draft-perkins-avtcore-rtp-circuit-breakers-00.txt   draft-perkins-avtcore-rtp-circuit-breakers-01.txt 
Network Working Group C. Perkins Network Working Group C. Perkins
Internet-Draft University of Glasgow Internet-Draft University of Glasgow
Intended status: Standards Track V. Singh Intended status: Standards Track V. Singh
Expires: September 5, 2012 Aalto University Expires: January 17, 2013 Aalto University
March 4, 2012 July 16, 2012
RTP Congestion Control: Circuit Breakers for Unicast Sessions RTP Congestion Control: Circuit Breakers for Unicast Sessions
draft-perkins-avtcore-rtp-circuit-breakers-00 draft-perkins-avtcore-rtp-circuit-breakers-01
Abstract Abstract
The Real-time Transport Protocol (RTP) is widely used for telephony, The Real-time Transport Protocol (RTP) is widely used in telephony,
video conferencing, and telepresence applications. These video conferencing, and telepresence applications. Such applications
applications are often used over best-effort UDP/IP networks. If are often run on best-effort UDP/IP networks. If congestion control
congestion control is not implemented then network congestion will is not implemented in the applications, then network congestion will
deteriorate the user's multimedia experience. This document does not deteriorate the user's multimedia experience. This document does not
propose a congestion control algorithm. Instead, it specifies a propose a congestion control algorithm; rather, it defines a minimal
minimal set of "circuit-breakers". Circuit-breakers are conditions set of "circuit-breakers". Circuit-breakers are conditions under
under which an RTP flow should cease to transmit media to protect the which an RTP flow is expected to stop transmiting media to protect
network from excessive congestion. It is expected that all RTP the network from excessive congestion. It is expected that all RTP
applications running on best-effort networks will be able to run applications running on best-effort networks will be able to run
without triggering these circuit breakers in normal operation. without triggering these circuit breakers in normal operation. Any
future RTP congestion control specification is expected to operate
within the envelope defined by these circuit breakers.
Status of this Memo Status of this Memo
This Internet-Draft is submitted in full conformance with the This Internet-Draft is submitted in full conformance with the
provisions of BCP 78 and BCP 79. provisions of BCP 78 and BCP 79.
Internet-Drafts are working documents of the Internet Engineering Internet-Drafts are working documents of the Internet Engineering
Task Force (IETF). Note that other groups may also distribute Task Force (IETF). Note that other groups may also distribute
working documents as Internet-Drafts. The list of current Internet- working documents as Internet-Drafts. The list of current Internet-
Drafts is at http://datatracker.ietf.org/drafts/current/. Drafts is at http://datatracker.ietf.org/drafts/current/.
Internet-Drafts are draft documents valid for a maximum of six months Internet-Drafts are draft documents valid for a maximum of six months
and may be updated, replaced, or obsoleted by other documents at any and may be updated, replaced, or obsoleted by other documents at any
time. It is inappropriate to use Internet-Drafts as reference time. It is inappropriate to use Internet-Drafts as reference
material or to cite them other than as "work in progress." material or to cite them other than as "work in progress."
This Internet-Draft will expire on September 5, 2012. This Internet-Draft will expire on January 17, 2013.
Copyright Notice Copyright Notice
Copyright (c) 2012 IETF Trust and the persons identified as the Copyright (c) 2012 IETF Trust and the persons identified as the
document authors. All rights reserved. document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal This document is subject to BCP 78 and the IETF Trust's Legal
Provisions Relating to IETF Documents Provisions Relating to IETF Documents
(http://trustee.ietf.org/license-info) in effect on the date of (http://trustee.ietf.org/license-info) in effect on the date of
publication of this document. Please review these documents publication of this document. Please review these documents
skipping to change at page 2, line 21 skipping to change at page 2, line 24
Table of Contents Table of Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 3 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 3
2. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . 3 2. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . 3
3. Background . . . . . . . . . . . . . . . . . . . . . . . . . . 3 3. Background . . . . . . . . . . . . . . . . . . . . . . . . . . 3
4. RTP Circuit Breakers for Systems Using the RTP/AVP Profile . . 6 4. RTP Circuit Breakers for Systems Using the RTP/AVP Profile . . 6
4.1. RTP/AVP Circuit Breaker #1: Timeout . . . . . . . . . . . 7 4.1. RTP/AVP Circuit Breaker #1: Timeout . . . . . . . . . . . 7
4.2. RTP/AVP Circuit Breaker #2: Congestion . . . . . . . . . . 8 4.2. RTP/AVP Circuit Breaker #2: Congestion . . . . . . . . . . 8
5. RTP Circuit Breakers for Systems Using the RTP/AVPF Profile . 10 5. RTP Circuit Breakers for Systems Using the RTP/AVPF Profile . 10
6. Impact of RTCP XR . . . . . . . . . . . . . . . . . . . . . . 10 6. Impact of RTCP XR . . . . . . . . . . . . . . . . . . . . . . 11
7. Impact of Explicit Congestion Notification (ECN) . . . . . . . 11 7. Impact of Explicit Congestion Notification (ECN) . . . . . . . 11
8. Session Timeout . . . . . . . . . . . . . . . . . . . . . . . 11 8. Session Timeout . . . . . . . . . . . . . . . . . . . . . . . 11
9. References . . . . . . . . . . . . . . . . . . . . . . . . . . 11 9. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 12
9.1. Normative References . . . . . . . . . . . . . . . . . . . 11 10. Security Considerations . . . . . . . . . . . . . . . . . . . 12
9.2. Informative References . . . . . . . . . . . . . . . . . . 12 11. Open Issues . . . . . . . . . . . . . . . . . . . . . . . . . 12
Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . . 13 12. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . 12
13. References . . . . . . . . . . . . . . . . . . . . . . . . . . 13
13.1. Normative References . . . . . . . . . . . . . . . . . . . 13
13.2. Informative References . . . . . . . . . . . . . . . . . . 13
Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . . 14
1. Introduction 1. Introduction
The Real-time Transport Protocol (RTP) [RFC3550] is widely used in The Real-time Transport Protocol (RTP) [RFC3550] is widely used in
voice-over-IP, video teleconferencing, and telepresence systems. voice-over-IP, video teleconferencing, and telepresence systems.
