draft-ietf-rmcat-video-traffic-model-03.txt   draft-ietf-rmcat-video-traffic-model-04.txt 
Network Working Group X. Zhu Network Working Group X. Zhu
Internet-Draft S. Mena Internet-Draft S. Mena
Intended status: Informational Cisco Systems Intended status: Informational Cisco Systems
Expires: January 18, 2018 Z. Sarker Expires: July 22, 2018 Z. Sarker
Ericsson AB Ericsson AB
July 17, 2017 January 18, 2018
Modeling Video Traffic Sources for RMCAT Evaluations Modeling Video Traffic Sources for RMCAT Evaluations
draft-ietf-rmcat-video-traffic-model-03 draft-ietf-rmcat-video-traffic-model-04
Abstract Abstract
This document describes two reference video traffic source models for This document describes two reference video traffic source models for
evaluating RMCAT candidate algorithms. The first model statistically evaluating RMCAT candidate algorithms. The first model statistically
characterizes the behavior of a live video encoder in response to characterizes the behavior of a live video encoder in response to
changing requests on target video rate. The second model is trace- changing requests on target video rate. The second model is trace-
driven, and emulates the encoder output by scaling the pre-encoded driven, and emulates the encoder output based on actual encoded video
video frame sizes from a widely used video test sequence. Both frame sizes from a high-resolution test sequence. Both models are
models are designed to strike a balance between simplicity, designed to strike a balance between simplicity, repeatability, and
repeatability, and authenticity in modeling the interactions between authenticity in modeling the interactions between a live video
a video traffic source and the congestion control module. traffic source and the congestion control module. Finally, the
document describes how both approaches can be combined into a hybrid
model.
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 https://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 January 18, 2018. This Internet-Draft will expire on July 22, 2018.
Copyright Notice Copyright Notice
Copyright (c) 2017 IETF Trust and the persons identified as the Copyright (c) 2018 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 (https://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
carefully, as they describe your rights and restrictions with respect carefully, as they describe your rights and restrictions with respect
to this document. Code Components extracted from this document must to this document. Code Components extracted from this document must
include Simplified BSD License text as described in Section 4.e of include Simplified BSD License text as described in Section 4.e of
the Trust Legal Provisions and are provided without warranty as the Trust Legal Provisions and are provided without warranty as
described in the Simplified BSD License. described in the Simplified BSD License.
Table of Contents Table of Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 2 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 2
skipping to change at page 2, line 25 skipping to change at page 2, line 29
3. Desired Behavior of A Synthetic Video Traffic Model . . . . . 3 3. Desired Behavior of A Synthetic Video Traffic Model . . . . . 3
4. Interactions Between Synthetic Video Traffic Source and 4. Interactions Between Synthetic Video Traffic Source and
Other Components at the Sender . . . . . . . . . . . . . . . 4 Other Components at the Sender . . . . . . . . . . . . . . . 4
5. A Statistical Reference Model . . . . . . . . . . . . . . . . 6 5. A Statistical Reference Model . . . . . . . . . . . . . . . . 6
5.1. Time-damped response to target rate update . . . . . . . 7 5.1. Time-damped response to target rate update . . . . . . . 7
5.2. Temporary burst and oscillation during transient . . . . 8 5.2. Temporary burst and oscillation during transient . . . . 8
5.3. Output rate fluctuation at steady state . . . . . . . . . 8 5.3. Output rate fluctuation at steady state . . . . . . . . . 8
5.4. Rate range limit imposed by video content . . . . . . . . 9 5.4. Rate range limit imposed by video content . . . . . . . . 9
6. A Trace-Driven Model . . . . . . . . . . . . . . . . . . . . 9 6. A Trace-Driven Model . . . . . . . . . . . . . . . . . . . . 9
6.1. Choosing the video sequence and generating the traces . . 10 6.1. Choosing the video sequence and generating the traces . . 10
6.2. Using the traces in the syntethic codec . . . . . . . . . 11 6.2. Using the traces in the synthetic codec . . . . . . . . . 11
6.2.1. Main algorithm . . . . . . . . . . . . . . . . . . . 11 6.2.1. Main algorithm . . . . . . . . . . . . . . . . . . . 11
6.2.2. Notes to the main algorithm . . . . . . . . . . . . . 13 6.2.2. Notes to the main algorithm . . . . . . . . . . . . . 13
