| RMCAT WG | V. Singh |
| Internet-Draft | M. Nagy |
| Intended status: Experimental | J. Ott |
| Expires: January 5, 2015 | Aalto University |
| L. Eggert | |
| NetApp | |
| July 4, 2014 |
Congestion Control Using FEC for Conversational Media
draft-singh-rmcat-adaptive-fec-00
This document describes a new mechanism for conversational multimedia flows. The proposed mechanism uses Forward Error Correction (FEC) encoded RTP packets (redundant packets) along side the media packets to probe for available network capacity. A straightforward interpretation is, the sending endpoint increases the transmission rate by keeping the media rate constant but increases the amount of FEC. If no losses and discards occur, the endpoint can then increase the media rate. If losses occur, the redundant FEC packets help in recovering the lost packets. Consequently, the endpoint can vary the FEC bit rate to conservatively (by a small amount) or aggressively (by a large amount) probe for available network capacity.
This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.
Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet-Drafts is at http://datatracker.ietf.org/drafts/current/.
Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress."
This Internet-Draft will expire on January 5, 2015.
Copyright (c) 2014 IETF Trust and the persons identified as the document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (http://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Simplified BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Simplified BSD License.
The Real-time Transport Protocol (RTP) [RFC3550] is widely used in voice telephony and video conferencing systems. Many of these systems run over best-effort UDP/IP networks, and are required to implement congestion to adapt the transmission rate of the RTP streams to match the available network capacity, while maintaing the user-experience [I-D.ietf-rmcat-cc-requirements]. The circuit breakers [I-D.ietf-avtcore-rtp-circuit-breakers] describe a minimal set of conditions when an RTP stream is causing severe congestion and should cease transmission. Consequently, the congestion control algorithm are expected to avoid triggering these conditions.
Conversational multimedia systems use Negative Acknowlegment (NACK), Forward Error Correction (FEC), and Reference Picture Selection (RPS) to protect against packet loss. These are used in addition to the codec-dependent resilience methods (for e.g., full intra-refresh and error-concealment). In this way, the multimedia system is anyway trading off part of the transmission rate for redundancy or retransmissions to reduce the effects of packet loss. An endpoint often prefers using FEC in high latency networks where retransmissions may arrive later than the playout time of the packet (due to the size of the dejitter buffer) [Holmer13]. Therefore, the endpoint needs to adapt the transmission rate to best fit the changing network capacity and the amount of redundancy based on the observed/expected loss rate and network latency. Figure 1 shows the applicatbility of different error-resilience schemes based on the end-to-end latency and the observed packet loss [Devadoss08].
^
| .__________.
| | |
| | UEP/FEC |
l |____________|____. |
a | | | |
t | RPS | | |
e |_______. | | |
n | | | | |
c | | |____|_____|
y | NACK | |
| | |
+------------------------------->
Packet loss
Figure 1: Applicability of Error Resilience Schemes based on the network delay and observed packet loss
In this document, we describe the use of FEC packets not only for error-resilience but also as a probing mechanism for congestion control (ramping up the transmission rate).
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in BCP 14, [RFC2119] and indicate requirement levels for compliant implementations.
The terminology defined in RTP [RFC3550], RTP Profile for Audio and Video Conferences with Minimal Control [RFC3551], RTCP Extended Report (XR) [RFC3611], Extended RTP Profile for RTCP-based Feedback (RTP/AVPF) [RFC4585], RTP Retransmission Payload Format [RFC4588], Forward Error Correction (FEC) Framework [RFC6363], and Support for Reduced-Size RTCP [RFC5506] apply.
FEC is one method for providing error-resilience, it improves reliability by adding redundant data to the primary media flow, which is used by received to recover packets that have been lost due to congestion or bit-errors. The congestion control algorithm on the other hand aims at maximizing the network path utilization, but risks over-estimating the avaiable end-to-end network capacity leading to congestion (and therefore losses).