Many of these systems run over best-effort IP networks, and can Many of these systems run over best-effort UDP/IP networks, and can
suffer from packet loss and increased latency due to network suffer from packet loss and increased latency if network congestion
congestion. Designing effective RTP congestion control algorithms, occurs. Designing effective RTP congestion control algorithms, to
to adapt the transmission of RTP-based media to match the available adapt the transmission of RTP-based media to match the available
network capacity, while also maintaining the user experience, is a network capacity, while also maintaining the user experience, is a
difficult but important problem. Many such congestion control and difficult but important problem. Many such congestion control and
media adaptation algorithms have been proposed, but to date there is media adaptation algorithms have been proposed, but to date there is
no consensus on the correct approach, or even that a single standard no consensus on the correct approach, or even that a single standard
algorithm is desirable. algorithm is desirable.
This memo does not attempt to propose a new RTP congestion control This memo does not attempt to propose a new RTP congestion control
algorithm. Rather, it proposes a minimal set of "circuit breakers"; algorithm. Rather, it proposes a minimal set of "circuit breakers";
conditions under which there should be general agreement that an RTP conditions under which there is general agreement that an RTP flow is
flow is causing serious congestion, and should cease transmission. causing serious congestion, and ought to cease transmission. It is
It is expected that any future standards-track congestion control expected that future standards-track congestion control algorithms
algorithms for RTP will operate within the envelope defined by this for RTP will operate within the envelope defined by this memo.
memo.
2. Terminology 2. Terminology
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
"SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this
document are to be interpreted as described in RFC 2119 [RFC2119]. document are to be interpreted as described in RFC 2119 [RFC2119].
This interpretation of these key words applies only when written in
ALL CAPS. Mixed- or lower-case uses of these key words are not to be
interpreted as carrying special significance in this memo.
3. Background 3. Background
We consider congestion control for unicast RTP traffic flows. This We consider congestion control for unicast RTP traffic flows. This
is the problem of adapting the transmission of an audio/visual data is the problem of adapting the transmission of an audio/visual data
flow, encapsulated within an RTP transport session, from one sender flow, encapsulated within an RTP transport session, from one sender
to one receiver so that it matches the available network bandwidth. to one receiver, so that it matches the available network bandwidth.
Such adaptation must be done in a way that limits the disruption to Such adaptation needs to be done in a way that limits the disruption
the user experience caused by both packet loss and excessive rate to the user experience caused by both packet loss and excessive rate
changes. changes.
Congestion control for unicast RTP traffic can be implemented in one Congestion control for unicast RTP traffic can be implemented in one
of two places in the protocol stack. One approach is to run the RTP of two places in the protocol stack. One approach is to run the RTP
traffic over a congestion controlled transport protocol, for example traffic over a congestion controlled transport protocol, for example
over TCP, and to adapt the media encoding to match the dictates of over TCP, and to adapt the media encoding to match the dictates of
the transport-layer congestion control algorithm. This is safe for the transport-layer congestion control algorithm. This is safe for
the network, but may be suboptimal for the media quality unless the the network, but can be suboptimal for the media quality unless the
transport protocol is designed to support real-time media flows. We transport protocol is designed to support real-time media flows. We
do not consider this class of applications further in this memo, as do not consider this class of applications further in this memo, as
their network safety is guaranteed by the underlying transport. their network safety is guaranteed by the underlying transport.
Alternatively, RTP flows can be run over a non-congestion controlled Alternatively, RTP flows can be run over a non-congestion controlled
transport protocol, for example UDP, performing rate adaptation at transport protocol, for example UDP, performing rate adaptation at
the application layer based on RTP Control Protocol (RTCP) feedback. the application layer based on RTP Control Protocol (RTCP) feedback.
With a well-designed, network-aware, application, this allows highly With a well-designed, network-aware, application, this allows highly
effective media quality adaptation, but there is potential to disrupt effective media quality adaptation, but there is potential to disrupt
the network operation if the application does not adapt its sending the network's operation if the application does not adapt its sending
rate in a timely and effective manner. This memo focusses on this rate in a timely and effective manner. We consider this class of
class of application. applications in this memo.
Congestion control relies on monitoring the delivery of a media flow, Congestion control relies on monitoring the delivery of a media flow,
and responding to adapt the transmission of that flow when there are and responding to adapt the transmission of that flow when there are
signs that the network path is congested. Network congestion may be signs that the network path is congested. Network congestion can be
detected in one of three ways: 1) a receiver may infer the onset of detected in one of three ways: 1) a receiver can infer the onset of
congestion by observing an increase in one-way delay caused by queue congestion by observing an increase in one-way delay caused by queue
build-up within the network; 2) if Explicit Congestion Notification build-up within the network; 2) if Explicit Congestion Notification
(ECN) [RFC3168] is supported, the network may signal the presence of (ECN) [RFC3168] is supported, the network can signal the presence of
congestion by marking packets with ECN Congestion Experienced (CE) congestion by marking packets using ECN Congestion Experienced (CE)
marks; or 3) in the extreme case, congestion will cause packet loss, marks; or 3) in the extreme case, congestion will cause packet loss
which can be detected by observing a gap in the received RTP sequence that can be detected by observing a gap in the received RTP sequence
numbers. Once the onset of congestion is observed, the receiver must numbers. Once the onset of congestion is observed, the receiver has
send feedback to the sender to indicate that the transmission rate to send feedback to the sender to indicate that the transmission rate
should be reduced. How the sender reduces the transmission rate is needs to be reduced. How the sender reduces the transmission rate is
highly dependent on the media codec being used, and is outside the highly dependent on the media codec being used, and is outside the
scope of this memo. scope of this memo.
