RFC 4586
Extended RTP Profile for Real-time Transport Control Protocol (RTCP)-Based Feedback: Results of the Timing Rule Simulations
Network Working Group C. Burmeister Request for Comments: 4586 R. Hakenberg Category: Informational A. Miyazaki Panasonic J. Ott Helsinki University of Technology N. Sato S. Fukunaga Oki July 2006 Extended RTP Profile for Real-time Transport Control Protocol (RTCP)-Based Feedback: Results of the Timing Rule Simulations Status of This Memo This memo provides information for the Internet community. It does not specify an Internet standard of any kind. Distribution of this memo is unlimited. Copyright Notice Copyright (C) The Internet Society (2006).Abstract
This document describes the results achieved when simulating the timing rules of the Extended RTP Profile for Real-time Transport Control Protocol (RTCP)-Based Feedback, denoted AVPF. Unicast and multicast topologies are considered as well as several protocol and environment configurations. The results show that the timing rules result in better performance regarding feedback delay and still preserve the well-accepted RTP rules regarding allowed bit rates for control traffic.
Table of Contents
1. Introduction ....................................................3 2. Timing Rules of the Extended RTP Profile for RTCP-Based Feedback ........................................................4 3. Simulation Environment ..........................................5 3.1. Network Simulator Version 2 ................................5 3.2. RTP Agent ..................................................5 3.3. Scenarios ..................................................5 3.4. Topologies .................................................6 4. RTCP Bit Rate Measurements ......................................6 4.1. Unicast ....................................................7 4.2. Multicast .................................................10 4.3. Summary of the RTCP Bit Rate Measurements .................10 5. Feedback Measurements ..........................................11 5.1. Unicast ...................................................11 5.2. Multicast .................................................12 5.2.1. Shared Losses vs. Distributed Losses ...............13 6. Investigations on "l" ..........................................14 6.1. Feedback Suppression Performance ..........................16 6.2. Loss Report Delay .........................................18 6.3. Summary of "l" Investigations .............................18 7. Applications Using AVPF ........................................19 7.1. NEWPRED Implementation in NS2 .............................19 7.2. Simulation ................................................21 7.2.1. Simulation A - Constant Packet Loss Rate ...........21 7.2.2. Simulation B - Packet Loss Due to Congestion .......23 7.3. Summary of Application Simulations ........................24 8. Summary ........................................................24 9. Security Considerations ........................................25 10. Normative References ..........................................26 11. Informative References ........................................26
1. Introduction
The Real-time Transport Protocol (RTP) is widely used for the transmission of real-time or near real-time media data over the Internet. While it was originally designed to work well for multicast groups in very large scales, its scope is not limited to that. More and more applications use RTP for small multicast groups (e.g., video conferences) or even unicast (e.g., IP telephony and media streaming applications). RTP comes together with its companion protocol Real-time Transport Control Protocol (RTCP), which is used to monitor the transmission of the media data and provide feedback of the reception quality. Furthermore, it can be used for loose session control. Having the scope of large multicast groups in mind, the rules regarding when to send feedback were carefully restricted to avoid feedback explosion or feedback-related congestion in the network. RTP and RTCP have proven to work well in the Internet, especially in large multicast groups, which is shown by their widespread usage today. However, the applications that transmit the media data only to small multicast groups or unicast may benefit from more frequent feedback. The source of the packets may be able to react to changes in the reception quality, which may be due to varying network utilization (e.g., congestion) or other changes. Possible reactions include transmission rate adaptation according to a congestion control algorithm or the invocation of error resilience features for the media stream (e.g., retransmissions, reference picture selection, NEWPRED, etc.). As mentioned before, more frequent feedback may be desirable to increase the reception quality, but RTP restricts the use of RTCP feedback. Hence it was decided to create a new extended RTP profile, which redefines some of the RTCP timing rules, but keeps most of the algorithms for RTP and RTCP, which have proven to work well. The new rules should scale from unicast to multicast, where unicast or small multicast applications have the most gain from it. A detailed description of the new profile and its timing rules can be found in [1]. This document investigates the new algorithms by the means of simulations. We show that the new timing rules scale well and behave in a network-friendly manner. Firstly, the key features of the new RTP profile that are important for our simulations are roughly described in Section 2. After that, we describe in Section 3 the environment that is used to conduct the simulations. Section 4 describes simulation results that show the backwards compatibility to RTP and that the new profile is network-friendly in terms of used
bandwidth for RTCP traffic. In Section 5, we show the benefit that applications could get from implementing the new profile. In Section 6, we investigated the effect of the parameter "l" (used to calculate the T_dither_max value) upon the algorithm performance, and finally, in Section 7, we show the performance gain we could get for a special application, namely, NEWPRED in [6] and [7].2. Timing Rules of the Extended RTP Profile for RTCP-Based Feedback
As said above, RTP restricts the usage of RTCP feedback. The main restrictions on RTCP are as follows: - RTCP messages are sent in compound packets, i.e., every RTCP packet contains at least one sender report (SR) or receiver report (RR) message and a source description (SDES) message. - The RTCP compound packets are sent in time intervals (T_rr), which are computed as a function of the average packet size, the number of senders and receivers in the group, and the session bandwidth (5% of the session bandwidth is used for RTCP messages; this bandwidth is shared between all session members, where the senders may get a larger share than the receivers.) - The average minimum interval between two RTCP packets from the same source is 5 seconds. We see that these rules prevent feedback explosion and scale well to large multicast groups. However, they do not allow timely feedback at all. While the second rule scales also to small groups or unicast (in this cases the interval might be as small as a few milliseconds), the third rule may prevent the receivers from sending feedback timely. The timing rules to send RTCP feedback from the new RTP profile [1] consist of two key components. First, the minimum interval of 5 seconds is abolished. Second, receivers get one chance during every other of their (now quite small) RTCP intervals to send an RTCP packet "early", i.e., not according to the calculated interval, but virtually immediately. It is important to note that the RTCP interval calculation is still inherited from the original RTP specification. The specification and all the details of the extended timing rules can be found in [1]. Rather than describing the algorithms here, we reference the original specification [1]. Therefore, we use also the same variable names and abbreviations as in [1].
