 716 1. Reducing N and/or M -- but at the expense of having less accurate 717 metrics, and/or 719 2. Exploiting the fact that more recent measurements are more 720 valuable than older measurements and weighting them accordingly. 722 Although more recent measurements are more valuable, older 723 measurements are still needed to gain an accurate estimate of the 724 distribution descriptor we are measuring. Unfortunately, the simple 725 exponentially weighted moving average weights drop off too quickly 726 for our requirements and have an infinite tail. A simple linearly 727 declining weighted moving average also does not provide enough weight 728 to the most recent measurements. We propose a piecewise linear 729 distribution of weights, such that the first section (samples 1:F) is 730 flat as in a simple moving average, and the second section (samples 731 F+1:M) is linearly declining weights to the end of the averaging 732 window. We choose integer weights, which allows incremental 733 calculation without introducing rounding errors. 735 4.1.1. Improving the response of the skewness estimate 737 The weighted moving average for skew_est, based on skew_est in 738 Section 3.2.2, can be calculated as follows: 740 skew_est = ((M-F+1)*sum(skew_base_T(1:F)) 742 + sum([(M-F):1].*skew_base_T(F+1:M))) 744 / ((M-F+1)*sum(numsampT(1:F)) 746 + sum([(M-F):1].*numsampT(F+1:M))) 748 where numsampT is an array of the number of OWD samples in each T 749 (i.e. num_T(OWD)), and numsampT(1) is the most recent; skew_base_T(1) 750 is the most recent calculation of skew_base_T; 1:F refers to the 751 integer values 1 through to F, and [(M-F):1] refers to an array of 752 the integer values (M-F) declining through to 1; and ".*" is the 753 array scalar dot product operator. 755 To calculate this weighted skew_est incrementally: 757 Notation: F_ - flat portion, D_ - declining portion, W_ - weighted 758 component 760 Initialize: sum_skewbase = 0, F_skewbase=0, W_D_skewbase=0 762 skewbase_hist = buffer length M initialize to 0 764 numsampT = buffer length M initialized to 0 766 Steps per iteration: 768 1. old_skewbase = skewbase_hist(M) 770 2. old_numsampT = numsampT(M) 772 3. cycle(skewbase_hist) 774 4. cycle(numsampT) 776 5. numsampT(1) = num_T(OWD) 778 6. skewbase_hist(1) = skew_base_T 780 7. F_skewbase = F_skewbase + skew_base_T - skewbase_hist(F+1) 782 8. W_D_skewbase = W_D_skewbase + (M-F)*skewbase_hist(F+1) 783 - sum_skewbase 785 9. W_D_numsamp = W_D_numsamp + (M-F)*numsampT(F+1) - sum_numsamp 786 + F_numsamp 788 10. F_numsamp = F_numsamp + numsampT(1) - numsampT(F+1) 790 11. sum_skewbase = sum_skewbase + skewbase_hist(F+1) - old_skewbase 792 12. sum_numsamp = sum_numsamp + numsampT(1) - old_numsampT 794 13. skew_est = ((M-F+1)*F_skewbase + W_D_skewbase) / 795 ((M-F+1)*F_numsamp+W_D_numsamp) 797 Where cycle(....) refers to the operation on a cyclic buffer where 798 the start of the buffer is now the next element in the buffer. 800 4.1.2. Improving the response of the variability estimate 802 Similarly the weighted moving average for var_est can be calculated 803 as follows: 805 var_est = ((M-F+1)*sum(var_base_T(1:F)) 807 + sum([(M-F):1].*var_base_T(F+1:M))) 809 / ((M-F+1)*sum(numsampT(1:F)) 811 + sum([(M-F):1].*numsampT(F+1:M))) 813 where numsampT is an array of the number of OWD samples in each T 814 (i.e. num_T(OWD)), and numsampT(1) is the most recent; skew_base_T(1) 815 is the most recent calculation of skew_base_T; 1:F refers to the 816 integer values 1 through to F, and [(M-F):1] refers to an array of 817 the integer values (M-F) declining through to 1; and ".*" is the 818 array scalar dot product operator. When removing oscillation noise 819 (see Section 4.2) this calculation must be adjusted to allow for 820 invalid var_base_T records. 822 Var_est can be calculated incrementally in the same way as skew_est 823 in Section 4.1.1. However, note that the buffer numsampT is used for 824 both calculations so the operations on it should not be repeated. 826 4.2. Removing oscillation noise 828 When a path has no bottleneck, var_est will be very small and the 829 recorded significant mean crossings will be the result of path noise. 