6.3. Varying frame rate and resolution . . . . . . . . . . . . 13 6.3. Varying frame rate and resolution . . . . . . . . . . . . 13
7. Combining The Two Models . . . . . . . . . . . . . . . . . . 14 7. Combining The Two Models . . . . . . . . . . . . . . . . . . 14
8. Implementation Status . . . . . . . . . . . . . . . . . . . . 15 8. Implementation Status . . . . . . . . . . . . . . . . . . . . 15
9. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 15 9. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 16
10. References . . . . . . . . . . . . . . . . . . . . . . . . . 16 10. References . . . . . . . . . . . . . . . . . . . . . . . . . 16
10.1. Normative References . . . . . . . . . . . . . . . . . . 16 10.1. Normative References . . . . . . . . . . . . . . . . . . 16
10.2. Informative References . . . . . . . . . . . . . . . . . 16 10.2. Informative References . . . . . . . . . . . . . . . . . 16
Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 17 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 17
1. Introduction 1. Introduction
When evaluating candidate congestion control algorithms designed for When evaluating candidate congestion control algorithms designed for
real-time interactive media, it is important to account for the real-time interactive media, it is important to account for the
characteristics of traffic patterns generated from a live video characteristics of traffic patterns generated from a live video
skipping to change at page 3, line 15 skipping to change at page 3, line 17
On the other hand, evaluation results of a candidate RMCAT algorithm On the other hand, evaluation results of a candidate RMCAT algorithm
should mostly reflect performance of the congestion control module, should mostly reflect performance of the congestion control module,
and somewhat decouple from peculiarities of any specific video codec. and somewhat decouple from peculiarities of any specific video codec.
It is also desirable that evaluation tests are repeatable, and be It is also desirable that evaluation tests are repeatable, and be
easily duplicated across different candidate algorithms. easily duplicated across different candidate algorithms.
One way to strike a balance between the above considerations is to One way to strike a balance between the above considerations is to
evaluate RMCAT algorithms using a synthetic video traffic source evaluate RMCAT algorithms using a synthetic video traffic source
model that captures key characteristics of the behavior of a live model that captures key characteristics of the behavior of a live
video encoder. To this end, this draft presents two reference video encoder. To this end, this draft presents two reference
models. The first is based on statistical modelling; the second is models. The first is based on statistical modeling; the second is
trace-driven. The draft also discusses the pros and cons of each trace-driven. The draft also discusses the pros and cons of each
approach, as well as how both approaches can be combined. approach, as well as how both approaches can be combined into a
hybrid model.
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 RFC2119 [RFC2119]. document are to be interpreted as described RFC2119 [RFC2119].
3. Desired Behavior of A Synthetic Video Traffic Model 3. Desired Behavior of A Synthetic Video Traffic Model
A live video encoder employs encoder rate control to meet a target A live video encoder employs encoder rate control to meet a target
skipping to change at page 4, line 31 skipping to change at page 4, line 37
o Statistical resemblance: The synthetic traffic should match the o Statistical resemblance: The synthetic traffic should match the
outcome of the real video encoder in terms of statistical outcome of the real video encoder in terms of statistical
characteristics, such as the mean, variance, peak, and characteristics, such as the mean, variance, peak, and
autocorrelation coefficients of the bitrate. It is also important autocorrelation coefficients of the bitrate. It is also important
that the statistical resemblance should hold across different time that the statistical resemblance should hold across different time
scales, ranging from tens of milliseconds to sub-seconds. scales, ranging from tens of milliseconds to sub-seconds.
o Wide range of coverage: The model should be easily configurable to o Wide range of coverage: The model should be easily configurable to
cover a wide range of codec behaviors (e.g., with either fast or cover a wide range of codec behaviors (e.g., with either fast or
slow reaction time in live encoder rate control) and video content slow reaction time in live encoder rate control) and video content
variations (e.g, ranging from high-motion to low-motion). variations (e.g., ranging from high-motion to low-motion).