3. Simulation Environment
This section describes the simulation testbed that was used for the investigations and its key features. The extensions to the simulator that were necessary are roughly described in the following sections.3.1. Network Simulator Version 2
The simulations were conducted using the network simulator version 2 (ns2). ns2 is an open source project, written in a combination of Tool Command Language (TCL) and C++. The scenarios are set up using TCL. Using the scripts, it is possible to specify the topologies (nodes and links, bandwidths, queue sizes, or error rates for links) and the parameters of the "agents", i.e., protocol configurations. The protocols themselves are implemented in C++ in the agents, which are connected to the nodes. The documentation for ns2 and the newest version can be found in [4].3.2. RTP Agent
We implemented a new agent, based on RTP/RTCP. RTP packets are sent at a constant packet rate with the correct header sizes. RTCP packets are sent according to the timing rules of [2] and [3], and also its algorithms for group membership maintenance are implemented. Sender and receiver reports are sent. Further, we extended the agent to support the extended profile [1]. The use of the new timing rules can be turned on and off via parameter settings in TCL.3.3. Scenarios
The scenarios that are simulated are defined in TCL scripts. We set up several different topologies, ranging from unicast with two session members to multicast with up to 25 session members. Depending on the sending rates used and the corresponding link bandwidths, congestion losses may occur. In some scenarios, bit errors are inserted on certain links. We simulated groups with RTP/AVP agents, RTP/AVPF agents, and mixed groups. The feedback messages are generally NACK messages as defined in [1] and are triggered by packet loss.
3.4. Topologies
Mainly, four different topologies are simulated to show the key features of the extended profile. However, for some specific simulations we used different topologies. This is then indicated in the description of the simulation results. The main four topologies are named after the number of participating RTP agents, i.e., T-2, T-4, T-8, and T-16, where T-2 is a unicast scenario, T-4 contains four agents, etc. Figure 1 below illustrates the main topologies. A5 A5 | A6 / | / / | /--A7 / |/ A2 A2-----A6 A2--A8 / / / A9 / / / / / / / /---A10 A1-----A2 A1-----A3 A1-----A3-----A7 A1------A3< \ \ \ \---A11 \ \ \ \ \ \ \ A12 A4 A4-----A8 A4--A13 |\ | \--A14 | \ | A15 A16 T-2 T-4 T-8 T-16 Figure 1: Simulated topologies4. RTCP Bit Rate Measurements
The new timing rules allow more frequent RTCP feedback for small multicast groups. In large groups, the algorithm behaves similarly to the normal RTCP timing rules. While it is generally good to have more frequent feedback, it cannot be allowed at all to increase the bit rate used for RTCP above a fixed limit, i.e., 5% of the total RTP bandwidth according to RTP. This section shows that the new timing rules keep RTCP bandwidth usage under the 5% limit for all investigated scenarios, topologies, and group sizes. Furthermore, we show that mixed groups (some members using AVP, some AVPF) can be allowed and that each session member behaves
fairly according to its corresponding specification. Note that other values for the RTCP bandwidth limit may be specified using the RTCP bandwidth modifiers as in [10].4.1. Unicast
First we measured the RTCP bandwidth share in the unicast topology T-2. Even for a fixed topology and group size, there are several protocol parameters that are varied to simulate a large range of different scenarios. We varied the configurations of the agents in the sense that the agents may use AVP or AVPF. Thereby it is possible that one agent uses AVP and the other AVPF in one RTP session. This is done to test the backwards compatibility of the AVPF profile. Next, we consider scenarios where no losses occur. In this case, both RTP session members transmit the RTCP compound packets at regular intervals, calculated as T_rr, if they use AVPF, and use a minimum interval of 5 seconds (on average) if they implement AVP. No early packets are sent, because the need to send early feedback is not given. Still it is important to see that not more than 5% of the session bandwidth is used for RTCP and that AVP and AVPF members can coexist without interference. The results can be found in Table 1.
| | | | | | Used RTCP Bit Rate |
| Session | Send | Rec. | AVP | AVPF | (% of session bw) |
|Bandwidth|Agents|Agents|Agents|Agents| A1 | A2 | sum |
+---------+------+------+------+------+------+------+------+
| 2 Mbps | 1 | 2 | - | 1,2 | 2.42 | 2.56 | 4.98 |
| 2 Mbps | 1,2 | - | - | 1,2 | 2.49 | 2.49 | 4.98 |
| 2 Mbps | 1 | 2 | 1 | 2 | 0.01 | 2.49 | 2.50 |
| 2 Mbps | 1,2 | - | 1 | 2 | 0.01 | 2.48 | 2.49 |
| 2 Mbps | 1 | 2 | 1,2 | - | 0.01 | 0.01 | 0.02 |
| 2 Mbps | 1,2 | - | 1,2 | - | 0.01 | 0.01 | 0.02 |
|200 kbps | 1 | 2 | - | 1,2 | 2.42 | 2.56 | 4.98 |
|200 kbps | 1,2 | - | - | 1,2 | 2.49 | 2.49 | 4.98 |
|200 kbps | 1 | 2 | 1 | 2 | 0.06 | 2.49 | 2.55 |
|200 kbps | 1,2 | - | 1 | 2 | 0.08 | 2.50 | 2.58 |
|200 kbps | 1 | 2 | 1,2 | - | 0.06 | 0.06 | 0.12 |
|200 kbps | 1,2 | - | 1,2 | - | 0.08 | 0.08 | 0.16 |
| 20 kbps | 1 | 2 | - | 1,2 | 2.44 | 2.54 | 4.98 |
| 20 kbps | 1,2 | - | - | 1,2 | 2.50 | 2.51 | 5.01 |
| 20 kbps | 1 | 2 | 1 | 2 | 0.58 | 2.48 | 3.06 |
| 20 kbps | 1,2 | - | 1 | 2 | 0.77 | 2.51 | 3.28 |
| 20 kbps | 1 | 2 | 1,2 | - | 0.58 | 0.61 | 1.19 |
| 20 kbps | 1,2 | - | 1,2 | - | 0.77 | 0.79 | 1.58 |
Table 1: Unicast simulations without packet loss
We can see that in configurations where both agents use the new
timing rules each of them uses, at most, about 2.5% of the session
bandwidth for RTP, which sums up to 5% of the session bandwidth for
both. This is achieved regardless of the agent being a sender or a
receiver. In the cases where agent A1 uses AVP and agent A2 AVPF,
the total RTCP session bandwidth decreases. This is because agent A1
can send RTCP packets only with an average minimum interval of 5
seconds. Thus, only a small fraction of the session bandwidth is
used for its RTCP packets. For a high-bit-rate session (session
bandwidth = 2 Mbps), the fraction of the RTCP packets from agent A1
is as small as 0.01%. For smaller session bandwidths, the fraction
increases because the same amount of RTCP data is sent. The
bandwidth share that is used by RTCP packets from agent A2 is not
different from what was used, when both agents implemented the AVPF.
