idnits 2.17.1 draft-ietf-rmcat-nada-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 == Line 461 has weird spacing: '... delta x_o...' -- The document date (July 2, 2018) is 2096 days in the past. Is this intentional? Checking references for intended status: Experimental ---------------------------------------------------------------------------- == Outdated reference: A later version (-10) exists of draft-ietf-rmcat-eval-test-06 == Outdated reference: A later version (-07) exists of draft-ietf-rmcat-video-traffic-model-04 == Outdated reference: A later version (-11) exists of draft-ietf-rmcat-wireless-tests-04 Summary: 0 errors (**), 0 flaws (~~), 5 warnings (==), 1 comment (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 Network Working Group X. Zhu 3 Internet-Draft R. Pan 4 Intended status: Experimental M. Ramalho 5 Expires: January 3, 2019 S. Mena 6 P. Jones 7 J. Fu 8 Cisco Systems 9 S. D'Aronco 10 EPFL 11 July 2, 2018 13 NADA: A Unified Congestion Control Scheme for Real-Time Media 14 draft-ietf-rmcat-nada-08 16 Abstract 18 This document describes NADA (network-assisted dynamic adaptation), a 19 novel congestion control scheme for interactive real-time media 20 applications, such as video conferencing. In the proposed scheme, 21 the sender regulates its sending rate based on either implicit or 22 explicit congestion signaling, in a unified approach. The scheme can 23 benefit from explicit congestion notification (ECN) markings from 24 network nodes. It also maintains consistent sender behavior in the 25 absence of such markings, by reacting to queuing delays and packet 26 losses instead. 28 Status of This Memo 30 This Internet-Draft is submitted in full conformance with the 31 provisions of BCP 78 and BCP 79. 33 Internet-Drafts are working documents of the Internet Engineering 34 Task Force (IETF). Note that other groups may also distribute 35 working documents as Internet-Drafts. The list of current Internet- 36 Drafts is at https://datatracker.ietf.org/drafts/current/. 38 Internet-Drafts are draft documents valid for a maximum of six months 39 and may be updated, replaced, or obsoleted by other documents at any 40 time. It is inappropriate to use Internet-Drafts as reference 41 material or to cite them other than as "work in progress." 43 This Internet-Draft will expire on January 3, 2019. 45 Copyright Notice 47 Copyright (c) 2018 IETF Trust and the persons identified as the 48 document authors. All rights reserved. 50 This document is subject to BCP 78 and the IETF Trust's Legal 51 Provisions Relating to IETF Documents 52 (https://trustee.ietf.org/license-info) in effect on the date of 53 publication of this document. Please review these documents 54 carefully, as they describe your rights and restrictions with respect 55 to this document. Code Components extracted from this document must 56 include Simplified BSD License text as described in Section 4.e of 57 the Trust Legal Provisions and are provided without warranty as 58 described in the Simplified BSD License. 60 Table of Contents 62 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 3 63 2. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . 3 64 3. System Overview . . . . . . . . . . . . . . . . . . . . . . . 3 65 4. Core Congestion Control Algorithm . . . . . . . . . . . . . . 5 66 4.1. Mathematical Notations . . . . . . . . . . . . . . . . . 5 67 4.2. Receiver-Side Algorithm . . . . . . . . . . . . . . . . . 8 68 4.3. Sender-Side Algorithm . . . . . . . . . . . . . . . . . . 10 69 5. Practical Implementation of NADA . . . . . . . . . . . . . . 12 70 5.1. Receiver-Side Operation . . . . . . . . . . . . . . . . . 12 71 5.1.1. Estimation of one-way delay and queuing delay . . . . 12 72 5.1.2. Estimation of packet loss/marking ratio . . . . . . . 13 73 5.1.3. Estimation of receiving rate . . . . . . . . . . . . 13 74 5.2. Sender-Side Operation . . . . . . . . . . . . . . . . . . 13 75 5.2.1. Rate shaping buffer . . . . . . . . . . . . . . . . . 14 76 5.2.2. Adjusting video target rate and sending rate . . . . 15 77 5.3. Feedback Message Requirements . . . . . . . . . . . . . . 15 78 6. Discussions and Further Investigations . . . . . . . . . . . 16 79 6.1. Choice of delay metrics . . . . . . . . . . . . . . . . . 16 80 6.2. Method for delay, loss, and marking ratio estimation . . 16 81 6.3. Impact of parameter values . . . . . . . . . . . . . . . 17 82 6.4. Sender-based vs. receiver-based calculation . . . . . . . 18 83 6.5. Incremental deployment . . . . . . . . . . . . . . . . . 18 84 7. Implementation Status . . . . . . . . . . . . . . . . . . . . 19 85 8. Suggested Experiments . . . . . . . . . . . . . . . . . . . . 19 86 9. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 20 87 10. Security Considerations . . . . . . . . . . . . . . . . . . . 20 88 11. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 20 89 12. References . . . . . . . . . . . . . . . . . . . . . . . . . 