There are several ways in which a receiver may send feedback to a There are several ways in which a receiver can send feedback to a
media sender within the RTP framework: media sender within the RTP framework:
o The base RTP specification [RFC3550] defines RTCP Reception Report o The base RTP specification [RFC3550] defines RTCP Reception Report
(RR) packets to convey reception quality feedback information, and (RR) packets to convey reception quality feedback information, and
Sender Report (SR) packets to convey information about the media Sender Report (SR) packets to convey information about the media
transmission. RTCP SR packets contain data that can be used to transmission. RTCP SR packets contain data that can be used to
reconstruct media timing at a receiver, along with a count of the reconstruct media timing at a receiver, along with a count of the
total number of octets and packets sent. RTCP RR packets report total number of octets and packets sent. RTCP RR packets report
on the fraction of packets lost in the last reporting interval, on the fraction of packets lost in the last reporting interval,
the cumulative number of packets lost, the highest sequence number the cumulative number of packets lost, the highest sequence number
received, and the inter-arrival jitter. The RTCP RR packets also received, and the inter-arrival jitter. The RTCP RR packets also
contain timing information that allows the sender to estimate the contain timing information that allows the sender to estimate the
network round trip time (RTT) to the receivers. RTCP reports are network round trip time (RTT) to the receivers. RTCP reports are
sent periodically, with the reporting interval being determined by sent periodically, with the reporting interval being determined by
the number of participants in the session and a configured session the number of participants in the session and a configured session
bandwidth estimate. The interval between reports sent from each bandwidth estimate. The interval between reports sent from each
receiver tends to be on the order of a few seconds on average, and receiver tends to be on the order of a few seconds on average, and
it is randomised to avoid synchronisation of reports from multiple it is randomised to avoid synchronisation of reports from multiple
receivers. If a receiver detects problems, the base RTP receivers. RTCP RR packets allow a receiver to report ongoing
specification contains no provisions for sending the feedback network congestion to the sender. However, if a receiver detects
report early and must wait until the next scheduled reporting the onset of congestion partway through a reporting interval, the
interval. base RTP specification contains no provision for sending the RTCP
RR packet early, and the receiver has to wait until the next
scheduled reporting interval.
o The RTCP Extended Reports (XR) [RFC3611] allow reporting of more o The RTCP Extended Reports (XR) [RFC3611] allow reporting of more
complex and sophisticated reception quality metrics, but do not complex and sophisticated reception quality metrics, but do not
change the RTCP timing rules. RTCP extended reports of potential change the RTCP timing rules. RTCP extended reports of potential
interest for congestion control purposes are 1) the extended interest for congestion control purposes are the extended packet
packet loss, discard or burst/gap metrics for reacting based on loss, discard, and burst metrics [RFC3611], and
loss patterns; and 2) the end-system delay metrics for delay-based [I-D.ietf-xrblock-rtcp-xr-discard],
congestion control. [I-D.ietf-xrblock-rtcp-xr-discard-rle-metrics],
[I-D.ietf-xrblock-rtcp-xr-burst-gap-discard],
[I-D.ietf-xrblock-rtcp-xr-burst-gap-loss]; and the extended delay
metrics [I-D.ietf-xrblock-rtcp-xr-delay],
[I-D.ietf-xrblock-rtcp-xr-pdv]. Other RTCP Extended Reports that
could be helpful for congestion control purposes might be
developed in future.
o Rapid feedback about the occurrence of congestion events can be o Rapid feedback about the occurrence of congestion events can be
achieved using the Extended RTP Profile for RTCP-Based Feedback achieved using the Extended RTP Profile for RTCP-Based Feedback
(RTP/AVPF) [RFC4585]. This modifies the RTCP timing rules to (RTP/AVPF) [RFC4585] in place of the more common RTP/AVP profile
allow RTCP reports to be sent early, in some cases immediately, [RFC3551]. This modifies the RTCP timing rules to allow RTCP
provided the average RTCP reporting interval remains unchanged. reports to be sent early, in some cases immediately, provided the
It also defines new transport-layer feedback messages, including average RTCP reporting interval remains unchanged. It also
negative acknowledgements (NACKs), that can be used to report on defines new transport-layer feedback messages, including negative
specific congestion events. The use of the RTP/AVPF profile is acknowledgements (NACKs), that can be used to report on specific
dependent on signalling, but is otherwise generally backwards congestion events. The use of the RTP/AVPF profile is dependent
compatible, as it keeps the same average RTCP reporting interval on signalling, but is otherwise generally backwards compatible, as
as the base RTP specification. The Codec control messages it keeps the same average RTCP reporting interval as the base RTP
[RFC5104] extend the RTP/AVPF profile with additional feedback specification. The RTP Codec Control Messages [RFC5104] extend
message that can be used to influence that way in which rate the RTP/AVPF profile with additional feedback messages that can be
adaptation occurs. The dynamics of how rapidly feedback can be used to influence that way in which rate adaptation occurs. The
sent are unchanged. dynamics of how rapidly feedback can be sent are unchanged.
o Finally, the RTP and RTCP extensions for Explicit Congestion o Finally, the RTP and RTCP extensions for Explicit Congestion
Notification (ECN) [I-D.ietf-avtcore-ecn-for-rtp] can be used to Notification (ECN) [I-D.ietf-avtcore-ecn-for-rtp] can be used to
provide feedback on the number of packets that received an ECN provide feedback on the number of packets that received an ECN
Congestion Experienced (CE) mark. This extension builds on the Congestion Experienced (CE) mark. This extension builds on the
RTP/AVPF profile to allow rapid congestion feedback. RTP/AVPF profile to allow rapid congestion feedback when ECN is
supported.
