Network Working Group X. Zhu
Internet Draft R. Pan
Intended Status: Informational Cisco Systems
Expires: September 13, 2013 March 12, 2013
NADA: A Unified Congestion Control Scheme for Real-Time Media
draft-zhu-rmcat-nada-01
Abstract
This document describes a scheme named network-assisted dynamic
adaptation (NADA), a novel congestion control approach for
interactive real-time media applications, such as video conferencing.
In the proposed scheme, the sender regulates its sending rate based
on either implicit or explicit congestion signaling, in a unified
approach. The scheme can benefit from explicit congestion
notification (ECN) markings from network nodes. It also maintains
consistent sender behavior in the absence of such markings, by
reacting to queuing delays and packet losses instead.
We present here the overall system architecture, recommended
behaviors at the sender and the receiver, as well as expected network
node operations. Results from extensive simulation studies of the
proposed scheme are available upon request.
Status of this Memo
This Internet-Draft is submitted to IETF in full conformance with the
provisions of BCP 78 and BCP 79.
Internet-Drafts are working documents of the Internet Engineering
Task Force (IETF), its areas, and its working groups. Note that
other groups may also distribute working documents as
Internet-Drafts.
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."
The list of current Internet-Drafts can be accessed at
http://www.ietf.org/1id-abstracts.html
The list of Internet-Draft Shadow Directories can be accessed at
http://www.ietf.org/shadow.html
Zhu & Pan Expires September 13, 2013 [Page 1]
INTERNET DRAFT NADA March 12, 2013
Copyright and License Notice
Copyright (c) 2012 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.
Table of Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 3
2. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . . 3
3. System Model . . . . . . . . . . . . . . . . . . . . . . . . . 3
4. Network Node Operations . . . . . . . . . . . . . . . . . . . . 4
4.1 Default behavior of drop tail . . . . . . . . . . . . . . . 4
4.2 ECN marking . . . . . . . . . . . . . . . . . . . . . . . . 4
4.3 PCN marking . . . . . . . . . . . . . . . . . . . . . . . . 5
4.4 Comments and Discussions . . . . . . . . . . . . . . . . . . 6
5. Receiver Behavior . . . . . . . . . . . . . . . . . . . . . . . 6
5.1 Monitoring per-packet statistics . . . . . . . . . . . . . . 6
5.2 Calculating time-smoothed values . . . . . . . . . . . . . . 6
5.3 Sending periodic feedback . . . . . . . . . . . . . . . . . 7
6. Sender Behavior . . . . . . . . . . . . . . . . . . . . . . . . 7
6.1 Encoder rate control . . . . . . . . . . . . . . . . . . . . 8
6.2 Rate shaping buffer . . . . . . . . . . . . . . . . . . . . 8
6.3 Target rate calculator . . . . . . . . . . . . . . . . . . . 8
6.3.1 Slow-start behavior . . . . . . . . . . . . . . . . . . 9
6.4 Sending rate calculator . . . . . . . . . . . . . . . . . . 9
7. Incremental Deployment . . . . . . . . . . . . . . . . . . . . 10
8. IANA Considerations . . . . . . . . . . . . . . . . . . . . . . 10
9. References . . . . . . . . . . . . . . . . . . . . . . . . . . 10
9.1 Normative References . . . . . . . . . . . . . . . . . . . 10
9.2 Informative References . . . . . . . . . . . . . . . . . . 11
Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . . 11
Zhu & Pan Expires September 13, 2013 [Page 2]
INTERNET DRAFT NADA March 12, 2013
1. Introduction
Interactive real-time media applications introduce a unique set of
challenges for congestion control. Unlike TCP, the mechanism used for
real-time media needs to adapt fast to instantaneous bandwidth
changes, accommodate fluctuations in the output of video encoder rate
control, and cause low queuing delay over the network. An ideal
scheme should also make effective use of all types of congestion
signals, including packet losses, queuing delay, and explicit
congestion notification (ECN) markings.
Based on the above considerations, we present a scheme named network-
assisted dynamic adaptation (NADA). The proposed design benefits from
explicit congestion control signals (e.g., ECN markings) from the
network, and remains compatible in the presence of implicit signals
(delay or loss) only. In addition, it supports weighted bandwidth
sharing among competing video flows.
This documentation describes the overall system architecture,
recommended designs at the sender and receiver, as well as expected
network nodes operations. The signaling mechanism consists of
standard RTP timestamp [RFC3550] and standard RTCP feedback reports.