20 90 12.1. Normative References . . . . . . . . . . . . . . . . . . 20 91 12.2. Informative References . . . . . . . . . . . . . . . . . 21 92 Appendix A. Network Node Operations . . . . . . . . . . . . . . 23 93 A.1. Default behavior of drop tail queues . . . . . . . . . . 23 94 A.2. RED-based ECN marking . . . . . . . . . . . . . . . . . . 23 95 A.3. Random Early Marking with Virtual Queues . . . . . . . . 24 96 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 25 98 1. Introduction 100 Interactive real-time media applications introduce a unique set of 101 challenges for congestion control. Unlike TCP, the mechanism used 102 for real-time media needs to adapt quickly to instantaneous bandwidth 103 changes, accommodate fluctuations in the output of video encoder rate 104 control, and cause low queuing delay over the network. An ideal 105 scheme should also make effective use of all types of congestion 106 signals, including packet loss, queuing delay, and explicit 107 congestion notification (ECN) [RFC3168] markings. The requirements 108 for the congestion control algorithm are outlined in 109 [I-D.ietf-rmcat-cc-requirements]. 111 This document describes an experimental congestion control scheme 112 called network-assisted dynamic adaptation (NADA). The NADA design 113 benefits from explicit congestion control signals (e.g., ECN 114 markings) from the network, yet also operates when only implicit 115 congestion indicators (delay and/or loss) are available. Such a 116 unified sender behavior distinguishes NADA from other congestion 117 control schemes for real-time media. In addition, its core 118 congestion control algorithm is designed to guarantee stability for 119 path round-trip-times (RTTs) below a prescribed bound (e.g., 250ms 120 with default parameter choices). It further supports weighted 121 bandwidth sharing among competing video flows with different 122 priorities. The signaling mechanism consists of standard RTP 123 timestamp [RFC3550] and RTCP feedback reports with non-standard 124 messages. 126 2. Terminology 128 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 129 "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and 130 "OPTIONAL" in this document are to be interpreted as described in BCP 131 14 [RFC2119] [RFC8174] when, and only when, they appear in all 132 capitals, as shown here. 134 3. System Overview 136 Figure 1 shows the end-to-end system for real-time media transport 137 that NADA operates in. Note that there also exist network nodes 138 along the reverse (potentially uncongested) path that the RTCP 139 feedback reports traverse. Those network nodes are not shown in the 140 figure for sake of abrevity. 142 +---------+ r_vin +--------+ +--------+ +----------+ 143 | Media |<--------| RTP | |Network | | RTP | 144 | Encoder |========>| Sender |=======>| Node |====>| Receiver | 145 +---------+ r_vout +--------+ r_send +--------+ +----------+ 146 /|\ | 147 | | 148 +---------------------------------+ 149 RTCP Feedback Report 151 Figure 1: System Overview 153 o Media encoder with rate control capabilities. It encodes raw 154 media (audio and video) frames into compressed bitstream which is 155 later packetized into RTP packets. As discussed in 156 [I-D.ietf-rmcat-video-traffic-model], the actual output rate from 157 the encoder r_vout may fluctuate around the target r_vin. 158 Furthermore, it is possible that the encoder can only react to bit 159 rate changes at rather coarse time intervals, e.g., once every 0.5 160 seconds. 162 o RTP sender: responsible for calculating the NADA reference rate 163 based on network congestion indicators (delay, loss, or ECN 164 marking reports from the receiver), for updating the video encoder 165 with a new target rate r_vin, and for regulating the actual 166 sending rate r_send accordingly. The RTP sender also generates a 167 sending timestamp for each outgoing packet. 169 o RTP receiver: responsible for measuring and estimating end-to-end 170 delay (based on sender timestamp), packet loss (based on RTP 171 sequence number), ECN marking ratios (based on [RFC6679]), and 172 receiving rate (r_recv) of the flow. It calculates the aggregated 173 congestion signal (x_curr) that accounts for queuing delay, ECN 174 markings, and packet losses. The receiver also determines the 175 mode for sender rate adaptation (rmode) based on whether the flow 176 has encountered any standing non-zero congestion. The receiver 177 sends periodic RTCP reports back to the sender, containing values 178 of x_curr, rmode, and r_recv. 180 o Network node with several modes of operation. The system can work 181 with the default behavior of a simple drop tail queue. It can 182 also benefit from advanced AQM features such as PIE, FQ-CoDel, 183 RED-based ECN marking, and PCN marking using a token bucket 184 algorithm. Note that network node operation is out of control for 185 the design of NADA. 187 4. Core Congestion Control Algorithm 189 Like TCP-Friendly Rate Control (TFRC) [Floyd-CCR00] [RFC5348], NADA 190 is a rate-based congestion control algorithm. In its simplest form, 191 the sender reacts to the collection of network congestion indicators 192 in the form of an aggregated congestion signal, and operates in one 193 of two modes: 195 o Accelerated ramp-up: when the bottleneck is deemed to be 196 underutilized, the rate increases multiplicatively with respect to 197 the rate of previously successful transmissions. The rate 198 increase mutliplier (gamma) is calculated based on observed round- 199 trip-time and target feedback interval, so as to limit self- 200 inflicted queuing delay. 