In addition to these mechanisms for providing feedback, the sender In addition to these mechanisms for providing feedback, the sender
can include an RTP header extension in each packet to record packet can include an RTP header extension in each packet to record packet
transmission times. There are two methods: [RFC5450] represents the transmission times. There are two methods: [RFC5450] represents the
transmission time in terms of a time-offset from the RTP timestamp of transmission time in terms of a time-offset from the RTP timestamp of
the packet, while [RFC6051] includes an explicit NTP-format sending the packet, while [RFC6051] includes an explicit NTP-format sending
timestamp (potentially more accurate, but a higher header overhead). timestamp (potentially more accurate, but a higher header overhead).
Accurate sending timestamps can be helpful for estimating queuing Accurate sending timestamps can be helpful for estimating queuing
delays, to get an early indication of the onset of congestion. delays, to get an early indication of the onset of congestion.
Taken together, these various mechanisms allow receivers to provide Taken together, these various mechanisms allow receivers to provide
feedback on the senders when congestion events occur, with varying feedback on the senders when congestion events occur, with varying
degrees of timeliness and accuracy. The key distinction is between degrees of timeliness and accuracy. The key distinction is between
systems that use only the basic RTCP mechanisms, without RTP/AVPF systems that use only the basic RTCP mechanisms, without RTP/AVPF
rapid feedback, and those that use the RTP/AVPF extensions, and can rapid feedback, and those that use the RTP/AVPF extensions and so can
respond to congestion more rapidly. respond to congestion more rapidly.
4. RTP Circuit Breakers for Systems Using the RTP/AVP Profile 4. RTP Circuit Breakers for Systems Using the RTP/AVP Profile
The feedback mechanisms defined in [RFC3550] are the minimum that can The feedback mechanisms defined in [RFC3550] and available under the
be required for a baseline circuit breaker mechanism suitable for all RTP/AVP profile [RFC3551] are the minimum that can be assumed for a
unicast applications of RTP. Accordingly, for an RTP circuit breaker baseline circuit breaker mechanism that is suitable for all unicast
to be useful, it should be able to detect that an RTP flow is causing applications of RTP. Accordingly, for an RTP circuit breaker to be
useful, it needs to be able to detect that an RTP flow is causing
excessive congestion using only basic RTCP features, without needing excessive congestion using only basic RTCP features, without needing
RTCP XR feedback or the RTP/AVPF profile for rapid RTCP reports. RTCP XR feedback or the RTP/AVPF profile for rapid RTCP reports.
Three potential congestion signals are available from the basic RTCP Three potential congestion signals are available from the basic RTCP
SR/RR packets, and are reported for each SSRC in the RTP session: SR/RR packets and are reported for each synchronisation source (SSRC)
in the RTP session:
1. The sender can estimate the network round-trip time once per RTCP 1. The sender can estimate the network round-trip time once per RTCP
reporting interval, based on the contents and timing of RTCP SR reporting interval, based on the contents and timing of RTCP SR
and RR packets. and RR packets.
2. Receivers report estimated jitter (the statistical variance of 2. Receivers report a jitter estimate (the statistical variance of
the RTP data packet inter-arrival time) calculated over the RTCP the RTP data packet inter-arrival time) calculated over the RTCP
reporting interval. Due to the nature of the jitter calculation reporting interval. Due to the nature of the jitter calculation
([RFC3550], section 6.4.4), the jitter is only meaningful for RTP ([RFC3550], section 6.4.4), the jitter is only meaningful for RTP
flows that send a single data packet for each RTP timestamp value flows that send a single data packet for each RTP timestamp value
(i.e., audio flows, or video flows where each frame comprises one (i.e., audio flows, or video flows where each frame comprises one
RTP packet). RTP packet).
3. Receivers report the fraction of packets lost during the RTCP 3. Receivers report the fraction of RTP data packets lost during the
reporting interval, and the cumulative number of packets lost RTCP reporting interval, and the cumulative number of RTP packets
over the entire RTP session. lost over the entire RTP session.
These congestion signals limit the possible circuit breakers, since These congestion signals limit the possible circuit breakers, since
they give only limited visibility into the behaviour of the network. they give only limited visibility into the behaviour of the network.
RTT estimates are widely used in congestion control algorithms, as a RTT estimates are widely used in congestion control algorithms, as a
proxy for queuing delay measures in delay-based congestion control, proxy for queuing delay measures in delay-based congestion control or
or to determine connection timeouts. RTT estimates derived from RTCP to determine connection timeouts. RTT estimates derived from RTCP SR
SR and RR packets send according to the RTP/AVP timing rules are too and RR packets sent according to the RTP/AVP timing rules are far too
infrequent to be useful though, and don't give enough information to infrequent to be useful though, and don't give enough information to
distinguish a delay change due to routing updates from queuing delay distinguish a delay change due to routing updates from queuing delay
caused by congestion. Accordingly, we do not use the RTT estimate caused by congestion. Accordingly, we do not use the RTT estimate
alone as an RTP circuit breaker. alone as an RTP circuit breaker.
Increased jitter can be a signal of transient network congestion, but Increased jitter can be a signal of transient network congestion, but
in the highly aggregated form reported in RTCP RR packets, it offers in the highly aggregated form reported in RTCP RR packets, it offers
insufficient information to estimate the extent or persistence of insufficient information to estimate the extent or persistence of
congestion. Jitter reports are a useful early warning of potential congestion. Jitter reports are a useful early warning of potential
network congestion, but provide an insufficiently strong signal to be network congestion, but provide an insufficiently strong signal to be
used as a circuit breaker. used as a circuit breaker.