2. Terminology
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 RFC 2119 [RFC2119].
3. System Model
The system consists of the following elements:
* Incoming media stream, in the form of consecutive raw video
frames and audio samples;
* Media encoder with rate control capabilities. It takes the
incoming media stream and encodes it to an RTP stream at a
target bit rate R_o. Note that the actual output rate from the
encoder R_v may fluctuate randomly around R_o. Also, the encoder
can only change its rate at rather coarse time intervals, on the
order of seconds.
* RTP sender, responsible for calculating the target bit rate
R_o based on network congestion signals (delay or ECN marking
reports from the receiver), and for regulating the actual
sending rate R_s accordingly. A rate shaping buffer is employed
Zhu & Pan Expires September 13, 2013 [Page 3]
INTERNET DRAFT NADA March 12, 2013
to absorb the instantaneous difference between video encoder
output rate R_v and sending rate R_s. The buffer size L_s,
together with R_o, influences the calculation of R_s. The RTP
sender also generates RTP timestamp in outgoing packets.
* RTP receiver, responsible for measuring and estimating end-to-
end delay d based on sender RTP timestamp. In the presence of
packet losses and ECN markings, it also records the individual
loss and marking events, and calculates the equivalent delay
d_tilde that accounts for queuing delay, ECN marking, and packet
losses. The receiver feeds such statistics back to the sender
via periodic RTCP reports.
* Network node, with several modes of operation. The system can
work with the default behavior of a simple drop tail queue. It
can also benefit from advanced AQM features such as RED-based
ECN marking, and PCN marking using a token bucket algorithm.
In the following, we will elaborate on the respective operations at the
network node, the receiver, and the sender.
4. Network Node Operations
We consider three variations of queue management behavior at the network
node, leading to either implicit or explicit congestion signals.
4.1 Default behavior of drop tail
In conventional network with drop tail or RED queues, congestion is
inferred from the estimation of end-to-end delay. No special action is
required at network node.
Packet drops at the queue are detected at the receiver, and contributes
to the calculation of the equivalent delay d_tilde.
4.2 ECN marking
In this mode, the network node randomly marks the ECN field in the IP
packet header following the Random Early Detection (RED) algorithm
[RFC2309]. Calculation of the marking probability involves the following
steps:
* upon packet arrival, update smoothed queue size q_avg as:
q_avg = alpha*q + (1-alpha)*q_avg.
The smoothing parameter alpha is a value between 0 and 1. A value of
alpha=1 corresponds to performing no smoothing at all.
Zhu & Pan Expires September 13, 2013 [Page 4]
INTERNET DRAFT NADA March 12, 2013
* calculate marking probability p as:
p = 0, if q < q_lo;
q_avg - q_lo
p = p_max*--------------, if q_lo <= q < q_hi;
q_hi - q_lo
p = 1, if q >= q_hi.
Here, q_lo and q_hi corresponds to the low and high thresholds of queue
occupancy. The maximum parking probability is p_max.
The ECN markings events will contribute to the calculation of an
equivalent delay d_tilde at the receiver. No changes are required at the
sender.
4.3 PCN marking
As a more advanced feature, we also envision network nodes which support
PCN marking based on virtual queues. In such a case, the marking
probability of the ECN bit in the IP packet header is calculated as
follows:
* upon packet arrival, meter packet against token bucket (r,b);
* update token level b_tk;
* calculate the marking probability as:
p = 0, if b-b_tk < b_lo;
b-b_tk-b_lo
p = p_max* --------------, if b_lo<= b-b_tk <b_hi;
b_hi-b_lo
p = 1, if b-b_tk>=b_hi.
Here, the token bucket lower and upper limits are denoted by b_lo and
b_hi, respectively. The parameter b indicates the size of the token
bucket. The parameter r is chosen as r=gamma*C, where gamma<1 is the
target utilization ratio and C designates link capacity. The maximum
marking probability is p_max.
The ECN markings events will contribute to the calculation of an
equivalent delay d_tilde at the receiver. No changes are required at the
sender. The virtual queuing mechanism from the PCN marking algorithm
will lead to additional benefits such as zero standing queues.
Zhu & Pan Expires September 13, 2013 [Page 5]
INTERNET DRAFT NADA March 12, 2013
4.4 Comments and Discussions
In all three flavors described above, the network queue operates with
the simple first-in-first-out (FIFO) principle. There is no need to
maintain per-flow state. Such a simple design ensures that the system
can scale easily with large number of video flows and high link
capacity.