The main idea behind using FEC for congestion control is as follows: the sending endpoint chooses a high FEC rate to aggressively probe for available capacity and conversely chooses a low FEC rate to conservatively probe for available capacity. During the ramp up, if a packet is lost and the FEC packet arrives in time for decoding, the receiver is be able to recover the lost packet; if no packet is lost, the sender is able to increase the media encoding rate by swapping out a part of the FEC rate. This method can be especially useful when the transmission rate is close to the bottleneck link rate: by choosing an appropriate FEC rate, the endpoint is able to probe for available capacity without changing the target media rate and therefore not affecting the user-experience. Therefore, the congestion control algorithm is always able to probe for available capacity, as improved reliability compensates for possible errors resulting from overuse (i.e., increase in observed latency and/or losses).
+------------+ (B) Good conditions +-----------+
| |------------------------------------>| |
| STEADY | | PROBE |
| |<------------------------------------| |
+------------+ Probed, but Loss recovered +-----------+
/\ | | /\ |
| |(A) | | |
| |_______________________________________________| | |(C)
(B) | | (A) | |
| \/ (B) | \/
+------------+ +------------+
| | (A) Unstable conditions | |
| REDUCE |<------------------------------------| INCREASE |
| | | |
+------------+ +------------+
Figure 2: State machine of a Congestion Control enabling FEC.
The Figure 2 illustrates the the state machine of a congestion control algorithm incorporating FEC for probing. The state-machine includes 4 states: STEADY, PROBE, INCREASE, and REDUCE.
The Figure 3 shows the interaction between the rate control module, the RTP and the FEC module.
At the sender, the rate control module calculates the new bit rate. If the new bit rate is higher than the previous than the previous bit rate indicates to the FEC module that the congestion control intends to probe. The FEC module depending on its internal state machine decides to add FEC for probing or not. Thereafter it indicates to the rate control module the bit rate remaining for the RTP media stream, which may be less than equal to the calculated bit rate.
At the reciver, the FEC module reconstructs lost packets in the primary stream from the packets received in the repair stream. If packets are repair it generates the post-repair loss report (discussed in Section 3.3) for the corresponding RTP packets.
At the sender, The FEC module also receives the RTCP Feedback related to the primary stream and any post-repair loss report. It uses the information from these RTCP reports to calculate the effectiveness of FEC for congestion control and is also the basis for changing its internal state.
+ - - - - - - - - - - - - - - - - - - - - - - - -+
| +--------------------------------------------+ |
| Media Encoder/Decoder |
| +--------------------------------------------+ |
| |
| +- -- -- -- -- -- -- -+ +- -- -- -- -+ |
| Rate Control | | RTP |
| | Module | | Queue | |
+- -- -- -- -- -- -- -+ +- -- -- -- -+
| ^ | | |
| | | Source
| | R +--------------------+ | RTP |
| T | |
| | C | | |
| P | |
| | +----------+ +----------------+ |
| F | FEC Code |<--->| FEC Module |
| | B +----------+ +----------------+ |
| | | |
| |------------------------+ | | |
| RTCP FB Repair | | Source
| | RTP | | RTP |
| | |
| +--------------------------------------------+ |
| Transport Layer (UDP) |
| +--------------------------------------------+ |
|
| +--------------------------------------------+ |
| IP |
| +--------------------------------------------+ |
| Endpoint |
+ - - - - - - - - - - - - - - - - - - - - - - - +
Figure 3: Interaction of Congestion Control and FEC Module.
[RFC6363] describes a framework for using Forward Error Correction (FEC) codes with RTP and allows any FEC code to be used with the framework. For this proposal, the FEC packets are created by XORing RTP media packets, the resulting redundant RTP packets are encoded using the scheme defined in [RFC5109].
The endpoint MAY use a single-frame FEC or a multi-frame FEC for protecting the primary RTP stream. A single-frame FEC protects against a single packet loss and fails when burst loss occurs. Using multi-frame FEC helps mitigate these issues at the cost of higher overhead and latency in recovering lost packets. [Holmer13] shows examples of using a single- and multi-frame FEC.