202 o Gradual rate update: in the presence of non-zero aggregate 203 congestion signal, the sending rate is adjusted in reaction to 204 both its value (x_curr) and its change in value (x_diff). 206 This section introduces the list of mathematical notations and 207 describes the core congestion control algorithm at the sender and 208 receiver, respectively. Additional details on recommended practical 209 implementations are described in Section 5.1 and Section 5.2. 211 4.1. Mathematical Notations 213 This section summarizes the list of variables and parameters used in 214 the NADA algorithm. 216 +--------------+-------------------------------------------------+ 217 | Notation | Variable Name | 218 +--------------+-------------------------------------------------+ 219 | t_curr | Current timestamp | 220 | t_last | Last time sending/receiving a feedback | 221 | delta | Observed interval between current and previous | 222 | | feedback reports: delta = t_curr-t_last | 223 | r_ref | Reference rate based on network congestion | 224 | r_send | Sending rate | 225 | r_recv | Receiving rate | 226 | r_vin | Target rate for video encoder | 227 | r_vout | Output rate from video encoder | 228 | d_base | Estimated baseline delay | 229 | d_fwd | Measured and filtered one-way delay | 230 | d_queue | Estimated queueing delay | 231 | d_tilde | Equivalent delay after non-linear warping | 232 | p_mark | Estimated packet ECN marking ratio | 233 | p_loss | Estimated packet loss ratio | 234 | x_curr | Aggregate congestion signal | 235 | x_prev | Previous value of aggregate congestion signal | 236 | x_diff | Change in aggregate congestion signal w.r.t. | 237 | | its previous value: x_diff = x_curr - x_prev | 238 | rmode | Rate update mode: (0 = accelerated ramp-up; | 239 | | 1 = gradual update) | 240 | gamma | Rate increase multiplier in accelerated ramp-up | 241 | | mode | 242 | loss_int | Measured average loss interval in packet count | 243 | loss_exp | Threshold value for setting the last observed | 244 | | packet loss to expiration | 245 | rtt | Estimated round-trip-time at sender | 246 | buffer_len | Rate shaping buffer occupancy measured in bytes | 247 +--------------+-------------------------------------------------+ 249 Figure 2: List of variables. 251 +--------------+----------------------------------+----------------+ 252 | Notation | Parameter Name | Default Value | 253 +--------------+----------------------------------+----------------+ 254 | PRIO | Weight of priority of the flow | 1.0 255 | RMIN | Minimum rate of application | 150 Kbps | 256 | | supported by media encoder | | 257 | RMAX | Maximum rate of application | 1.5 Mbps | 258 | | supported by media encoder | | 259 | XREF | Reference congestion level | 10ms | 260 | KAPPA | Scaling parameter for gradual | 0.5 | 261 | | rate update calculation | | 262 | ETA | Scaling parameter for gradual | 2.0 | 263 | | rate update calculation | | 264 | TAU | Upper bound of RTT in gradual | 500ms | 265 | | rate update calculation | | 266 | DELTA | Target feedback interval | 100ms | 267 +..............+..................................+................+ 268 | LOGWIN | Observation window in time for | 500ms | 269 | | calculating packet summary | | 270 | | statistics at receiver | | 271 | QEPS | Threshold for determining queuing| 10ms | 272 | | delay build up at receiver | | 273 | DFILT | Bound on filtering delay | 120ms | 274 | GAMMA_MAX | Upper bound on rate increase | 0.5 | 275 | | ratio for accelerated ramp-up | | 276 | QBOUND | Upper bound on self-inflicted | 50ms | 277 | | queuing delay during ramp up | | 278 +..............+..................................+................+ 279 | MULTILOSS | Multiplier for self-scaling the | 7. | 280 | | expiration threshold of the last | | 281 | | observed loss (loss_exp) based on| | 282 | | measured average loss interval | | 283 | | (loss_int) | | 284 | QTH | Delay threshold for invoking | 50ms | 285 | | non-linear warping | | 286 | LAMBDA | Scaling parameter in the | 0.5 | 287 | | exponent of non-linear warping | | 288 +..............+..................................+................+ 289 | PLRREF | Reference packet loss ratio | 0.01 | 290 | PMRREF | Reference packet marking ratio | 0.01 | 291 | DLOSS | Reference delay penalty for loss | 10ms | 292 | | when packet loss ratio is at | | 293 | | PLRREF | | 294 | DMARK | Reference delay penalty for ECN | 2ms | 295 | | marking when packet marking | | 296 | | is at PMRREF | | 297 +..............