The remaining congestion signals are the packet loss fraction and the The remaining congestion signals are the packet loss fraction and the
cumulative number of packets lost. These are robust indicators of cumulative number of packets lost. These are robust indicators of
congestion in networks where packet loss is primarily due to queue congestion in a network where packet loss is primarily due to queue
overflows, although less accurate in networks where losses can be overflows, although less accurate in networks where losses can be
caused by non-congestive packet corruption. TCP also uses packet caused by non-congestive packet corruption. TCP uses packet loss as
loss as a congestion signal. a congestion signal.
Two packet loss regimes can be observed: 1) RTCP RR packets show a Two packet loss regimes can be observed: 1) RTCP RR packets show a
non-zero packet loss fraction, while the extended highest sequence non-zero packet loss fraction, while the extended highest sequence
number received continues to increment; and 2) RR packets show a loss number received continues to increment; and 2) RR packets show a loss
fraction of zero, but the extended highest sequence number received fraction of zero, but the extended highest sequence number received
does not increment even though the sender has been transmitting RTP does not increment even though the sender has been transmitting RTP
data packets. The former corresponds to the TCP congestion avoidance data packets. The former corresponds to the TCP congestion avoidance
state, and indicates a congested path that is still delivering data; state, and indicates a congested path that is still delivering data;
the latter corresponds to a TCP timeout, and is most likely due to a the latter corresponds to a TCP timeout, and is most likely due to a
path failure. We derive circuit breaker conditions for these two path failure. We derive circuit breaker conditions for these two
skipping to change at page 7, line 31 skipping to change at page 7, line 46
4.1. RTP/AVP Circuit Breaker #1: Timeout 4.1. RTP/AVP Circuit Breaker #1: Timeout
If RTP data packets are being sent while the corresponding RTCP RR If RTP data packets are being sent while the corresponding RTCP RR
packets report a non-increasing extended highest sequence number packets report a non-increasing extended highest sequence number
received, this is an indication that those RTP data packets are not received, this is an indication that those RTP data packets are not
reaching the receiver. This could be a short-term issue affecting reaching the receiver. This could be a short-term issue affecting
only a few packets, perhaps caused by a slow-to-open firewall or a only a few packets, perhaps caused by a slow-to-open firewall or a
transient connectivity problem, but if the issue persists, it is a transient connectivity problem, but if the issue persists, it is a
sign of a more ongoing and significant problem. Accordingly, if a sign of a more ongoing and significant problem. Accordingly, if a
sender of RTP data packets receives two or more consecutive RTCP RR sender of RTP data packets receives two or more consecutive RTCP RR
packets from the same receiver that correspond to its transmission packets from the same receiver that correspond to its transmission,
and have a non-increasing extended highest sequence number received and have a non-increasing extended highest sequence number received
field, then that sender SHOULD cease transmission. field (i.e., at least three RTCP RR packets that report the same
value in the extended highest sequence number received field, when
the sender has sent data packets that would have caused an increase
in the reported value of the extended highest sequence number
received if they had reached the receiver), then that sender SHOULD
cease transmission.
Systems that usually send at a high data rate, but which can reduce Systems that usually send at a high data rate, but which can reduce
their data rate significantly (i.e., by an order of magnitude), MAY their data rate significantly (i.e., by at least a factor of ten),
first reduce their sending rate to this lower value, but MUST then MAY first reduce their sending rate to this lower value to see if
cease transmission if the problem does not resolve itself within a this resolves the congestion, but MUST then cease transmission if the
further two RTCP reporting intervals. An example of this might be a problem does not resolve itself within a further two RTCP reporting
video conferencing system that backs off to sending audio only, intervals. An example of this might be a video conferencing system
before completely dropping the call. that backs off to sending audio only, before completely dropping the
call. If such a reduction in sending rate resolves the congestion
problem, the sender MAY gradually increase the rate at which it sends
data after a reasonable amount of time has passed, provided it takes
care not to cause the problem to recur ("reasonable" is intentionally
not defined here).
The choice of two RTCP reporting intervals is to give enough time for The choice of two RTCP reporting intervals is to give enough time for
transient problems to resolve themselves, but to stop problem flows transient problems to resolve themselves, but to stop problem flows
quickly enough to avoid causing serious problems. A single RTCP quickly enough to avoid causing serious ongoing network congestion.
report showing no reception could be caused by numerous transient A single RTCP report showing no reception could be caused by numerous
faults, and so should not stop transmission. More than two RTCP transient faults, and so will not cease transmission. Waiting for
reports could avoid false positives, but would lead to problematic more than two RTCP reports before stopping a flow might avoid some
flows running for a long time before being cut off. false positives, but would lead to problematic flows running for a
long time before being cut off.
4.2. RTP/AVP Circuit Breaker #2: Congestion 4.2. RTP/AVP Circuit Breaker #2: Congestion
If RTP data packets are being sent, and the corresponding RTCP RR If RTP data packets are being sent, and the corresponding RTCP RR
packets show non-zero packet loss fraction and increasing extended packets show non-zero packet loss fraction and increasing extended
highest sequence number received, then the RTP data packets are highest sequence number received, then the RTP data packets are
arriving at the receiver, but some degree of congestion is occurring. arriving at the receiver, but some degree of congestion is occurring.
The RTP/AVP profile [RFC3551] states that: The RTP/AVP profile [RFC3551] states that:
If best-effort service is being used, RTP receivers SHOULD monitor If best-effort service is being used, RTP receivers SHOULD monitor
skipping to change at page 8, line 34 skipping to change at page 9, line 10
The comparison to TCP cannot be specified exactly, but is intended The comparison to TCP cannot be specified exactly, but is intended
as an "order-of-magnitude" comparison in timescale and throughput. as an "order-of-magnitude" comparison in timescale and throughput.