These distinct behavior features can be characterized via simple These distinct behavior features can be characterized via simple
statistical models, or a trace-driven approach. We present an statistical modelling, or a trace-driven approach. Section 5 and
example of each in Section 5 and Section 6 Section 6 provide an example of each approach, respectively.
Section 7 discusses how both models can be combined together.
4. Interactions Between Synthetic Video Traffic Source and Other 4. Interactions Between Synthetic Video Traffic Source and Other
Components at the Sender Components at the Sender
Figure 1 depitcs the interactions of the synthetic video encoder with Figure 1 depicts the interactions of the synthetic video encoder with
other components at the sender, such as the application, the other components at the sender, such as the application, the
congestion control module, the media packet transport module, etc. congestion control module, the media packet transport module, etc.
Both reference models, as described later in Section 5 and Section 6, Both reference models, as described later in Section 5 and Section 6,
follow the same set of interactions. follow the same set of interactions.
The synthetic video encoder takes in raw video frames captured by the The synthetic video encoder takes in raw video frames captured by the
camera and then dynamically generates a sequence of encoded video camera and then dynamically generates a sequence of encoded video
frames with varying size and interval. These encoded frames are frames with varying size and interval. These encoded frames are
processed by other modules in order to transmit the video stream over processed by other modules in order to transmit the video stream over
the network. During the lifetime of a video transmission session, the network. During the lifetime of a video transmission session,
skipping to change at page 6, line 28 skipping to change at page 6, line 33
| | | |
-------------------+ +--------------------> -------------------+ +-------------------->
interface from interface to interface from interface to
other modules (a) other modules (b) other modules (a) other modules (b)
Figure 1: Interaction between synthetic video encoder and other Figure 1: Interaction between synthetic video encoder and other
modules at the sender modules at the sender
5. A Statistical Reference Model 5. A Statistical Reference Model
In this section, we describe one simple statistical model of the live This section describes one simple statistical model of the live video
video encoder traffic source. Figure 2 summarizes the list of encoder traffic source. Figure 2 summarizes the list of tunable
tunable parameters in this statistical model. A more comprehensive parameters in this statistical model. A more comprehensive survey of
survey of popular methods for modelling video traffic source behavior popular methods for modeling video traffic source behavior can be
can be found in [Tanwir2013]. found in [Tanwir2013].
+==============+====================================+================+ +==============+====================================+================+
| Notation | Parameter Name | Example Value | | Notation | Parameter Name | Example Value |
+==============+====================================+================+ +==============+====================================+================+
| R_v | Target rate request to encoder | 1 Mbps | | R_v | Target rate request to encoder | 1 Mbps |
+--------------+------------------------------------+----------------+ +--------------+------------------------------------+----------------+
| FPS | Target frame rate of encoder output| 30 Hz | | FPS | Target frame rate of encoder output| 30 Hz |
+--------------+------------------------------------+----------------+ +--------------+------------------------------------+----------------+
| tau_v | Encoder reaction latency | 0.2 s | | tau_v | Encoder reaction latency | 0.2 s |
+--------------+------------------------------------+----------------+ +--------------+------------------------------------+----------------+
skipping to change at page 7, line 48 skipping to change at page 7, line 48
* Example value of K_B for a video stream encoded at 720p and 30 frames * Example value of K_B for a video stream encoded at 720p and 30 frames
per second, using H.264/AVC encoder. per second, using H.264/AVC encoder.
Figure 2: List of tunable parameters in a statistical video traffic Figure 2: List of tunable parameters in a statistical video traffic
source model. source model.