+..................................+................+ 298 | FPS | Frame rate of incoming video | 30 | 299 | BETA_S | Scaling parameter for modulating | 0.1 | 300 | | outgoing sending rate | | 301 | BETA_V | Scaling parameter for modulating | 0.1 | 302 | | video encoder target rate | | 303 | ALPHA | Smoothing factor in exponential | 0.1 | 304 | | smoothing of packet loss and | | 305 | | marking ratios | 306 +--------------+----------------------------------+----------------+ 308 Figure 3: List of algorithm parameters. 310 4.2. Receiver-Side Algorithm 312 The receiver-side algorithm can be outlined as below: 314 On initialization: 315 set d_base = +INFINITY 316 set p_loss = 0 317 set p_mark = 0 318 set r_recv = 0 319 set both t_last and t_curr as current time 321 On receiving a media packet: 322 obtain current timestamp t_curr from system clock 323 obtain from packet header sending time stamp t_sent 324 obtain one-way delay measurement: d_fwd = t_curr - t_sent 325 update baseline delay: d_base = min(d_base, d_fwd) 326 update queuing delay: d_queue = d_fwd - d_base 327 update packet loss ratio estimate p_loss 328 update packet marking ratio estimate p_mark 329 update measurement of receiving rate r_recv 331 On time to send a new feedback report (t_curr - t_last > DELTA): 332 calculate non-linear warping of delay d_tilde if packet loss exists 333 calculate current aggregate congestion signal x_curr 334 determine mode of rate adaptation for sender: rmode 335 send feedback containing values of: rmode, x_curr, and r_recv 336 update t_last = t_curr 338 In order for a delay-based flow to hold its ground when competing 339 against loss-based flows (e.g., loss-based TCP), it is important to 340 distinguish between different levels of observed queuing delay. For 341 instance, over wired connections, a moderate queuing delay value 342 below 100ms is likely self-inflicted or induced by other delay-based 343 flows, whereas a high queuing delay value of several hundreds of 344 milliseconds may indicate the presence of a loss-based flow that does 345 not refrain from increased delay. 347 If the last observed packet loss is within the expiration window of 348 loss_exp (measured in terms of packet counts), the estimated queuing 349 delay follows a non-linear warping: 351 / d_queue, if d_queue |||||||||=================> 599 +----------+ -----------+ RTP packets 600 Rate Shaping Buffer 602 Figure 4: NADA Sender Structure 604 5.2.1. Rate shaping buffer 606 The operation of the live video encoder is out of the scope of the 607 design for the congestion control scheme in NADA. Instead, its 608 behavior is treated as a black box. 610 A rate shaping buffer is employed to absorb any instantaneous 611 mismatch between encoder rate output r_vout and regulated sending 612 rate r_send. Its current level of occupancy is measured in bytes and 613 is denoted as buffer_len. 615 A large rate shaping buffer contributes to higher end-to-end delay, 616 which may harm the performance of real-time media communications. 617 Therefore, the sender has a strong incentive to prevent the rate 618 shaping buffer from building up. The mechanisms adopted are: 620 o To deplete the rate shaping buffer faster by increasing the 621 sending rate r_send; and 623 o To limit incoming packets of the rate shaping buffer by reducing 624 the video encoder target rate r_vin. 626 5.2.2. Adjusting video target rate and sending rate 628 The target rate for the live video encoder deviates from the network 629 congestion control rate r_ref based on the level of occupancy in the 630 rate shaping buffer: 632 r_vin = r_ref - BETA_V*8*buffer_len*FPS. (11) 634 The actual sending rate r_send is regulated in a similar fashion: 636 r_send = r_ref + BETA_S*8*buffer_len*FPS. (12) 638 In (11) and (12), the first term indicates the rate calculated from 639 network congestion feedback alone. The second term indicates the 640 influence of the rate shaping buffer. A large rate shaping buffer 641 nudges the encoder target rate slightly below -- and the sending rate 642 slightly above -- the reference rate r_ref. 644 Intuitively, the amount of extra rate offset needed to completely 645 drain the rate shaping buffer within the duration of a single video 646 frame is given by 8*buffer_len*FPS, where FPS stands for the frame 647 rate of the video. The scaling parameters BETA_V and BETA_S can be 648 tuned to balance between the competing goals of maintaining a small 649 rate shaping buffer and deviating from the reference rate point. 651 5.3. Feedback Message Requirements 653 The following list of information is required for NADA congestion 654 control to function properly: 656 o Recommended rate adaptation mode (rmode): a 1-bit flag indicating 657 whether the sender should operate in accelerated ramp-up mode 658 (rmode=0) or gradual update mode (rmode=1). 