The timescale on which TCP throughput is measured is the round- The timescale on which TCP throughput is measured is the round-
trip time of the connection. In essence, this requirement states trip time of the connection. In essence, this requirement states
that it is not acceptable to deploy an application (using RTP or that it is not acceptable to deploy an application (using RTP or
any other transport protocol) on the best-effort Internet which any other transport protocol) on the best-effort Internet which
consumes bandwidth arbitrarily and does not compete fairly with consumes bandwidth arbitrarily and does not compete fairly with
TCP within an order of magnitude. TCP within an order of magnitude.
(The phase "order of magnitude" in the above means a factor of ten).
The throughput of a long-lived TCP connection can be estimated using The throughput of a long-lived TCP connection can be estimated using
the TCP throughput equation: the TCP throughput equation:
s s
X = -------------------------------------------------------------- X = --------------------------------------------------------------
R*sqrt(2*b*p/3) + (t_RTO * (3*sqrt(3*b*p/8) * p * (1+32*p^2))) R*sqrt(2*b*p/3) + (t_RTO * (3*sqrt(3*b*p/8) * p * (1+32*p^2)))
Where: Where:
X is the transmit rate in bytes/second. X is the transmit rate in bytes/second.
s is the packet size in bytes. If the RTP data packets vary in s is the packet size in bytes. If the RTP data packets vary in
size, then the average size should be used. size, then the average size is to be used.
R is the round trip time in seconds. R is the round trip time in seconds.
p is the loss event rate, between 0 and 1.0, of the number of loss p is the loss event rate, between 0 and 1.0, of the number of loss
events as a fraction of the number of packets transmitted. events as a fraction of the number of packets transmitted.
t_RTO is the TCP retransmission timeout value in seconds, t_RTO is the TCP retransmission timeout value in seconds,
approximated by setting t_RTO = 4*R. approximated by setting t_RTO = 4*R.
b is the number of packets acknowledged by a single TCP b is the number of packets acknowledged by a single TCP
acknowledgement ([RFC3448] recommends the use of b=1 since many acknowledgement ([RFC3448] recommends the use of b=1 since many
TCP implementations do not use delayed acknowledgements). TCP implementations do not use delayed acknowledgements).
This is the same approach to estimated TCP throughput that is used in This is the same approach to estimated TCP throughput that is used in
[RFC3448]. Two parameters must be estimated in order to calculate [RFC3448]. Under conditions of low packet loss, this formula can be
the throughput: the round trip time, R, and the loss event rate, p. approximated as follows with reasonable accuracy:
The round trip time can be estimated from RTCP. This is done too
infrequently for accurate statistics, but is the best that can be s
done with the standard RTCP mechanisms. X = ---------------
R * sqrt(p*2/3)
It is RECOMMENDED that this simplified throughout equation be used,
since the reduction in accuracy is small, and it is much simpler to
calculate than the full equation.
Given this TCP equation, two parameters need to be estimated in order
to calculate the throughput: the round trip time, R, and the loss
event rate, p (the packet size, s, is known to the sender). The
round trip time can be estimated from RTCP SR and RR packets. This
is done too infrequently for accurate statistics, but is the best
that can be done with the standard RTCP mechanisms.
RTCP RR packets contain the packet loss fraction, rather than the RTCP RR packets contain the packet loss fraction, rather than the
loss event rate, so p cannot be reported (TCP typically treats the loss event rate, so p cannot be reported (TCP typically treats the
loss of multiple packets within a single RTT as one loss event, but loss of multiple packets within a single RTT as one loss event, but
RTCP RR packets report the overall fraction of packets lost, not RTCP RR packets report the overall fraction of packets lost, not
caring about when the losses occurred). Using the loss fraction in caring about when the losses occurred). Using the loss fraction in
place of the loss event rate can overestimate the loss. We believe place of the loss event rate can overestimate the loss. We believe
that this overestimate will not be significant, given that we are that this overestimate will not be significant, given that we are
only interested in order of magnitude comparison (Floyd et al, only interested in order of magnitude comparison (Floyd et al,
"Equation-Based Congestion Control for Unicast Applications", Proc. "Equation-Based Congestion Control for Unicast Applications", Proc.
SIGCOMM 2000, section 3.2.1, show that the difference is small for SIGCOMM 2000, section 3.2.1, show that the difference is small for
steady-state conditions and random loss, but using the loss fraction steady-state conditions and random loss, but using the loss fraction
is more conservative in the case of bursty loss). is more conservative in the case of bursty loss).
The congestion circuit breaker is therefore: when RTCP RR packets are The congestion circuit breaker is therefore: when RTCP RR packets are
received, estimate the TCP throughput using the above equation and received, estimate the TCP throughput using the simplified equation
the measured R, p (approximated by the loss fraction), and s. above, and the measured R, p (approximated by the loss fraction), and
Compare this with the actual sending rate. If the actual sending s. Compare this with the actual sending rate. If the actual sending
rate has been more than an order of magnitude greater than the rate has been more than a factor of ten greater than the throughput
throughput equation estimate for two or more RTCP reporting equation estimate for two or more RTCP reporting intervals, stop
intervals, stop transmitting. transmitting.
Again, we use two reporting intervals to avoid triggering the circuit Again, we use two reporting intervals to avoid triggering the circuit
breaker on transient failures. This circuit breaker is a worst-case breaker on transient failures. This circuit breaker is a worst-case
condition, and congestion control should be performed to keep well condition, and congestion control needs to be performed to keep well
within this bound. It is expected that the circuit breaker will only within this bound. It is expected that the circuit breaker will only
be triggered if the usual congestion control fails for some reason. be triggered if the usual congestion control fails for some reason.