Thus, the interaction of AVP and AVPF agents is not problematic in
these scenarios at all.
In our second unicast experiment, we show that the allowed RTCP
bandwidth share is not exceeded, even if packet loss occurs. We
simulated a constant byte error rate (BYER) on the link. The byte
errors are inserted randomly according to a uniform distribution.
Packets with byte errors are discarded on the link; hence the receiving agents will not see the loss immediately. The agents detect packet loss by a gap in the sequence number. When an AVPF agent detects a packet loss, the early feedback procedure is started. As described in AVPF [1], in unicast T_dither_max is always zero, hence an early packet can be sent immediately if allow_early is true. If the last packet was already an early one (i.e., allow_early = false), the feedback might be appended to the next regularly scheduled receiver report. The max_feedback_delay parameter (which we set to 1 second in our simulations) determines if that is allowed. The results are shown in Table 2, where we can see that there is no difference in the RTCP bandwidth share, whether or not losses occur. This is what we expected, because even though the RTCP packet size grows and early packets are sent, the interval between the packets increases and thus the RTCP bandwidth stays the same. Only the RTCP bandwidth of the agents that use the AVP increases slightly. This is because the interval between the packets is still 5 seconds (in average), but the packet size increased because of the feedback that is appended. | | | | | | Used RTCP Bit Rate | | Session | Send | Rec. | AVP | AVPF | (% of session bw) | |Bandwidth|Agents|Agents|Agents|Agents| A1 | A2 | sum | +---------+------+------+------+------+------+------+------+ | 2 Mbps | 1 | 2 | - | 1,2 | 2.42 | 2.56 | 4.98 | | 2 Mbps | 1,2 | - | - | 1,2 | 2.49 | 2.49 | 4.98 | | 2 Mbps | 1 | 2 | 1 | 2 | 0.01 | 2.49 | 2.50 | | 2 Mbps | 1,2 | - | 1 | 2 | 0.01 | 2.48 | 2.49 | | 2 Mbps | 1 | 2 | 1,2 | - | 0.01 | 0.02 | 0.03 | | 2 Mbps | 1,2 | - | 1,2 | - | 0.01 | 0.01 | 0.02 | |200 kbps | 1 | 2 | - | 1,2 | 2.42 | 2.56 | 4.98 | |200 kbps | 1,2 | - | - | 1,2 | 2.50 | 2.49 | 4.99 | |200 kbps | 1 | 2 | 1 | 2 | 0.06 | 2.50 | 2.56 | |200 kbps | 1,2 | - | 1 | 2 | 0.08 | 2.49 | 2.57 | |200 kbps | 1 | 2 | 1,2 | - | 0.06 | 0.07 | 0.13 | |200 kbps | 1,2 | - | 1,2 | - | 0.09 | 0.08 | 0.17 | | 20 kbps | 1 | 2 | - | 1,2 | 2.42 | 2.57 | 4.99 | | 20 kbps | 1,2 | - | - | 1,2 | 2.52 | 2.51 | 5.03 | | 20 kbps | 1 | 2 | 1 | 2 | 0.58 | 2.54 | 3.12 | | 20 kbps | 1,2 | - | 1 | 2 | 0.83 | 2.43 | 3.26 | | 20 kbps | 1 | 2 | 1,2 | - | 0.58 | 0.73 | 1.31 | | 20 kbps | 1,2 | - | 1,2 | - | 0.86 | 0.84 | 1.70 | Table 2: Unicast simulations with packet loss
4.2. Multicast
Next, we investigated the RTCP bandwidth share in multicast scenarios; i.e., we simulated the topologies T-4, T-8, and T-16 and measured the fraction of the session bandwidth that was used for RTCP packets. Again we considered different situations and protocol configurations (e.g., with or without bit errors, groups with AVP and/or AVPF agents, etc.). For reasons of readability, we present only selected results. For a documentation of all results, see [5]. The simulations of the different topologies in scenarios where no losses occur (neither through bit errors nor through congestion) show a similar behavior as in the unicast case. For all group sizes, the maximum RTCP bit rate share used is 5.06% of the session bandwidth in a simulation of 16 session members in a low-bit-rate scenario (session bandwidth = 20 kbps) with several senders. In all other scenarios without losses, the RTCP bit rate share used is below that. Thus, the requirement that not more than 5% of the session bit rate should be used for RTCP is fulfilled with reasonable accuracy. Simulations where bit errors are randomly inserted in RTP and RTCP packets and the corrupted packets are discarded give the same results. The 5% rule is kept (at maximum 5.07% of the session bandwidth is used for RTCP). Finally, we conducted simulations where we reduced the link bandwidth and thereby caused congestion-related losses. These simulations are different from the previous bit error simulations, in that the losses occur more in bursts and are more correlated, also between different agents. The correlation and "burstiness" of the packet loss is due to the queuing discipline in the routers we simulated; we used simple FIFO queues with a drop-tail strategy to handle congestion. Random Early Detection (RED) queues may enhance the performance, because the burstiness of the packet loss might be reduced; however, this is not the subject of our investigations, but is left for future study. The delay between the agents, which also influences RTP and RTCP packets, is much more variable because of the added queuing delay. Still the RTCP bit rate share used does not increase beyond 5.09% of the session bandwidth. Thus, also for these special cases the requirement is fulfilled.4.3. Summary of the RTCP Bit Rate Measurements
We have shown that for unicast and reasonable multicast scenarios, feedback implosion does not happen. The requirement that at maximum 5% of the session bandwidth is used for RTCP is fulfilled for all investigated scenarios.