The receiving endpoint may report the post-repair loss (or residual loss) using either the report block defined in [RFC5725] (Run-length encoding of packets repaired) or in [I-D.ietf-xrblock-rtcp-xr-post-repair-loss-count] (packet count of repaired packets).
[Open issue: The current implementation is delay based and is documented in [Nagy14]. However, we would like to generalize the concept and apply it to different RMCAT algorithms for e.g., Google's Congestion Control algorithm [I-D.alvestrand-rmcat-congestion], SCReaM [I-D.johansson-rmcat-scream-cc], etc.]
The security considerations of [RFC3550], RTP/AVPF profile for rapid RTCP feedback [RFC4585], circuit breaker [I-D.ietf-avtcore-rtp-circuit-breakers], and Generic Forward Error Correction [RFC5109] apply.
If non-authenticated RTCP reports are used, an on-path attacker can send forged RTCP feedback packets that can disrupt the operation of the underlying congestion control. Additionally, the forged packets can either indicate no packet loss causing the congestion control to ramp-up quickly, or indicate high packet loss or RTT causing the circuit breaker to trigger.
There are no IANA impacts in this memo.
This document is based on the results published in [Nagy14].
The work of Varun Singh, and Joerg Ott has been partially supported by the European Institute of Innovation and Technology (EIT) ICT Labs activity RCLD 11882. The views expressed here are those of the author(s) only. Neither the European Commission nor the EITICT labs is liable for any use that may be made of the information in this document.
| [RFC4588] | Rey, J., Leon, D., Miyazaki, A., Varsa, V. and R. Hakenberg, "RTP Retransmission Payload Format", RFC 4588, July 2006. |
| [RFC6363] | Watson, M., Begen, A. and V. Roca, "Forward Error Correction (FEC) Framework", RFC 6363, October 2011. |
| [I-D.ietf-rmcat-cc-requirements] | Jesup, R., "Congestion Control Requirements For RMCAT", Internet-Draft draft-ietf-rmcat-cc-requirements-01, December 2013. |
| [I-D.alvestrand-rmcat-congestion] | Holmer, S., Cicco, L., Mascolo, S. and H. Alvestrand, "A Google Congestion Control Algorithm for Real-Time Communication", Internet-Draft draft-alvestrand-rmcat-congestion-02, February 2014. |
| [I-D.johansson-rmcat-scream-cc] | Johansson, I. and Z. Sarker, "Self-Clocked Rate Adaptation for Multimedia", Internet-Draft draft-johansson-rmcat-scream-cc-02, June 2014. |
| [I-D.sarker-rmcat-eval-test] | Sarker, Z., Singh, V., Zhu, X. and M. Ramalho, "Test Cases for Evaluating RMCAT Proposals", Internet-Draft draft-sarker-rmcat-eval-test-00, February 2014. |
| [Nagy14] | Nagy, M., Singh, V., Ott, J. and L. Eggert, "Congestion Control using FEC for Conversational Multimedia Communication", Proc. of 5th ACM Internation Conference on Multimedia Systems (MMSys 2014) , 3 2014. |
| [Devadoss08] | Devadoss, J., Singh, V., Ott, J., Liu, C., Wang, Y-K. and I. Curcio, "Evaluation of Error Resilience Mechanisms for 3G Conversational Video", Proc. of IEEE International Symposium on Multimedia (ISM 2008) , 3 2014. |
| [Holmer13] | Holmer, S., Shemer, M. and M. Paniconi, "Handling Packet Loss in WebRTC", Proc. of IEEE International Conference on Image Processing (ICIP 2013) , 9 2013. |
This document is based on the results published in [Nagy14]. See the paper for ns-2 and testbed results; more results based on the scenarios listed in [I-D.sarker-rmcat-eval-test] will be published shorty.