5.1. Time-damped response to target rate update 5.1. Time-damped response to target rate update
While the congestion control module can update its target rate While the congestion control module can update its target rate
request R_v at any time, our model dictates that the encoder will request R_v at any time, the statistical model dictates that the
only react to such changes after tau_v seconds from a previous rate encoder will only react to such changes tau_v seconds after a
transition. In other words, when the encoder has reacted to a rate previous rate transition. In other words, when the encoder has
change request at time t, it will simply ignore all subsequent rate reacted to a rate change request at time t, it will simply ignore all
change requests until time t+tau_v. subsequent rate change requests until time t+tau_v.
5.2. Temporary burst and oscillation during transient 5.2. Temporary burst and oscillation during transient
The output rate R_o during the period [t, t+tau_v] is considered to The output rate R_o during the period [t, t+tau_v] is considered to
be in transient. Based on observations from video encoder output be in transient. Based on observations from video encoder output
data, we model the transient behavior of an encoder upon reacting to data, the transient behavior of an encoder upon reacting to a new
a new target rate request in the form of high variation in output target rate request is modelled in the form of high variation in
frame sizes. It is assumed that the overall average output rate R_o output frame sizes. It is assumed that the overall average output
during this period matches the target rate R_v. Consequently, the rate R_o during this period matches the target rate R_v.
occasional burst of large frames are followed by smaller-than average Consequently, the occasional burst of large frames are followed by
encoded frames. smaller-than-average encoded frames.
This temporary burst is characterized by two parameters: This temporary burst is characterized by two parameters:
o burst duration K_d: number of frames in the burst event; and o burst duration K_d: number of frames in the burst event; and
o burst frame size K_B: size of the initial burst frame which is o burst frame size K_B: size of the initial burst frame which is
typically significantly larger than average frame size at steady typically significantly larger than average frame size at steady
state. state.
It can be noted that these burst parameters can also be used to mimic It can be noted that these burst parameters can also be used to mimic
the insertion of a large on-demand I frame in the presence of severe the insertion of a large on-demand I frame in the presence of severe
packet losses. The values of K_d and K_B typically depend on the packet losses. The values of K_d and K_B typically depend on the
type of video codec, spatial and temporal resolution of the encoded type of video codec, spatial and temporal resolution of the encoded
stream, as well as the video content activity level. stream, as well as the video content activity level.
5.3. Output rate fluctuation at steady state 5.3. Output rate fluctuation at steady state
We model output rate R_o during steady state as randomly fluctuating The output rate R_o during steady state is modelled as randomly
around the target rate R_v. The output traffic can be characterized fluctuating around the target rate R_v. The output traffic can be
as the combination of two random processes denoting the frame characterized as the combination of two random processes denoting the
interval t and output frame size B over time. These two random frame interval t and output frame size B over time. These two random
processes capture two sources of variations in the encoder output: processes capture two sources of variations in the encoder output:
o Fluctuations in frame interval: the intervals between adjacent o Fluctuations in frame interval: the intervals between adjacent
frames have been observed to fluctuate around the reference frames have been observed to fluctuate around the reference
interval of t0 = 1/FPS. Deviations in normalized frame interval interval of t0 = 1/FPS. Deviations in normalized frame interval
DELTA_t = (t-t0)/t0 can be modelled by a zero-mean Laplacian DELTA_t = (t-t0)/t0 can be modelled by a zero-mean Laplacian
distribution with scaling parameter SCALE_t. The value of SCALE_t distribution with scaling parameter SCALE_t. The value of SCALE_t
dictates the "width" of the Laplacian distribution and therefore dictates the "width" of the Laplacian distribution and therefore
the amount of fluctuations in actual frame intervals (t) with the amount of fluctuations in actual frame intervals (t) with
respect to the reference t0. respect to the reference frame interval t0.
o Fluctuations in frame size: size of the output encoded frames also o Fluctuations in frame size: size of the output encoded frames also
tend to fluctuate around the reference frame size B0=R_v/8/FPS. tend to fluctuate around the reference frame size B0=R_v/8/FPS.