(tbd -- we need to base the circuit breaker condition on something,
so TCP seems a logical choice. Following TCP limits too closely is
inappropriate for many applications of RTP, though, since they have
different dynamics. Is the above lax enough to not disrupt valid
applications, but tight enough to provide meaningful protection for
the network?)
5. RTP Circuit Breakers for Systems Using the RTP/AVPF Profile 5. RTP Circuit Breakers for Systems Using the RTP/AVPF Profile
More rapid feedback allows more responsiveness. The receiver SHOULD More rapid feedback allows more responsiveness. The receiver SHOULD
provide feedback more often during, or at onset of, congestion, and provide feedback more often during, or at onset of, congestion, and
provide feedback less often when there is no congestion. provide feedback less often when there is no congestion.
(tbd -- mechanisms may probably need to be designed in conjunction (tbd -- mechanisms probably need to be designed in conjunction with
with the different classes of congestion control that can leverage the different classes of congestion control that can leverage RTP/
RTP/AVPF; e.g., we might need to specify limits for TFRC-like or AVPF; e.g., we might need to specify limits for TFRC-like or delay-
delay-based algorithms using RTP/AVPF feedback.) based algorithms using RTP/AVPF feedback.)
(tbd -- a high-level question to be answered is whether we need to (tbd -- a high-level question to be answered is whether we need to
specify anything different for the circuit breaker for AVPF, or if we specify anything different for the circuit breaker for AVPF, or if we
leave that unchanged, and focus solely on the dynamics, to ensure the leave that unchanged, and focus solely on the dynamics, to ensure the
circuit breaker is never triggered.) circuit breaker is never triggered.)
6. Impact of RTCP XR 6. Impact of RTCP XR
(tbd) (tbd)
This improves the information, but doesn't change the dynamics of the This improves the information, but doesn't change the dynamics of the
congestion control loop. Suspect the impact will actually be quite congestion control loop. Suspect the impact will actually be quite
small. small.
Packets discarded [I-D.ietf-xrblock-rtcp-xr-discard] or bytes Packets discarded [I-D.ietf-xrblock-rtcp-xr-discard] or bytes
discarded [I-D.ietf-xrblock-rtcp-xr-discard-rle-metrics] due to late discarded [I-D.ietf-xrblock-rtcp-xr-discard-rle-metrics] due to late
arrival by the receiver may indicate congestion. Congestion control arrival by the receiver might indicate congestion. Congestion
should consider the discarded packets as if they were lost packets. control needs to consider the discarded packets as if they were lost
packets.
The RTCP RR reports the loss fraction over an RTCP interval which is The RTCP RR reports the loss fraction over an RTCP interval which is
insufficient to distinguish between solitary or bursty losses. To insufficient to distinguish between solitary or bursty losses. To
provide rough sense of duration of losses or discards, an endpoint provide rough sense of duration of losses or discards, an endpoint
may use burst/gap reporting for loss can use burst/gap reporting for loss
[I-D.ietf-xrblock-rtcp-xr-burst-gap-loss] and discard [I-D.ietf-xrblock-rtcp-xr-burst-gap-loss] and discard
[I-D.ietf-xrblock-rtcp-xr-burst-gap-discard]. For more accurate [I-D.ietf-xrblock-rtcp-xr-burst-gap-discard]. For more accurate
reporting the receiver may use Run-length encoded (RLE) lost reporting the receiver can use Run-length encoded (RLE) lost
[RFC3611] or discarded [I-D.ietf-xrblock-rtcp-xr-discard-rle-metrics] [RFC3611] or discarded [I-D.ietf-xrblock-rtcp-xr-discard-rle-metrics]
packets. packets.
For precise measurement of network roundtrip delay the receiver can For precise measurement of network roundtrip delay the receiver can
signal its end-system delay [I-D.ietf-xrblock-rtcp-xr-delay] signal its end-system delay [I-D.ietf-xrblock-rtcp-xr-delay]
[RFC3611]. [RFC3611].
A receiver may also indicate onset or end of congestion by reporting A receiver can also indicate onset or end of congestion by reporting
the distribution of the inter-packet delay variation the distribution of the inter-packet delay variation
[I-D.ietf-xrblock-rtcp-xr-pdv] [RFC3611]. [I-D.ietf-xrblock-rtcp-xr-pdv] [RFC3611].
7. Impact of Explicit Congestion Notification (ECN) 7. Impact of Explicit Congestion Notification (ECN)
ECN-CE marked packets SHOULD be treated as if it were lost for the ECN-CE marked packets SHOULD be treated as if it were lost for the
purposes of congestion control, when determining the optimal rate at purposes of congestion control, when determining the optimal rate at
which to send. However, it seems unwise to treat the receipt of which to send. However, it seems unwise to treat the receipt of
multiple ECN-CE marked packets as a circuit breaker, since it is multiple ECN-CE marked packets as a circuit breaker, since it is
likely that ECN-capable and non-ECN-capable paths will exist for a likely that ECN-capable and non-ECN-capable paths will exist for a
long time to come. Rather, consider packet loss as the circuit long time to come. Rather, consider packet loss as the circuit
breaker condition as for non-ECN flows. breaker condition as for non-ECN flows.
8. Session Timeout 8. Session Timeout
From a usability perspective, if there is no audio or video response From a usability perspective, if there is no audio or video response
from the other peer, it is likely that the user may terminate the from the other peer, it is likely that the user will terminate the
session. session.