5. Feedback Measurements
In this section we describe the results of feedback delay measurements, which we conducted in the simulations. Therefore, we use two metrics for measuring the performance of the algorithms; these are the "mean waiting time" (MWT) and the number of feedback packets that are sent, suppressed, or not allowed. The waiting time is the time, measured at a certain agent, between the detection of a packet loss event and the time when the corresponding feedback is sent. Assuming that the value of the feedback decreases with its delay, we think that the mean waiting time is a good metric to measure the performance gain we could get by using AVPF instead of AVP. The feedback an RTP/AVPF agent wants to send can be either sent or not sent. If it was not sent, this could be due to feedback suppression (i.e., another receiver already sent the same feedback) or because the feedback was not allowed (i.e., the max_feedback_delay was exceeded). We traced for every detected loss, if the agent sent the corresponding feedback or not and if not, why. The more feedback was not allowed, the worse the performance of the algorithm. Together with the waiting times, this gives us a good hint of the overall performance of the scheme.5.1. Unicast
In the unicast case, the maximum dithering interval T_dither_max is fixed and set to zero. This is because it does not make sense for a unicast receiver to wait for other receivers if they have the same feedback to send. But still feedback can be delayed or might not be permitted to be sent at all. The regularly scheduled packets are spaced according to T_rr, which depends in the unicast case mainly on the session bandwidth. Table 3 shows the mean waiting times (MWTs) measured in seconds for some configurations of the unicast topology T-2. The number of feedback packets that are sent or discarded is listed also (feedback sent (sent) or feedback discarded (disc)). We do not list suppressed packets, because for the unicast case feedback suppression does not apply. In the simulations, agent A1 was a sender and agent A2 was a pure receiver.
| | | Feedback Statistics |
| Session | | AVP | AVPF |
|Bandwidth| PLR | sent |disc| MWT | sent |disc| MWT |
+---------+-------+------+----+-------+------+----+-------+
| 2 Mbps | 0.001 | 781 | 0 | 2.604 | 756 | 0 | 0.015 |
| 2 Mbps | 0.01 | 7480 | 0 | 2.591 | 7548 | 2 | 0.006 |
| 2 Mbps | cong. | 25 | 0 | 2.557 | 1741 | 0 | 0.001 |
| 20 kbps | 0.001 | 79 | 0 | 2.472 | 74 | 2 | 0.034 |
| 20 kbps | 0.01 | 780 | 0 | 2.605 | 709 | 64 | 0.163 |
| 20 kbps | cong. | 780 | 0 | 2.590 | 687 | 70 | 0.162 |
Table 3: Feedback statistics for the unicast simulations
From the table above we see that the mean waiting time can be
decreased dramatically by using AVPF instead of AVP. While the
waiting times for agents using AVP is always around 2.5 seconds (half
the minimum interval average), it can be decreased to a few ms for
most of the AVPF configurations.
In the configurations with high session bandwidth, normally all
triggered feedback is sent. This is because more RTCP bandwidth is
available. There are only very few exceptions, which are probably
due to more than one packet loss within one RTCP interval, where the
first loss was by chance sent quite early. In this case, it might be
possible that the second feedback is triggered after the early packet
was sent, but possibly too early to append it to the next regularly
scheduled report, because of the limitation of the
max_feedback_delay. This is different for the cases with a small
session bandwidth, where the RTCP bandwidth share is quite low and
T_rr thus larger. After an early packet was sent, the time to the
next regularly scheduled packet can be very high. We saw that in
some cases the time was larger than the max_feedback_delay, and in
these cases the feedback is not allowed to be sent at all.
With a different setting of max_feedback_delay, it is possible to
have either more feedback that is not allowed and a decreased mean
waiting time or more feedback that is sent but an increased waiting
time. Thus, the parameter should be set with care according to the
application's needs.
5.2. Multicast
In this section, we describe some measurements of feedback statistics
in the multicast simulations. We picked out certain characteristic
and representative results. We considered the topology T-16.
Different scenarios and applications are simulated for this topology.
The parameters of the different links are set as follows. The agents
A2, A3, and A4 are connected to the middle node of the multicast
tree, i.e., agent A1, via high bandwidth and low-delay links. The other agents are connected to the nodes 2, 3, and 4 via different link characteristics. The agents connected to node 2 represent mobile users. They suffer in certain configurations from a certain byte error rate on their access links and the delays are high. The agents that are connected to node 3 have low-bandwidth access links, but do not suffer from bit errors. The last agents, which are connected to node 4, have high bandwidth and low delay.5.2.1. Shared Losses vs. Distributed Losses
In our first investigation, we wanted to see the effect of the loss characteristic on the algorithm's performance. We investigate the cases where packet loss occurs for several users simultaneously (shared losses) or totally independently (distributed losses). We first define agent A1 to be the sender. In the case of shared losses, we inserted a constant byte error rate on one of the middle links, i.e., the link between A1 and A2. In the case of distributed losses, we inserted the same byte error rate on all links downstream of A2. These scenarios are especially interesting because of the feedback suppression algorithm. When all receivers share the same loss, it is only necessary for one of them to send the loss report. Hence if a member receives feedback with the same content that it has scheduled to be sent, it suppresses the scheduled feedback. Of course, this suppressed feedback does not contribute to the mean waiting times. So we expect reduced waiting times for shared losses, because the probability is high that one of the receivers can send the feedback more or less immediately. The results are shown in the following table. | | Feedback Statistics | | | Shared Losses | Distributed Losses | |Agent|sent|fbsp|disc|sum | MWT |sent|fbsp|disc|sum | MWT | +-----+----+----+----+----+-----+----+----+----+----+-----+ | A2 | 274| 351| 25| 650|0.267| -| -| -| -| -| | A5 | 231| 408| 11| 650|0.243| 619| 2| 32| 653|0.663| | A6 | 234| 407| 9| 650|0.235| 587| 2| 32| 621|0.701| | A7 | 223| 414| 13| 650|0.253| 594| 6| 41| 641|0.658| | A8 | 188| 443| 19| 650|0.235| 596| 1| 32| 629|0.677| Table 4: Feedback statistics for multicast simulations Table 4 shows the feedback statistics for the simulation of a large group size. All 16 agents of topology T-16 joined the RTP session. However, only agent A1 acts as an RTP sender; the other agents are pure receivers. Only 4 or 5 agents suffer from packet loss, i.e.,