Likewise, deviations in the normalized frame size DELTA_B = Likewise, deviations in the normalized frame size DELTA_B =
(B-B0)/B0 can be modelled by a zero-mean Laplacian distribution (B-B0)/B0 can be modelled by a zero-mean Laplacian distribution
with scaling parameter SCALE_B. The value of SCALE_B dictates the with scaling parameter SCALE_B. The value of SCALE_B dictates the
"width" of this second Laplacian distribution and correspondingly "width" of this second Laplacian distribution and correspondingly
the amount of fluctuations in output frame sizes (B) with respect the amount of fluctuations in output frame sizes (B) with respect
to the reference target B0. to the reference target B0.
Both values of SCALE_t and SCALE_B can be obtained via parameter Both values of SCALE_t and SCALE_B can be obtained via parameter
fitting from empirical data captured for a given video encoder. fitting from empirical data captured for a given video encoder.
Example values are listed in Figure 2 based on empirical data Example values are listed in Figure 2 based on empirical data
presented in [IETF-Interim]. presented in [IETF-Interim].
5.4. Rate range limit imposed by video content 5.4. Rate range limit imposed by video content
The output rate R_o is further clipped within the dynamic range The output rate R_o is further clipped within the dynamic range
[R_min, R_max], which in reality are dictated by scene and motion [R_min, R_max], which in reality are dictated by scene and motion
complexity of the captured video content. In our model, these complexity of the captured video content. In the proposed
parameters are specified by the application. statistical model, these parameters are specified by the application.
6. A Trace-Driven Model 6. A Trace-Driven Model
We now present the second approach to model a video traffic source. The second approach for modelling a video traffic source is trace-
This approach is based on running an actual live video encoder on a driven. This can be achieved by running an actual live video encoder
set of chosen raw video sequences and using the encoder's output on a set of chosen raw video sequences and using the encoder's output
traces for constructing a synthetic live encoder. With this traces for constructing a synthetic live encoder. With this
approach, the recorded video traces naturally exhibit temporal approach, the recorded video traces naturally exhibit temporal
fluctuations around a given target rate request R_v from the fluctuations around a given target rate request R_v from the
congestion control module. congestion control module.
The following list summarizes the main steps of this approach: The following list summarizes the main steps of this approach:
1) Choose one or more representative raw video sequences. 1. Choose one or more representative raw video sequences.
2) Encode the sequence(s) using an actual live video encoder. Repeat 2. Encode the sequence(s) using an actual live video encoder.
the process for a number of bitrates. Keep only the sequence of Repeat the process for a number of bitrates. Keep only the
frame sizes for each bitrate. sequence of frame sizes for each bitrate.
3) Construct a data structure that contains the output of the 3. Construct a data structure that contains the output of the
previous step. The data structure should allow for easy bitrate previous step. The data structure should allow for easy bitrate
lookup. lookup.
4) Upon a target bitrate request R_v from the controller, look up the 4. Upon a target bitrate request R_v from the controller, look up
closest bitrates among those previously stored. Use the frame size the closest bitrates among those previously stored. Use the
sequences stored for those bitrates to approximate the frame sizes to frame size sequences stored for those bitrates to approximate the
output. frame sizes to output.
5) The output of the synthetic encoder contains "encoded" frames with 5. The output of the synthetic encoder contains "encoded" frames
zeros as contents but with realistic sizes. with zeros as contents but with realistic sizes.
Section 6.1 explains steps 1), 2), and 3), Section 6.2 elaborates on In the following, Section 6.1 explains the first three steps (1-3),
steps 4) and 5). Finally, Section 6.3 briefly discusses the Section 6.2 elaborates on the remaining two steps (4-5). Finally,
possibility to extend the model for supporting variable frame rate Section 6.3 briefly discusses the possibility to extend the trace-
and/or variable frame resolution. driven model for supporting time-varying frame rate and/or time-
varying frame resolution.