660 o Aggregated congestion signal (x_curr): the most recently updated 661 value, calculated by the receiver according to Section 4.2. This 662 information is expressed with a unit of 100 microsecond (i.e., 663 1/10 of a millisecond) in 15 bits. This allows a maximum value of 664 x_curr at approximately 3.27 second. 666 o Receiving rate (r_recv): the most recently measured receiving rate 667 according to Section 5.1.3. This information is expressed with a 668 unit of bits per second (bps) in 32 bits (unsigned int). This 669 allows a maximum rate of approximately 4.3Gbps. 671 The above list of information can be accommodated by 48 bits, or 6 672 bytes, in total. Choice of the feedback message interval DELTA is 673 discussed in Section 6.3 A target feedback interval of DELTA=100ms is 674 recommended. 676 6. Discussions and Further Investigations 678 6.1. Choice of delay metrics 680 The current design works with relative one-way-delay (OWD) as the 681 main indication of congestion. The value of the relative OWD is 682 obtained by maintaining the minimum value of observed OWD over a 683 relatively long time horizon and subtract that out from the observed 684 absolute OWD value. Such an approach cancels out the fixed 685 difference between the sender and receiver clocks. It has been 686 widely adopted by other delay-based congestion control approaches 687 such as [RFC6817]. As discussed in [RFC6817], the time horizon for 688 tracking the minimum OWD needs to be chosen with care: it must be 689 long enough for an opportunity to observe the minimum OWD with zero 690 standing queue along the path, and sufficiently short so as to timely 691 reflect "true" changes in minimum OWD introduced by route changes and 692 other rare events. 694 The potential drawback in relying on relative OWD as the congestion 695 signal is that when multiple flows share the same bottleneck, the 696 flow arriving late at the network experiencing a non-empty queue may 697 mistakenly consider the standing queuing delay as part of the fixed 698 path propagation delay. This will lead to slightly unfair bandwidth 699 sharing among the flows. 701 Alternatively, one could move the per-packet statistical handling to 702 the sender instead and use relative round-trip-time (RTT) in lieu of 703 relative OWD, assuming that per-packet acknowledgements are 704 available. The main drawback of RTT-based approach is the noise in 705 the measured delay in the reverse direction. 707 Note that the choice of either delay metric (relative OWD vs. RTT) 708 involves no change in the proposed rate adaptation algorithm. 709 Therefore, comparing the pros and cons regarding which delay metric 710 to adopt can be kept as an orthogonal direction of investigation. 712 6.2. Method for delay, loss, and marking ratio estimation 714 Like other delay-based congestion control schemes, performance of 715 NADA depends on the accuracy of its delay measurement and estimation 716 module. Appendix A in [RFC6817] provides an extensive discussion on 717 this aspect. 719 The current recommended practice of applying minimum filter with a 720 window size of 15 samples suffices in guarding against processing 721 delay outliers observed in wired connections. For wireless 722 connections with a higher packet delay variation (PDV), more 723 sophisticated techniques on de-noising, outlier rejection, and trend 724 analysis may be needed. 726 More sophisticated methods in packet loss ratio calculation, such as 727 that adopted by [Floyd-CCR00], will likely be beneficial. These 728 alternatives are currently under investigation. 730 6.3. Impact of parameter values 732 In the gradual rate update mode, the parameter TAU indicates the 733 upper bound of round-trip-time (RTT) in feedback control loop. 734 Typically, the observed feedback interval delta is close to the 735 target feedback interval DELTA, and the relative ratio of delta/TAU 736 versus ETA dictates the relative strength of influence from the 737 aggregate congestion signal offset term (x_offset) versus its recent 738 change (x_diff), respectively. These two terms are analogous to the 739 integral and proportional terms in a proportional-integral (PI) 740 controller. The recommended choice of TAU=500ms, DELTA=100ms and ETA 741 = 2.0 corresponds to a relative ratio of 1:10 between the gains of 742 the integral and proportional terms. Consequently, the rate 743 adaptation is mostly driven by the change in the congestion signal 744 with a long-term shift towards its equilibrium value driven by the 745 offset term. Finally, the scaling parameter KAPPA determines the 746 overall speed of the adaptation and needs to strike a balance between 747 responsiveness and stability. 