The sender behavior stays the same in the presence of all types of
congestion signals: delay, loss, ECN marking due to either RED/ECN or
PCN algorithms. This unified approach allows a graceful transition of
the scheme as the level of congestion in the network shifts dynamically
between different regimes.
5. Receiver Behavior
The role of the receiver is fairly straightforward. It is in charge of
four steps: a) monitoring end-to-end delay/loss/marking statistics on a
per-packet basis; b) aggregating all forms of congestion signals in
terms of the equivalent delay; c) calculating time-smoothed value of the
congestion signal; and d) sending periodic reports back to the sender.
5.1 Monitoring per-packet statistics
The receiver observes and estimates one-way delay d_n for the n-th
packet, ECN marking event 1_M, and packet loss event 1_L. Here, 1_M and
1_L are binary indicators: the value of 1 corresponding to a marked or
lost packet and value of 0 indicates no marking or loss.
The equivalent delay d_tilde is calculated as follows:
d_tilde = d_n + 1_M d_M + 1_M d_L,
where d_M is a prescribed fictitious delay value corresponding to the
ECN marking event (e.g., d_M = 200 ms), and d_L is a prescribed
fictitious delay value corresponding to the packet loss event (e.g., d_L
= 1 second). By introducing a large fictitious delay penalty for ECN
marking and packet losses, our proposed scheme leads to low end-to-end
actual delays in the presence of such events.
5.2 Calculating time-smoothed values
The receiver smoothes its observations via exponential averaging:
x_n = alpha*d_tilde + (1-alpha)*x_n.
The weighting parameter alpha adjusts the level of smoothing.
Zhu & Pan Expires September 13, 2013 [Page 6]
INTERNET DRAFT NADA March 12, 2013
5.3 Sending periodic feedback
Periodically, the receiver sends back the updated value of x in RTCP
messages, to aid the sender in its calculation of target rate. The size
of acknowledgement packets are typically on the order of tens of bytes,
and are significantly smaller than average video packet sizes.
Therefore, the bandwidth overhead of the receiver acknowledgement stream
is sufficiently low.
6. Sender Behavior
As illustrated in Fig. 1, the sender comprises four modules: a) encoder
rate control; b) rate shaping buffer; c) encoder target rate calculator,
and d) sending rate calculator.
The following sections describe these modules in further details, and
explain how they interact with each other.
------------
| | R_v --------------
| |----------------> | | | | | -------------->
| Encoder | | | | | | / \
| | -------------- |
------------ Rate Shaping Buffer |
/ \ | |
| | |
| | | R_s
| R_o | |
| L_s | |
| | |
----------------- | -----------------
| | | | |
| Target Rate | R_o |----->| Sending Rate |
| Calculator |------------------------> | Calculator |
| | -----------------
----------------
/ \
|
|
--------------- RTCP report (loss/delay/marking)
Figure 1 Sender Structure
Zhu & Pan Expires September 13, 2013 [Page 7]
INTERNET DRAFT NADA March 12, 2013
6.1 Encoder rate control
The encoder rate control procedure has the following characteristics:
* Rate changes can happen only at large intervals, on the order of
seconds.
* Given a target rate R_o, the encoder output rate may randomly
fluctuate around it.
* The encoder output rate is further constrained by video content
complexity. The range of the final rate output is [R_min, R_max].
Note that it's content-dependent, and may change over time.
6.2 Rate shaping buffer
A rate shaping buffer is employed to absorb any instantaneous mismatch
between encoder rate output R_v and regulated sending rate R_s. The size
of the buffer evolves from time t-tau to time t as:
L_s(t) = max [0, L_s(t-tau)+R_v*tau-R_s*tau].
A large rate shaping buffer contributes to higher end-to-end delay,
which may harm the performance of real-time media communications.
Therefore, the sender has a strong incentive to constrain the size of
the shaping buffer. It can either deplete it faster by increasing the
sending rate R_s, or limit its growth by reducing the target rate for
the video encoder rate control. On the other hand, the sender still
needs to regulate its outgoing rate R_s so as not to incur excessive
congestion over the network. We will describe in Section 6.2 how to
balance between these two conflicting goals.