According to RFC 3550 [RFC3550], any participant that has not sent an According to RFC 3550 [RFC3550], any participant that has not sent an
RTP packet within the last two RTCP interval is removed from the RTP packet within the last two RTCP interval is removed from the
sender list. To avoid timing out the specific flow, the endpoint sender list. To avoid timing out the specific flow, the endpoint
MUST send corresponding RTCP reports. Interactive Connectivity MUST send corresponding RTCP reports. Interactive Connectivity
Establishment (ICE) [RFC5245] recommends that the timeout MUST NOT be Establishment (ICE) [RFC5245] recommends that the timeout MUST NOT be
less than 15 seconds. less than 15 seconds.
If no RTCP RR arrives for two complete SR intervals, the sender If no RTCP RR arrives for two complete SR intervals, the sender
SHOULD cease transmission. However, if the endpoint can reduce the SHOULD cease transmission. However, if the endpoint can reduce the
media rate then it MAY first reduce the rate to the lower value, but media rate then it MAY first reduce the rate to the lower value, but
terminate the transmission if still no RTCP RR is received in the terminate the transmission if still no RTCP RR is received in the
next two SR intervals. next two SR intervals.
9. References 9. IANA Considerations
9.1. Normative References There are no actions for IANA.
10. Security Considerations
(tbd: Security considerations: how to protect against fake RTCP
reports being used to force sessions to close? SRTCP is one option,
but are there any lighter weight options?)
11. Open Issues
o Clarify: when will the recipient end a call, if it receives no
data?
o When we say "cease transmission", do we need some minimum interval
before we're allowed to restart?
o What does "cease transmission" mean? Do we send an RTCP BYE and
leave the session, or is it more temporary than that?
o Add a receiver-based circuit-breaker condition. Note that this is
dependent on the signalling still working, since the receiver
needs to be able to inform the sender.
12. Acknowledgements
The authors would like to thank Harald Alvestrand, Randell Jesup, and
Abheek Saha for their valuable feedback.
13. References
13.1. Normative References
[RFC2119] Bradner, S., "Key words for use in RFCs to Indicate [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate
Requirement Levels", BCP 14, RFC 2119, March 1997. Requirement Levels", BCP 14, RFC 2119, March 1997.
[RFC3448] Handley, M., Floyd, S., Padhye, J., and J. Widmer, "TCP [RFC3448] Handley, M., Floyd, S., Padhye, J., and J. Widmer, "TCP
Friendly Rate Control (TFRC): Protocol Specification", Friendly Rate Control (TFRC): Protocol Specification",
RFC 3448, January 2003. RFC 3448, January 2003.
[RFC3550] Schulzrinne, H., Casner, S., Frederick, R., and V. [RFC3550] Schulzrinne, H., Casner, S., Frederick, R., and V.
Jacobson, "RTP: A Transport Protocol for Real-Time Jacobson, "RTP: A Transport Protocol for Real-Time
Applications", STD 64, RFC 3550, July 2003. Applications", STD 64, RFC 3550, July 2003.
[RFC3551] Schulzrinne, H. and S. Casner, "RTP Profile for Audio and [RFC3551] Schulzrinne, H. and S. Casner, "RTP Profile for Audio and
Video Conferences with Minimal Control", STD 65, RFC 3551, Video Conferences with Minimal Control", STD 65, RFC 3551,
July 2003. July 2003.
[RFC3611] Friedman, T., Caceres, R., and A. Clark, "RTP Control [RFC3611] Friedman, T., Caceres, R., and A. Clark, "RTP Control
Protocol Extended Reports (RTCP XR)", RFC 3611, Protocol Extended Reports (RTCP XR)", RFC 3611,
November 2003. November 2003.
skipping to change at page 12, line 21 skipping to change at page 13, line 33
[RFC3611] Friedman, T., Caceres, R., and A. Clark, "RTP Control [RFC3611] Friedman, T., Caceres, R., and A. Clark, "RTP Control
Protocol Extended Reports (RTCP XR)", RFC 3611, Protocol Extended Reports (RTCP XR)", RFC 3611,
November 2003. November 2003.
[RFC4585] Ott, J., Wenger, S., Sato, N., Burmeister, C., and J. Rey, [RFC4585] Ott, J., Wenger, S., Sato, N., Burmeister, C., and J. Rey,
"Extended RTP Profile for Real-time Transport Control "Extended RTP Profile for Real-time Transport Control
Protocol (RTCP)-Based Feedback (RTP/AVPF)", RFC 4585, Protocol (RTCP)-Based Feedback (RTP/AVPF)", RFC 4585,
July 2006. July 2006.
9.2. Informative References 13.2. Informative References
[I-D.ietf-avtcore-ecn-for-rtp] [I-D.ietf-avtcore-ecn-for-rtp]
Westerlund, M., Johansson, I., Perkins, C., and K. Westerlund, M., Johansson, I., Perkins, C., and K.
Carlberg, "Explicit Congestion Notification (ECN) for RTP Carlberg, "Explicit Congestion Notification (ECN) for RTP
over UDP", draft-ietf-avtcore-ecn-for-rtp-06 (work in over UDP", draft-ietf-avtcore-ecn-for-rtp-06 (work in
progress), February 2012. progress), February 2012.
[I-D.ietf-xrblock-rtcp-xr-burst-gap-discard] [I-D.ietf-xrblock-rtcp-xr-burst-gap-discard]
Clark, A., Hunt, G., Wu, W., and R. Huang, "RTCP XR Report Clark, A., Hunt, G., Wu, W., and R. Huang, "RTCP XR Report
Block for Burst/Gap Discard metric Reporting", Block for Burst/Gap Discard metric Reporting",
 End of changes. 49 change blocks. 
129 lines changed or deleted 196 lines changed or added

This html diff was produced by rfcdiff 1.33. The latest version is available from http://tools.ietf.org/tools/rfcdiff/