749 The choice of the target feedback interval DELTA needs to strike the 750 right balance between timely feedback and low RTCP feedback message 751 counts. A target feedback interval of DELTA=100ms is recommended, 752 corresponding to a feedback bandwidth of 16Kbps with 200 bytes per 753 feedback message --- approximately 1.6% overhead for a 1 Mbps flow. 754 Furthermore, both simulation studies and frequency-domain analysis 755 have established that a feedback interval below 250ms (i.e., more 756 frequently than 4 feedback messages per second) will not break up the 757 feedback control loop of NADA congestion control. 759 In calculating the non-linear warping of delay in (1), the current 760 design uses fixed values of QTH for determining whether to perform 761 the non-linear warping). Its value may need to be tuned for 762 different operational enviornments (e.g., over wired vs. wireless 763 connections). It is possible to adapt its value based on past 764 observed patterns of queuing delay in the presence of packet losses. 765 It needs to be noted that the non-linear warping mechanism may lead 766 to multiple NADA streams stuck in loss-based mode when competing 767 against each other. 769 In calculating the aggregate congestion signal x_curr, the choice of 770 DMARK and DLOSS influence the steady-state packet loss/marking ratio 771 experienced by the flow at a given available bandwidth. Higher 772 values of DMARK and DLOSS result in lower steady-state loss/marking 773 ratios, but are more susceptible to the impact of individual packet 774 loss/marking events. While the value of DMARK and DLOSS are fixed 775 and predetermined in the current design, a scheme for automatically 776 tuning these values based on desired bandwidth sharing behavior in 777 the presence of other competing loss-based flows (e.g., loss-based 778 TCP) is under investigation. 780 6.4. Sender-based vs. receiver-based calculation 782 In the current design, the aggregated congestion signal x_curr is 783 calculated at the receiver, keeping the sender operation completely 784 independent of the form of actual network congestion indications 785 (delay, loss, or marking). Alternatively, one can move the logics of 786 (1) and (2) to the sender. Such an approach requires slightly higher 787 overhead in the feedback messages, which should contain individual 788 fields on queuing delay (d_queue), packet loss ratio (p_loss), packet 789 marking ratio (p_mark), receiving rate (r_recv), and recommended rate 790 adaptation mode (rmode). 792 6.5. Incremental deployment 794 One nice property of NADA is the consistent video endpoint behavior 795 irrespective of network node variations. This facilitates gradual, 796 incremental adoption of the scheme. 798 To start off with, the proposed congestion control mechanism can be 799 implemented without any explicit support from the network, and relies 800 solely on observed one-way delay measurements and packet loss ratios 801 as implicit congestion signals. 803 When ECN is enabled at the network nodes with RED-based marking, the 804 receiver can fold its observations of ECN markings into the 805 calculation of the equivalent delay. The sender can react to these 806 explicit congestion signals without any modification. 808 Ultimately, networks equipped with proactive marking based on token 809 bucket level metering can reap the additional benefits of zero 810 standing queues and lower end-to-end delay and work seamlessly with 811 existing senders and receivers. 813 7. Implementation Status 815 The NADA scheme has been implemented in [ns-2] and [ns-3] simulation 816 platforms. Extensive ns-2 simulation evaluations of an earlier 817 version of the draft are documented in [Zhu-PV13]. Evaluation 818 results of the current draft over several test cases in 819 [I-D.ietf-rmcat-eval-test] have been presented at recent IETF 820 meetings [IETF-90][IETF-91]. Evaluation results of the current draft 821 over several test cases in [I-D.ietf-rmcat-wireless-tests] have been 822 presented at [IETF-93]. An open source implementation of NADA as 823 part of a ns-3 module is available at [ns3-rmcat] 825 The scheme has also been implemented and evaluated in a lab setting 826 as described in [IETF-90]. Preliminary evaluation results of NADA in 827 single-flow and multi-flow scenarios have been presented in 828 [IETF-91]. 830 8. Suggested Experiments 832 NADA has been extensively evaluated under various test scenarios, 833 including the collection of test cases specified by 834 [I-D.ietf-rmcat-eval-test] and the subset of WiFi-based test cases in 835 [I-D.ietf-rmcat-wireless-tests]. Additional evaluations have been 836 carried out to characterize how NADA interacts with various active 837 queue management (AQM) schemes such as RED, CoDel, and PIE. Most of 838 these evaluations have been carried out in simulators. A few key 839 test cases have also bee evaluated in implementations embedded in 840 video conferencing clients. 842 Further experiments are suggested for the following scenarios: 844 o Experiments reflecting the set up of a typical WAN connection. 846 o Experiments with ECN marking capability turned on at the network 847 for existing test cases. 