6.3 Target rate calculator
The sender calculates the target rate R_o based on network congestion
information from receiver RTCP reports. It first compensates the effect
of delayed observation by one round-trip time (RTT) via a linear
predictor:
x_n - x_n-1
x_hat = x_n + ---------------*tau_o (1)
delta
Zhu & Pan Expires September 13, 2013 [Page 8]
INTERNET DRAFT NADA March 12, 2013
In (1), the arrival interval between the (n-1)-th the n-th packets is
designated by delta. The parameter tau_o indicates the reference round-
trip-time, hence the prediction step size.
The target rate is then calculated as:
R_max-R_min
R_o = R_min + w*---------------*x_ref (2)
x_hat
Here, R_min and R_max denote the content-dependent rate range the
encoder can produce. The weight of priority level is w. The reference
congestion signal x_ref is chosen so that the maximum rate of R_max can
be achieved when x_hat = w*x_ref. Note that the combination of w and
x_ref determines how sensitive the rate adaptation scheme is in reaction
to fluctuations in observed signal x. The final target rate R_o is
clipped within the range of [R_min, R_max].
Note that the sender does not need to differentiate whether the network
node operates with drop tail, RED-based ECN marking, or token-bucket-
level-based PCN marking. It reacts to the aggregated congestion signal
information from the receiver in exactly the same manner.
6.3.1 Slow-start behavior
In addition, the initial sending rate of a stream is regulated to grow
linearly, no more than R_ss at time t:
t-t_0
R_ss(t) = R_min + -------(R_max-R_min).
T
The start time of the stream is t_0, and T represents the time horizon
over which the slow-start mechanism is effective. The encoder target
rate is chosen to be the minimum of R_o and R_ss in the first T
seconds.
6.4 Sending rate calculator
Finally, the actual outgoing rate over the network is R_s. Its value is
calculated based on both the encoder target rate R_o and the rate
shaping buffer size L_s in the following manner:
Zhu & Pan Expires September 13, 2013 [Page 9]
INTERNET DRAFT NADA March 12, 2013
L_s
R_s = R_o + beta * -------.
tau_v
The first term indicates the rate calculated from network congestion
feedback alone. The second term indicates the additional pressure to
send out more packets by the rate shaping buffer, if it is building up.
The scaling factor beta can be tuned to balance between these two
competing goals. The parameter tau_v is a reference latency value, to
keep beta dimensionless.
7. Incremental Deployment
One nice property of proposed design is the consistent behavior of video
end points regardless of variations in network node operations. This
facilitates gradual, incremental adoption of the scheme.
To start off with, the proposed encoder congestion control mechanism can
be implemented without any explicit support from the network, and react
to observed congestion signals both in terms of delay and loss.
When ECN is enabled at the network nodes with RED-based marking, the
receiver can fold its observations of ECN markings into the calculation
of the equivalent delay. The sender can react to these explicit
congestion signals without any modification.
Ultimately, networks equipped with proactive marking based on token
bucket level metering can reap the additional benefits of zero standing
queues and lower end-to-end delay and work seamlessly with existing
senders and receivers.
8. IANA Considerations
There are no actions for IANA.
9. References
9.1 Normative References
[RFC2119] Bradner, S., "Key words for use in RFCs to Indicate
Requirement Levels", BCP 14, RFC 2119, March 1997.
[RFC3550] Schulzrinne, H., Casner, S., Frederick, R., and V.
Jacobson, "RTP: A Transport Protocol for Real-Time
Applications", STD 64, RFC 3550, July 2003.
Zhu & Pan Expires September 13, 2013 [Page 10]
INTERNET DRAFT NADA March 12, 2013
9.2 Informative References
[RFC3168] Ramakrishnan, K., Floyd, S., and D. Black, "The Addition
of Explicit Congestion Notification (ECN) to IP",
RFC 3168, September 2001.
[RFC2309] Braden, B., Clark, D., Crowcroft, J., Davie, B., Deering,
S., Estrin, D., Floyd, S., Jacobson, V., Minshall, G.,
Partridge, C., Peterson, L., Ramakrishnan, K., Shenker,
S., Wroclawski, J., and L. Zhang, "Recommendations on
Queue Management and Congestion Avoidance in the
Internet", RFC 2309, April 1998.
Authors' Addresses
Xiaoqing Zhu
Cisco Systems,
510 McCarthy Blvd,
Milpitas, CA 95134, USA
EMail: xiaoqzhu@cisco.com
Rong Pan
Cisco Systems
510 McCarthy Blvd,
Milpitas, CA 95134, USA
Email: ropan@cisco.com
Zhu & Pan Expires September 13, 2013 [Page 11]