830 Thus up to N-1 meaningless mean crossings can be a source of error at 831 the point a link becomes a bottleneck and flows traversing it begin 832 to be grouped. 834 To remove this source of noise from freq_est: 836 1. Set the current var_base_T = NaN (a value representing an invalid 837 record, i.e. Not a Number) for flows that are deemed to not be 838 transiting a bottleneck by the first skew_est based grouping test 839 (see Section 3.3.1). 841 2. Then var_est = sum_MT(var_base_T != NaN) / num_MT(OWD) 843 3. For freq_est, only record a significant mean crossing if flow 844 deemed to be transiting a bottleneck. 846 These three changes can help to remove the non-bottleneck noise from 847 freq_est. 849 5. Measuring OWD 851 This section discusses the OWD measurements required for this 852 algorithm to detect shared bottlenecks. 854 The SBD mechanism described in this document relies on differences 855 between OWD measurements to avoid the practical problems with 856 measuring absolute OWD (see [Hayes-LCN14] section IIIC). Since all 857 summary statistics are relative to the mean OWD and sender/receiver 858 clock offsets should be approximately constant over the measurement 859 periods, the offset is subtracted out in the calculation. 861 5.1. Time stamp resolution 863 The SBD mechanism requires timing information precise enough to be 864 able to make comparisons. As a rule of thumb, the time resolution 865 should be less than one hundredth of a typical path's range of 866 delays. In general, the coarser the time resolution, the more care 867 that needs to be taken to ensure rounding errors do not bias the 868 skewness calculation. Timing information described by 869 [I-D.dt-rmcat-feedback-message] should be sufficient for the sender 870 to calculate relative OWD. 872 5.2. Clock skew 874 Generally sender and receiver clock skew will be too small to cause 875 significant errors in the estimators. Skew_est and freq_est are the 876 most sensitive to this type of noise due to their use of a mean OWD 877 calculated over a longer interval. In circumstances where clock skew 878 is high, basing skew_est only on the previous T's mean and ignoring 879 freq_est provides a noisier but reliable signal. 881 A more sophisticated method is to estimate the effect the clock skew 882 is having on the summary statistics, and then adjust statistics 883 accordingly. There are a number of techniques in the literature, 884 including [Zhang-Infocom02]. 886 6. Expected feedback from experiments 888 The algorithm described in this memo has so far been evaluated using 889 simulations. Real network tests using the proposed congestion 890 control algorithms will help confirm the default parameter choice. 891 For example, the time interval T may need to be made longer if the 892 packet rate is very low. Implementers and testers are invited to 893 document their findings in an Internet draft. 895 7. Acknowledgments 897 This work was part-funded by the European Community under its Seventh 898 Framework Programme through the Reducing Internet Transport Latency 899 (RITE) project (ICT-317700). The views expressed are solely those of 900 the authors. 902 8. IANA Considerations 904 This memo includes no request to IANA. 906 9. Security Considerations 908 The security considerations of RFC 3550 [RFC3550], RFC 4585 909 [RFC4585], and RFC 5124 [RFC5124] are expected to apply. 911 Non-authenticated RTCP packets carrying OWD measurements, shared 912 bottleneck indications, and/or summary statistics could allow 913 attackers to alter the bottleneck sharing characteristics for private 914 gain or disruption of other parties communication. 916 10. Change history 918 Changes made to this document: 920 WG-07->WG-08 : Updates addressing https://www.ietf.org/mail- 921 archive/web/rmcat/current/msg01671.html Mainly 922 clarifications. 924 WG-06->WG-07 : Updates addressing 925 https://mailarchive.ietf.org/arch/msg/ 926 rmcat/80B6q4nI7carGcf_ddBwx7nKvOw. Mainly 927 clarifications. Figure 2 to supplement grouping 928 algorithm description. 