849 o Experiments with multiple RMCAT streams bearing different user- 850 specified priorities. 852 o Experiments with additional access technologies, especially over 853 cellular networks such as 3G/LTE. 855 o Experiments with various media source contents, including audio 856 only, audio and video, and application content sharing (e.g., 857 slide shows). 859 9. IANA Considerations 861 This document makes no request of IANA. 863 10. Security Considerations 865 TBD 867 11. Acknowledgements 869 The authors would like to thank Randell Jesup, Luca De Cicco, Piers 870 O'Hanlon, Ingemar Johansson, Stefan Holmer, Cesar Ilharco Magalhaes, 871 Safiqul Islam, Michael Welzl, Mirja Kuhlewind, Karen Elisabeth Egede 872 Nielsen, Julius Flohr, Roland Bless, and Andreas Smas for their 873 various valuable review comments and feedback. Thanks to Charles 874 Ganzhorn for contributing to the testbed-based evaluation of NADA 875 during an early stage of its development. 877 12. References 879 12.1. Normative References 881 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 882 Requirement Levels", BCP 14, RFC 2119, 883 DOI 10.17487/RFC2119, March 1997, 884 . 886 [RFC3168] Ramakrishnan, K., Floyd, S., and D. Black, "The Addition 887 of Explicit Congestion Notification (ECN) to IP", 888 RFC 3168, DOI 10.17487/RFC3168, September 2001, 889 . 891 [RFC3550] Schulzrinne, H., Casner, S., Frederick, R., and V. 892 Jacobson, "RTP: A Transport Protocol for Real-Time 893 Applications", STD 64, RFC 3550, DOI 10.17487/RFC3550, 894 July 2003, . 896 [RFC5348] Floyd, S., Handley, M., Padhye, J., and J. Widmer, "TCP 897 Friendly Rate Control (TFRC): Protocol Specification", 898 RFC 5348, DOI 10.17487/RFC5348, September 2008, 899 . 901 [RFC6679] Westerlund, M., Johansson, I., Perkins, C., O'Hanlon, P., 902 and K. Carlberg, "Explicit Congestion Notification (ECN) 903 for RTP over UDP", RFC 6679, DOI 10.17487/RFC6679, August 904 2012, . 906 [RFC8174] Leiba, B., "Ambiguity of Uppercase vs Lowercase in RFC 907 2119 Key Words", BCP 14, RFC 8174, DOI 10.17487/RFC8174, 908 May 2017, . 910 12.2. Informative References 912 [Budzisz-TON11] 913 Budzisz, L., Stanojevic, R., Schlote, A., Baker, F., and 914 R. Shorten, "On the Fair Coexistence of Loss- and Delay- 915 Based TCP", IEEE/ACM Transactions on Networking vol. 19, 916 no. 6, pp. 1811-1824, December 2011. 918 [Floyd-CCR00] 919 Floyd, S., Handley, M., Padhye, J., and J. Widmer, 920 "Equation-based Congestion Control for Unicast 921 Applications", ACM SIGCOMM Computer Communications 922 Review vol. 30, no. 4, pp. 43-56, October 2000. 924 [I-D.ietf-rmcat-cc-codec-interactions] 925 Zanaty, M., Singh, V., Nandakumar, S., and Z. Sarker, 926 "Congestion Control and Codec interactions in RTP 927 Applications", draft-ietf-rmcat-cc-codec-interactions-02 928 (work in progress), March 2016. 930 [I-D.ietf-rmcat-cc-requirements] 931 Jesup, R. and Z. Sarker, "Congestion Control Requirements 932 for Interactive Real-Time Media", draft-ietf-rmcat-cc- 933 requirements-09 (work in progress), December 2014. 935 [I-D.ietf-rmcat-eval-test] 936 Sarker, Z., Singh, V., Zhu, X., and M. Ramalho, "Test 937 Cases for Evaluating RMCAT Proposals", draft-ietf-rmcat- 938 eval-test-06 (work in progress), June 2018. 940 [I-D.ietf-rmcat-video-traffic-model] 941 Zhu, X., Cruz, S., and Z. Sarker, "Modeling Video Traffic 942 Sources for RMCAT Evaluations", draft-ietf-rmcat-video- 943 traffic-model-04 (work in progress), January 2018. 945 [I-D.ietf-rmcat-wireless-tests] 946 Sarker, Z., Johansson, I., Zhu, X., Fu, J., Tan, W., and 947 M. Ramalho, "Evaluation Test Cases for Interactive Real- 948 Time Media over Wireless Networks", draft-ietf-rmcat- 949 wireless-tests-04 (work in progress), May 2017. 951 [IETF-90] Zhu, X., Ramalho, M., Ganzhorn, C., Jones, P., and R. Pan, 952 "NADA Update: Algorithm, Implementation, and Test Case 953 Evalua6on Results", July 2014, 954 . 957 [IETF-91] Zhu, X., Pan, R., Ramalho, M., Mena, S., Ganzhorn, C., 958 Jones, P., and S. D'Aronco, "NADA Algorithm Update and 959 Test Case Evaluations", November 2014, 960 . 963 [IETF-93] Zhu, X., Pan, R., Ramalho, M., Mena, S., Ganzhorn, C., 964 Jones, P., D'Aronco, S., and J. Fu, "Updates on NADA", 965 July 2015, . 968 [ns-2] "The Network Simulator - ns-2", 969 . 971 [ns-3] "The Network Simulator - ns-3", . 973 [ns3-rmcat] 974 Fu, J., Mena, S., and X. Zhu, "NS3 open source module of 975 IETF RMCAT congestion control protocols", November 2017, 976 . 978 [RFC6660] Briscoe, B., Moncaster, T., and M. Menth, "Encoding Three 979 Pre-Congestion Notification (PCN) States in the IP Header 980 Using a Single Diffserv Codepoint (DSCP)", RFC 6660, 981 DOI 10.17487/RFC6660, July 2012, 982 . 984 [RFC6817] Shalunov, S., Hazel, G., Iyengar, J., and M. Kuehlewind, 985 "Low Extra Delay Background Transport (LEDBAT)", RFC 6817, 986 DOI 10.17487/RFC6817, December 2012, 987 . 