A2, A5, A6, A7, and A8 for the case of shared losses and A5, A6, A7, and A8 in the case of distributed losses. Since the number of session members is the same for both cases, T_rr is also the same on the average. Still the mean waiting times are reduced by more than 50% in the case of shared losses. This proves our assumption that shared losses enhance the performance of the algorithm, regardless of the loss characteristic. The feedback suppression mechanism seems to be working quite well. Even though some feedback is sent from different receivers (i.e., 1150 loss reports are sent in total and only 650 packets were lost, resulting in loss reports being received on the average 1.8 times), most of the redundant feedback was suppressed. That is, 2023 loss reports were suppressed from 3250 individual detected losses, which means that more than 60% of the feedback was actually suppressed.6. Investigations on "l"
In this section, we want to investigate the effect of the parameter "l" on the T_dither_max calculation in RTP/AVPF agents. We investigate the feedback suppression performance as well as the report delay for three sample scenarios. For all receivers, the T_dither_max value is calculated as T_dither_max = l * T_rr, with l = 0.5. The rationale for this is that, in general, if the receiver has no round-trip time (RTT) estimation, it does not know how long it should wait for other receivers to send feedback. The feedback suppression algorithm would certainly fail if the time selected is too short. However, the waiting time is increased unnecessarily (and thus the value of the feedback is decreased) in case the chosen value is too large. Ideally, the optimum time value could be found for each case, but this is not always feasible. On the other hand, it is not dangerous if the optimum time is not used. A decreased feedback value and a failure of the feedback suppression mechanism do not hurt the network stability. We have shown for the cases of distributed losses that the overall bandwidth constraints are kept in any case and thus we could only lose some performance by choosing the wrong time value. On the other hand, a good measure for T_dither_max is the RTCP interval T_rr. This value increases with the number of session members. Also, we know that we can send feedback at least every T_rr. Thus, increasing T_dither max beyond T_rr would certainly make no sense. So by choosing T_rr/2, we guarantee that at least sometimes (i.e., when a loss is detected in the first half of the interval between two regularly scheduled RTCP packets) we are allowed to send early packets. Because of the randomness of T_dither, we still have a good chance of sending the early packet in time.
The AVPF profile specifies that the calculation of T_dither_max, as given above, is common to session members having an RTT estimation and to those not having it. If this were not so, participants using different calculations for T_dither_max might also have very different mean waiting times before sending feedback, which translates into different reporting priorities. For example, in a scenario where T_rr = 1 s and the RTT = 100 ms, receivers using the RTT estimation would, on average, send more feedback than those not using it. This might partially cancel out the feedback suppression mechanism and even cause feedback implosion. Also note that, in a general case where the losses are shared, the feedback suppression mechanism works if the feedback packets from each receiver have enough time to reach each of the other ones before the calculated T_dither_max seconds. Therefore, in scenarios of very high bandwidth (small T_rr), the calculated T_dither_max could be much smaller than the propagation delay between receivers, which would translate into a failure of the feedback suppression mechanism. In these cases, one solution could be to limit the bandwidth available to receivers (see [10]) such that this does not happen. Another solution could be to develop a mechanism for feedback suppression based on the RTT estimation between senders. This will not be discussed here and may be the subject of another document. Note, however, that a really high bandwidth media stream is not that likely to rely on this kind of error repair in the first place. In the following, we define three representative sample scenarios. We use the topology from the previous section, T-16. Most of the agents contribute only little to the simulations, because we introduced an error rate only on the link between the sender A1 and the agent A2. The first scenario represents those cases, where losses are shared between two agents. One agent is located upstream on the path between the other agent and the sender. Therefore, agent A2 and agent A5 see the same losses that are introduced on the link between the sender and agent A2. Agents A6, A7, and A8 do not join the RTP session. From the other agents, only agents A3 and A9 join. All agents are pure receivers, except A1, which is the sender. The second scenario also represents cases where losses are shared between two agents, but this time the agents are located on different branches of the multicast tree. The delays to the sender are roughly of the same magnitude. Agents A5 and A6 share the same losses. Agents A3 and A9 join the RTP session, but are pure receivers and do not see any losses. Finally, in the third scenario, the losses are shared between two agents, A5 and A6. The same agents as in the second scenario are
active. However, the delays of the links are different. The delay of the link between agents A2 and A5 is reduced to 20 ms and between A2 and A6 to 40 ms. All agents beside agent A1 are pure RTP receivers. Thus, these agents do not have an RTT estimation to the source. T_dither_max is calculated with the above given formula, depending only on T_rr and l, which means that all agents should calculate roughly the same T_dither_max.6.1. Feedback Suppression Performance
The feedback suppression rate for an agent is defined as the ratio of the total number of feedback packets not sent out of the total number of feedback packets the agent intended to send (i.e., the sum of sent and not sent). The reasons for not sending a packet include: the receiver already saw the same loss reported in a receiver report coming from another session member or the max_feedback_delay (application-specific) was surpassed. The results for the feedback suppression rate of the agent Af that is further away from the sender are depicted in Table 5. In general, it can be seen that the feedback suppression rate increases as l increases. However there is a threshold, depending on the environment, from which the additional gain is not significant anymore. | | Feedback Suppression Rate | | l | Scen. 1 | Scen. 2 | Scen. 3 | +------+---------+---------+---------+ | 0.10 | 0.671 | 0.051 | 0.089 | | 0.25 | 0.582 | 0.060 | 0.210 | | 0.50 | 0.524 | 0.114 | 0.361 | | 0.75 | 0.523 | 0.180 | 0.370 | | 1.00 | 0.523 | 0.204 | 0.369 | | 1.25 | 0.506 | 0.187 | 0.372 | | 1.50 | 0.536 | 0.213 | 0.414 | | 1.75 | 0.526 | 0.215 | 0.424 | | 2.00 | 0.535 | 0.216 | 0.400 | | 3.00 | 0.522 | 0.220 | 0.405 | | 4.00 | 0.522 | 0.220 | 0.405 | Table 5: Fraction of feedback that was suppressed at agent (Af) of the total number of feedback messages the agent wanted to send Similar results can be seen in Table 6 for the agent An that is nearer to the sender.