930 WG-05->WG-06 : Updates addressing WG reviews 931 https://mailarchive.ietf.org/arch/msg/rmcat/- 932 1JdrTMq1Y5T6ZNlOkrQJQ27TzE and 933 https://mailarchive.ietf.org/arch/msg/rmcat/ 934 eI2Q1f8NL2SxbJgjFLR4_rEmJ_g. This has mainly 935 involved minor clarifications, including the moving 936 of 3.4.1 and 3.5 into the new Section 4, and 3.4.1 937 into Section 5 939 WG-04->WG-05 : Fix ToC formatting. Add section on expected 940 feedback from experiments replacing short section 941 on implementation status. Added comment on ECN as 942 a signal. Clarification of lost packet signaling. 944 Change term "draft" to "document" where 945 appropriate. American spelling. Some tightening 946 of the text. 948 WG-03->WG-04 : Add M to terminology table, suggest skew_est based 949 on previous T and no freq_est in clock skew 950 section, feedback requirements as a separate sub 951 section. 953 WG-02->WG-03 : Correct misspelled author 955 WG-01->WG-02 : Removed ambiguity associated with the term 956 "congestion". Expanded the description of 957 initialization messages. Removed PDV metric. 958 Added description of incremental weighted metric 959 calculations for skew_est. Various clarifications 960 based on implementation work. Fixed typos and 961 tuned parameters. 963 WG-00->WG-01 : Moved unbiased skew section to replace skew 964 estimate, more robust variability estimator, the 965 term variance replaced with variability, clock 966 drift term corrected to clock skew, revision to 967 clock skew section with a place holder, description 968 of parameters. 970 02->WG-00 : Fixed missing 0.5 in 3.3.2 and missing brace in 971 3.3.3 973 01->02 : New section describing improvements to the key 974 metric calculations that help to remove noise, 975 bias, and reduce lag. Some revisions to the 976 notation to make it clearer. Some tightening of 977 the thresholds. 979 00->01 : Revisions to terminology for clarity 981 11. References 983 11.1. Normative References 985 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 986 Requirement Levels", BCP 14, RFC 2119, 987 DOI 10.17487/RFC2119, March 1997, 988 . 990 11.2. Informative References 992 [Hayes-LCN14] 993 Hayes, D., Ferlin, S., and M. Welzl, "Practical Passive 994 Shared Bottleneck Detection using Shape Summary 995 Statistics", Proc. the IEEE Local Computer Networks 996 (LCN) pp150-158, September 2014, 997 . 1000 [I-D.dt-rmcat-feedback-message] 1001 Sarker, Z., Perkins, C., Singh, V., and M. Ramalho, "RTP 1002 Control Protocol (RTCP) Feedback for Congestion Control", 1003 draft-dt-rmcat-feedback-message-02 (work in progress), May 1004 2017. 1006 [I-D.ietf-rmcat-coupled-cc] 1007 Islam, S., Welzl, M., and S. Gjessing, "Coupled congestion 1008 control for RTP media", draft-ietf-rmcat-coupled-cc-06 1009 (work in progress), March 2017. 1011 [RFC3550] Schulzrinne, H., Casner, S., Frederick, R., and V. 1012 Jacobson, "RTP: A Transport Protocol for Real-Time 1013 Applications", STD 64, RFC 3550, DOI 10.17487/RFC3550, 1014 July 2003, . 1016 [RFC4585] Ott, J., Wenger, S., Sato, N., Burmeister, C., and J. Rey, 1017 "Extended RTP Profile for Real-time Transport Control 1018 Protocol (RTCP)-Based Feedback (RTP/AVPF)", RFC 4585, 1019 DOI 10.17487/RFC4585, July 2006, 1020 . 1022 [RFC5124] Ott, J. and E. Carrara, "Extended Secure RTP Profile for 1023 Real-time Transport Control Protocol (RTCP)-Based Feedback 1024 (RTP/SAVPF)", RFC 5124, DOI 10.17487/RFC5124, February 1025 2008, . 1027 [RFC6817] Shalunov, S., Hazel, G., Iyengar, J., and M. Kuehlewind, 1028 "Low Extra Delay Background Transport (LEDBAT)", RFC 6817, 1029 DOI 10.17487/RFC6817, December 2012, 1030 . 1032 [Zhang-Infocom02] 1033 Zhang, L., Liu, Z., and H. Xia, "Clock synchronization 1034 algorithms for network measurements", Proc. the IEEE 1035 International Conference on Computer Communications 1036 (INFOCOM) pp160-169, September 2002, 1037 . 1039 Authors' Addresses 1041 David Hayes (editor) 1042 Simula Research Laboratory 1043 P.O. Box 134 1044 Lysaker 1325 1045 Norway 1047 Phone: +47 2284 5566 1048 Email: davidh@simula.no 1050 Simone Ferlin 1051 Simula Research Laboratory 1052 P.O.Box 134 1053 Lysaker 1325 1054 Norway 1056 Phone: +47 4072 0702 1057 Email: ferlin@simula.no 1059 Michael Welzl 1060 University of Oslo 1061 PO Box 1080 Blindern 1062 Oslo N-0316 1063 Norway 1065 Phone: +47 2285 2420 1066 Email: michawe@ifi.uio.no 1068 Kristian Hiorth 1069 University of Oslo 1070 PO Box 1080 Blindern 1071 Oslo N-0316 1072 Norway 1074 Email: kristahi@ifi.uio.no