989 [RFC7567] Baker, F., Ed. and G. Fairhurst, Ed., "IETF 990 Recommendations Regarding Active Queue Management", 991 BCP 197, RFC 7567, DOI 10.17487/RFC7567, July 2015, 992 . 994 [Zhu-PV13] 995 Zhu, X. and R. Pan, "NADA: A Unified Congestion Control 996 Scheme for Low-Latency Interactive Video", in Proc. IEEE 997 International Packet Video Workshop (PV'13) San Jose, CA, 998 USA, December 2013. 1000 Appendix A. Network Node Operations 1002 NADA can work with different network queue management schemes and 1003 does not assume any specific network node operation. As an example, 1004 this appendix describes three variants of queue management behavior 1005 at the network node, leading to either implicit or explicit 1006 congestion signals. It needs to be acknowledged that NADA has not 1007 yet been tested with non-probabilistic ECN marking behaviors. 1009 In all three flavors described below, the network queue operates with 1010 the simple first-in-first-out (FIFO) principle. There is no need to 1011 maintain per-flow state. The system can scale easily with a large 1012 number of video flows and at high link capacity. 1014 A.1. Default behavior of drop tail queues 1016 In a conventional network with drop tail or RED queues, congestion is 1017 inferred from the estimation of end-to-end delay and/or packet loss. 1018 Packet drops at the queue are detected at the receiver, and 1019 contributes to the calculation of the aggregated congestion signal 1020 x_curr. No special action is required at network node. 1022 A.2. RED-based ECN marking 1024 In this mode, the network node randomly marks the ECN field in the IP 1025 packet header following the Random Early Detection (RED) algorithm 1026 [RFC7567]. Calculation of the marking probability involves the 1027 following steps: 1029 on packet arrival: 1030 update smoothed queue size q_avg as: 1031 q_avg = w*q + (1-w)*q_avg. 1033 calculate marking probability p as: 1035 / 0, if q < q_lo; 1036 | 1037 | q_avg - q_lo 1038 p= < p_max*--------------, if q_lo <= q < q_hi; 1039 | q_hi - q_lo 1040 | 1041 \ p = 1, if q >= q_hi. 1043 Here, q_lo and q_hi corresponds to the low and high thresholds of 1044 queue occupancy. The maximum marking probability is p_max. 1046 The ECN markings events will contribute to the calculation of an 1047 equivalent delay x_curr at the receiver. No changes are required at 1048 the sender. 1050 A.3. Random Early Marking with Virtual Queues 1052 Advanced network nodes may support random early marking based on a 1053 token bucket algorithm originally designed for Pre-Congestion 1054 Notification (PCN) [RFC6660]. The early congestion notification 1055 (ECN) bit in the IP header of packets are marked randomly. The 1056 marking probability is calculated based on a token-bucket algorithm 1057 originally designed for the Pre-Congestion Notification (PCN) 1058 [RFC6660]. The target link utilization is set as 90%; the marking 1059 probability is designed to grow linearly with the token bucket size 1060 when it varies between 1/3 and 2/3 of the full token bucket limit. 1062 Calculation of the marking probability involves the following steps: 1064 upon packet arrival: 1065 meter packet against token bucket (r,b); 1067 update token level b_tk; 1069 calculate the marking probability as: 1071 / 0, if b-b_tk < b_lo; 1072 | 1073 | b-b_tk-b_lo 1074 p = < p_max* --------------, if b_lo<= b-b_tk =b_hi. 1079 Here, the token bucket lower and upper limits are denoted by b_lo and 1080 b_hi, respectively. The parameter b indicates the size of the token 1081 bucket. The parameter r is chosen to be below capacity, resulting in 1082 slight under-utilization of the link. The maximum marking 1083 probability is p_max. 1085 The ECN markings events will contribute to the calculation of an 1086 equivalent delay x_curr at the receiver. No changes are required at 1087 the sender. The virtual queuing mechanism from the PCN-based marking 1088 algorithm will lead to additional benefits such as zero standing 1089 queues. 1091 Authors' Addresses 1093 Xiaoqing Zhu 1094 Cisco Systems 1095 12515 Research Blvd., Building 4 1096 Austin, TX 78759 1097 USA 1099 Email: xiaoqzhu@cisco.com 1101 Rong Pan * 1102 * Pending affiliation change. 1104 Email: rong.pan@gmail.com 1106 Michael A. Ramalho 1107 Cisco Systems, Inc. 1108 8000 Hawkins Road 1109 Sarasota, FL 34241 1110 USA 1112 Phone: +1 919 476 2038 1113 Email: mramalho@cisco.com 1115 Sergio Mena de la Cruz 1116 Cisco Systems 1117 EPFL, Quartier de l'Innovation, Batiment E 1118 Ecublens, Vaud 1015 1119 Switzerland 1121 Email: semena@cisco.com 1123 Paul E. Jones 1124 Cisco Systems 1125 7025 Kit Creek Rd. 1126 Research Triangle Park, NC 27709 1127 USA 1129 Email: paulej@packetizer.com 1130 Jiantao Fu 1131 Cisco Systems 1132 707 Tasman Drive 1133 Milpitas, CA 95035 1134 USA 1136 Email: jianfu@cisco.com 1138 Stefano D'Aronco 1139 Ecole Polytechnique Federale de Lausanne 1140 EPFL STI IEL LTS4, ELD 220 (Batiment ELD), Station 11 1141 Lausanne CH-1015 1142 Switzerland 1144 Email: stefano.daronco@epfl.ch