| | Feedback Suppression Rate |
| l | Scen. 1 | Scen. 2 | Scen. 3 |
+------+---------+---------+---------+
| 0.10 | 0.056 | 0.056 | 0.090 |
| 0.25 | 0.063 | 0.055 | 0.166 |
| 0.50 | 0.116 | 0.099 | 0.255 |
| 0.75 | 0.141 | 0.141 | 0.312 |
| 1.00 | 0.179 | 0.175 | 0.352 |
| 1.25 | 0.206 | 0.176 | 0.361 |
| 1.50 | 0.193 | 0.193 | 0.337 |
| 1.75 | 0.197 | 0.204 | 0.341 |
| 2.00 | 0.207 | 0.207 | 0.368 |
| 3.00 | 0.196 | 0.203 | 0.359 |
| 4.00 | 0.196 | 0.203 | 0.359 |
Table 6: Fraction of feedback that was suppressed at agent (An) of
the total number of feedback messages the agent wanted to send
The rate of feedback suppression failure is depicted in Table 7. The
trend of additional performance increase is not significant beyond a
certain threshold. Dependence on the scenario is noticeable here as
well.
| |Feedback Suppr. Failure Rate |
| l | Scen. 1 | Scen. 2 | Scen. 3 |
+------+---------+---------+---------+
| 0.10 | 0.273 | 0.893 | 0.822 |
| 0.25 | 0.355 | 0.885 | 0.624 |
| 0.50 | 0.364 | 0.787 | 0.385 |
| 0.75 | 0.334 | 0.679 | 0.318 |
| 1.00 | 0.298 | 0.621 | 0.279 |
| 1.25 | 0.289 | 0.637 | 0.267 |
| 1.50 | 0.274 | 0.595 | 0.249 |
| 1.75 | 0.274 | 0.580 | 0.235 |
| 2.00 | 0.258 | 0.577 | 0.233 |
| 3.00 | 0.282 | 0.577 | 0.236 |
| 4.00 | 0.282 | 0.577 | 0.236 |
Table 7: The ratio of feedback suppression failures.
Summarizing the feedback suppression results, it can be said that in
general the feedback suppression performance increases as l
increases. However, beyond a certain threshold, depending on
environment parameters such as propagation delays or session
bandwidth, the additional increase is not significant anymore. This
threshold is not uniform across all scenarios; a value of l=0.5 seems
to produce reasonable results with acceptable (though not optimal)
overhead.
6.2. Loss Report Delay
In this section, we show the results for the measured report delay during the simulations of the three sample scenarios. This measurement is a metric of the performance of the algorithms, because the value of the feedback for the sender typically decreases with the delay of its reception. The loss report delay is measured as the time at the sender between sending a packet and receiving the first corresponding loss report. | | Mean Loss Report Delay | | l | Scen. 1 | Scen. 2 | Scen. 3 | +------+---------+---------+---------+ | 0.10 | 0.124 | 0.282 | 0.210 | | 0.25 | 0.168 | 0.266 | 0.234 | | 0.50 | 0.243 | 0.264 | 0.284 | | 0.75 | 0.285 | 0.286 | 0.325 | | 1.00 | 0.329 | 0.305 | 0.350 | | 1.25 | 0.351 | 0.329 | 0.370 | | 1.50 | 0.361 | 0.363 | 0.388 | | 1.75 | 0.360 | 0.387 | 0.392 | | 2.00 | 0.367 | 0.412 | 0.400 | | 3.00 | 0.368 | 0.507 | 0.398 | | 4.00 | 0.368 | 0.568 | 0.398 | Table 8: The mean loss report delay, measured at the sender. As can be seen from Table 8, the delay increases, in general, as l increases. Also, a similar effect as for the feedback suppression performance is present: beyond a certain threshold, the additional increase in delay is not significant anymore. The threshold is environment dependent and seems to be related to the threshold, where the feedback suppression gain would not increase anymore.6.3. Summary of "l" Investigations
We have shown experimentally that the performance of the feedback suppression mechanisms increases as l increases. The same applies for the report delay, which also increases as l increases. This leads to a threshold where both the performance and the delay do not increase any further. The threshold is dependent upon the environment. So finding an optimum value of l is not possible because it is always a trade-off between delay and feedback suppression performance. With l=0.5, we think that a trade-off was found that is acceptable for typical applications and environments.
7. Applications Using AVPF
NEWPRED is one of the error resilience tools, which is defined in both ISO/IEC MPEG-4 visual part and ITU-T H.263. NEWPRED achieves fast error recovery using feedback messages. We simulated the behavior of NEWPRED in the network simulator environment as described above and measured the waiting time statistics, in order to verify that the extended RTP profile for RTCP-based feedback (AVPF) [1] is appropriate for the NEWPRED feedback messages. Simulation results, which are presented in the following sections, show that the waiting time is small enough to get the expected performance of NEWPRED.7.1. NEWPRED Implementation in NS2
The agent that performs the NEWPRED functionality, called NEWPRED agent, is different from the RTP agent we described above. Some of the added features and functionalities are described in the following points: Application Feedback The "Application Layer Feedback Messages" format is used to transmit the NEWPRED feedback messages. Thereby the NEWPRED functionality is added to the RTP agent. The NEWPRED agent creates one NACK message for each lost segment of a video frame, and then assembles multiple NACK messages corresponding to the segments in the same video frame into one Application Layer Feedback Message. Although there are two modes, namely, NACK mode and ACK mode, in NEWPRED [6][7], only NACK mode is used in these simulations. In this simulation, the RTP layer doesn't generate feedback messages. Instead, the decoder (NEWPRED) generates a NACK message when the segment cannot be decoded because the data hasn't arrived or loss of reference picture has occurred. Those conditions are detected in the decoder with frame number, segment number, and existence of reference pictures in the decoder. The parameters of NEWPRED agent are as follows: f: Frame Rate(frames/sec) seg: Number of segments in one video frame bw: RTP session bandwidth(kbps) Generation of NEWPRED's NACK Messages The NEWPRED agent generates NACK messages when segments are lost.
a. The NEWPRED agent generates multiple NACK messages per one
video frame when multiple segments are lost. These are
assembled into one Feedback Control Information (FCI) message
per video frame. If there is no lost segment, no message is
generated and sent.
b. The length of one NACK message is 4 bytes. Let num be the
number of NACK messages in one video frame (1 <= num <= seg).
Thus, 12+4*num bytes is the size of the low-delay RTCP feedback
message in a compound RTCP packet.
Measurements
We defined two values to be measured:
- Recovery time
The recovery time is measured as the time between the detection
of a lost segment and reception of a recovered segment. We
measured this "recovery time" for each lost segment.
- Waiting time
The waiting time is the additional delay due to the feedback
limitation of RTP.
Figure 2 depicts the behavior of a NEWPRED agent when a loss occurs.
The recovery time is approximated as follows:
(Recovery time) = (Waiting time) +
(Transmission time for feedback message) +
(Transmission time for media data)
Therefore, the waiting time is derived as follows:
(Waiting time) = (Recovery time) - (Round-trip delay), where
(Round-trip delay ) = (Transmission time for feedback message) +
(Transmission time for media data)
Picture Reference |: Picture Segment ____________________ %: Lost Segment /_ _ _ _ \ v/ \ / \ / \ / \ \ v \v \v \v \ \ Sender ---|----|----|----|----|----|---|-------------> \ \ ^ \ \ \ / \ \ \ / \ \ v / \ \ x / \ \ Lost / \ \ x / \ _____ v x / NACK v Receiver ---------------|----%===-%----%----%----|-----> |-a-| | |------- b -------| a: Waiting time b: Recover time (%: Video segments are lost) Figure 2: Relation between the measured values at the NEWPRED agent7.2. Simulation
We conducted two simulations (Simulation A and Simulation B). In Simulation A, the packets are dropped with a fixed packet loss rate on a link between two NEWPRED agents. In Simulation B, packet loss occurs due to congestion from other traffic sources, i.e., ftp sessions.7.2.1. Simulation A - Constant Packet Loss Rate
The network topology used for this simulation is shown in Figure 3. Link 1 Link 2 Link 3 +--------+ +------+ +------+ +--------+ | Sender |------|Router|-------|Router|------|Receiver| +--------+ +------+ +------+ +--------+ 10(msec) x(msec) 10(msec) Figure 3: Network topology that is used for Simulation A Link1 and link3 are error free, and each link delay is 10 msec. Packets may get dropped on link2. The packet loss rates (Plr) and link delay (D) are as follows:
D [ms] = {10, 50, 100, 200, 500}
Plr = {0.005, 0.01, 0.02, 0.03, 0.05, 0.1, 0.2}
Session bandwidth, frame rate, and the number of segments are shown
in Table 9.
+------------+----------+-------------+-----+
|Parameter ID| bw(kbps) |f (frame/sec)| seg |
+------------+----------+-------------+-----+
| 32k-4-3 | 32 | 4 | 3 |
| 32k-5-3 | 32 | 5 | 3 |
| 64k-5-3 | 64 | 5 | 3 |
| 64k-10-3 | 64 | 10 | 3 |
| 128k-10-6 | 128 | 10 | 6 |
| 128k-15-6 | 128 | 15 | 6 |
| 384k-15-6 | 384 | 15 | 6 |
| 384k-30-6 | 384 | 30 | 6 |
| 512k-30-6 | 512 | 30 | 6 |
| 1000k-30-9 | 1000 | 30 | 9 |
| 2000k-30-9 | 2000 | 30 | 9 |
+------------+----------+-------------+-----+
Table 9: Parameter sets of the NEWPRED agents
Figure 4 shows the key values of the result (packet loss rate vs.
mean of waiting time).
When the packet loss rate is 5% and the session bandwidth is 32 kbps,
the waiting time is around 400 msec, which is just allowable for
reasonable NEWPRED performance.
When the packet loss rate is less than 1%, the waiting time is less
than 200 msec. In such a case, the NEWPRED allows as much as
200-msec additional link delay.
When the packet loss rate is less than 5% and the session bandwidth
is 64 kbps, the waiting time is also less than 200 msec.
In 128-kbps cases, the result shows that when the packet loss rate is
20%, the waiting time is around 200 msec. In cases with more than
512-kbps session bandwidth, there is no significant delay. This
means that the waiting time due to the feedback limitation of RTCP is
negligible for the NEWPRED performance.
+------------------------------------------------------------+ | | Packet Loss Rate = | | Bandwidth | 0.005| 0.01 | 0.02 | 0.03 | 0.05 |0.10 |0.20 | |-----------+------+------+------+------+------+------+------| | 32k |130- |200- |230- |280- |350- |470- |560- | | | 180| 250| 320| 390| 430| 610| 780| | 64k | 80- |100- |120- |150- |180- |210- |290- | | | 130| 150| 180| 190| 210| 300| 400| | 128k | 60- | 70- | 90- |110- |130- |170- |190- | | | 70| 80| 100| 120| 140| 190| 240| | 384k | 30- | 30- | 30- | 40- | 50- | 50- | 50- | | | 50| 50| 50| 50| 60| 70| 90| | 512k | < 50 | < 50 | < 50 | < 50 | < 50 | < 50 | < 60 | | | | | | | | | | | 1000k | < 50 | < 50 | < 50 | < 50 | < 50 | < 50 | < 55 | | | | | | | | | | | 2000k | < 30 | < 30 | < 30 | < 30 | < 30 | < 35 | < 35 | +------------------+------+------+------+------+------+------+ Figure 4: The result of simulation A7.2.2. Simulation B - Packet Loss Due to Congestion
The configurations of link1, link2, and link3 are the same as in Simulation A except that link2 is also error-free, regarding bit errors. However, in addition, some FTP agents are deployed to overload link2. See Figure 5 for the simulation topology. Link1 Link2 Link3 +--------+ +------+ +------+ +--------+ | Sender |------|Router|-------|Router|------|Receiver| +--------+ /|+------+ +------+|\ +--------+ +---+/ | | \+---+ +-|FTP|+---+ +---+|FTP|-+ | +---+|FTP| ... |FTP|+---+ | ... +---+ +---+ +---+ +---+ FTP Agents FTP Agents Figure 5: Network Topology of Simulation B The parameters are defined as for Simulation A with the following values assigned: D[ms] ={10, 50, 100, 200, 500} 32 FTP agents are deployed at each edge, for a total of 64 FTP agents active.
The sets of session bandwidth, frame rate, and the number of segments are the same as in Simulation A (Table 9). We provide the results for the cases with 64 FTP agents, because these are the cases where packet losses could be detected to be stable. The results are similar to those for Simulation A except for a constant additional offset of 50..100 ms. This is due to the delay incurred by the routers' buffers.7.3. Summary of Application Simulations
We have shown that the limitations of RTP AVPF profile do not generate such high delay in the feedback messages that the performance of NEWPRED is degraded for sessions from 32 kbps to 2 Mbps. We could see that the waiting time increases with a decreasing session bandwidth and/or an increasing packet loss rate. The cause of the packet loss is not significant; congestion and constant packet loss rates behave similarly. Still we see that for reasonable conditions and parameters the AVPF is well suited to support the feedback needed for NEWPRED. For more information about NEWPRED, see [8] and [9].8. Summary
The new RTP profile AVPF was investigated regarding performance and potential risks to the network stability. Simulations were conducted using the network simulator ns2, simulating unicast and several differently sized multicast topologies. The results were shown in this document. Regarding the network stability, it was important to show that the new profile does not lead to any feedback implosion or use more bandwidth than it is allowed. We measured the bandwidth that was used for RTCP in relation to the RTP session bandwidth. We have shown that, more or less exactly, 5% of the session bandwidth is used for RTCP, in all considered scenarios. Other RTCP bandwidth values could be set using the RTCP bandwidth modifiers [10]. The scenarios included unicast with and without errors, differently sized multicast groups, with and without errors or congestion on the links. Thus, we can say that the new profile behaves in a network-friendly manner in the sense that it uses only the allowed RTCP bandwidth, as defined by RTP. Secondly, we have shown that receivers using the new profile experience a performance gain. This was measured by capturing the delay that the sender sees for the received feedback. Using the new profile, this delay can be decreased by orders of magnitude.
In the third place, we investigated the effect of the parameter "l" on the new algorithms. We have shown that there does not exist an optimum value for it but only a trade-off can be achieved. The influence of this parameter is highly environment-specific and a trade-off between performance of the feedback suppression algorithm and the experienced delay has to be met. The recommended value of l=0.5 given in this document seems to be reasonable for most applications and environments.9. Security Considerations
This document describes the simulation work carried out to verify the correct working of the RTCP timing rules specified in the AVPF profile [1]. Consequently, security considerations concerning these timing rules are described in that document.
10. Normative References
[1] Ott, J., Wenger, S., Sato, N., Burmeister, C., and J. Rey, "Extended RTP Profile for Real-time Transport Control Protocol (RTCP)-Based Feedback (RTP/AVPF)", RFC 4585, July 2006.11. Informative References
[2] Schulzrinne, H., Casner, S., Frederick, R., and V. Jacobson, "RTP: A Transport Protocol for Real-Time Applications", STD 64, RFC 3550, July 2003. [3] Schulzrinne, H. and S. Casner, "RTP Profile for Audio and Video Conferences with Minimal Control", STD 65, RFC 3551, July 2003. [4] Network Simulator Version 2 - ns-2, available from http://www.isi.edu/nsnam/ns. [5] C. Burmeister, T. Klinner, "Low Delay Feedback RTCP - Timing Rules Simulation Results". Technical Report of the Panasonic European Laboratories, September 2001, available from: http://www.informatik.uni-bremen.de/~jo/misc/ SimulationResults-A.pdf. [6] ISO/IEC 14496-2:1999/Amd.1:2000, "Information technology - Coding of audio-visual objects - Part2: Visual", July 2000. [7] ITU-T Recommendation, H.263. Video encoding for low bitrate communication. 1998. [8] S. Fukunaga, T. Nakai, and H. Inoue, "Error Resilient Video Coding by Dynamic Replacing of Reference Pictures", IEEE Global Telecommunications Conference (GLOBECOM), pp.1503-1508, 1996. [9] H. Kimata, Y. Tomita, H. Yamaguchi, S. Ichinose, T. Ichikawa, "Receiver-Oriented Real-Time Error Resilient Video Communication System: Adaptive Recovery from Error Propagation in Accordance with Memory Size at Receiver", Electronics and Communications in Japan, Part 1, vol. 84, no. 2, pp.8-17, 2001. [10] Casner, S., "Session Description Protocol (SDP) Bandwidth Modifiers for RTP Control Protocol (RTCP) Bandwidth", RFC 3556, July 2003.
Authors' Addresses
Carsten Burmeister Panasonic R&D Center Germany GmbH Monzastr. 4c D-63225 Langen, Germany EMail: carsten.burmeister@eu.panasonic.com Rolf Hakenberg Panasonic R&D Center Germany GmbH Monzastr. 4c D-63225 Langen, Germany EMail: rolf.hakenberg@eu.panasonic.com Akihiro Miyazaki Matsushita Electric Industrial Co., Ltd 1006, Kadoma, Kadoma City, Osaka, Japan EMail: miyazaki.akihiro@jp.panasonic.com Joerg Ott Helsinki University of Technology, Networking Laboratory PO Box 3000, 02015 TKK, Finland EMail: jo@acm.org Noriyuki Sato Oki Electric Industry Co., Ltd. 1-16-8 Chuo, Warabi, Saitama 335-8510 Japan EMail: sato652@oki.com Shigeru Fukunaga Oki Electric Industry Co., Ltd. 2-5-7 Hommachi, Chuo-ku, Osaka 541-0053 Japan EMail: fukunaga444@oki.com
Full Copyright Statement Copyright (C) The Internet Society (2006). This document is subject to the rights, licenses and restrictions contained in BCP 78, and except as set forth therein, the authors retain all their rights. This document and the information contained herein are provided on an "AS IS" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY AND THE INTERNET ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. Intellectual Property The IETF takes no position regarding the validity or scope of any Intellectual Property Rights or other rights that might be claimed to pertain to the implementation or use of the technology described in this document or the extent to which any license under such rights might or might not be available; nor does it represent that it has made any independent effort to identify any such rights. Information on the procedures with respect to rights in RFC documents can be found in BCP 78 and BCP 79. Copies of IPR disclosures made to the IETF Secretariat and any assurances of licenses to be made available, or the result of an attempt made to obtain a general license or permission for the use of such proprietary rights by implementers or users of this specification can be obtained from the IETF on-line IPR repository at http://www.ietf.org/ipr. The IETF invites any interested party to bring to its attention any copyrights, patents or patent applications, or other proprietary rights that may cover technology that may be required to implement this standard. Please address the information to the IETF at ietf-ipr@ietf.org. Acknowledgement Funding for the RFC Editor function is provided by the IETF Administrative Support Activity (IASA).