6.1. Choosing the video sequence and generating the traces 6.1. Choosing the video sequence and generating the traces
The first step we need to perform is a careful choice of a set of The first step is a careful choice of a set of video sequences that
video sequences that are representative of the use cases we want to are representative of the target use cases for the video traffic
model. Our use case here is video conferencing, so we must choose a model. For the example use case of interactive video conferencing,
low-motion sequence that resembles a "talking head", for instance a it is recommended to choose a low-motion sequence that resembles a
news broadcast or a video capture of an actual conference call. "talking head", e.g. from a news broadcast or recording of an actual
video conferencing call.
The length of the chosen video sequence is a tradeoff. If it is too The length of the chosen video sequence is a tradeoff. If it is too
long, it will be difficult to manage the data structures containing long, it will be difficult to manage the data structures containing
the traces. If it is too short, there will be an obvious periodic the traces. If it is too short, there will be an obvious periodic
pattern in the output frame sizes, leading to biased results when pattern in the output frame sizes, leading to biased results when
evaluating congestion controller performance. In our experience, a evaluating congestion control performance. In our experience, a
sequence whose length is between 2 and 4 minutes is a fair tradeoff. sequence with a length between 2 and 4 minutes is a fair tradeoff.
Once we have chosen the raw video sequence, denoted S, we use a live Given the chosen raw video sequence, denoted S, one can use a live
encoder, e.g. [H264] or [HEVC] to produce a set of encoded encoder, e.g. some implementation of [H264] or [HEVC], to produce a
sequences. As discussed in Section 3, a live encoder's output set of encoded sequences. As discussed in Section 3, the output
bitrate can be tuned by varying three input parameters, namely, bitrate of the live encoder can be achieved by tuning three input
quantization step size, frame rate, and picture resolution. In order parameters: quantization step size, frame rate, and picture
to simplify the choice of these parameters for a given target rate, resolution. In order to simplify the choice of these parameters for
we assume a fixed frame rate (e.g. 30 fps) and a fixed resolution a given target rate, one can typically assume a fixed frame rate
(e.g., 720p). See section 6.3 for a discussion on how to relax these (e.g. 30 fps) and a fixed resolution (e.g., 720p) when configuring
assumptions. the live encoder. See Section 6.3 for a discussion on how to relax
these assumptions.
Following these simplifications, we run the chosen encoder by setting Following these simplifications, the chosen encoder can be configured
a constant target bitrate at the beginning, then letting the encoder to start at a constant target bitrate, then vary the quantization
vary the quantization step size internally while encoding the input step size (internally via the video encoder rate controller) to meet
video sequence. Besides, we assume that the first frame is encoded various externally specified target rates. It can be further assumed
as an I-frame and the rest are P-frames. We further assume that the the first frame is encoded as an I-frame and the rest are P-frames.
encoder algorithm does not use knowledge of frames in the future when For live encoding, the encoder rate control algorithm typically does
encoding a given frame. not use knowledge of frames in the future when encoding a given
frame.
Given R_min and R_max, which are the minimum and maximum bitrates at Given the minimum and maximum bitrates at which the synthetic codec
which the synthetic codec is to operate (see Section 4), we divide is to operate (denoted as R_min and R_max, see Section 4), the entire
the bitrate range between R_min and R_max in n_s + 1 bitrate steps of range of target bitrates can be divided into n_s + 1 bitrate steps of
length l = (R_max - R_min) / n_s. We then use the following simple length l = (R_max - R_min) / n_s. The following simple algorithm is
algorithm to encode the raw video sequence. used to encode the raw video sequence.
r = R_min r = R_min
while r <= R_max do while r <= R_max do
Traces[r] = encode_sequence(S, r, e) Traces[r] = encode_sequence(S, r, e)
r = r + l r = r + l
where function encode_sequence takes as parameters, respectively, a The function encode_sequence takes as input parameters, respectively,
raw video sequence, a constant target rate, and an encoder algorithm; a raw video sequence (S), a constant target rate (r), and an encoder
it returns a vector with the sizes of frames in the order they were rate control algorithm (e); it returns a vector with the sizes of
encoded. The output vector is stored in a map structure called frames in the order they were encoded. The output vector is stored
Traces, whose keys are bitrates and whose values are vectors of frame in a map structure called Traces, whose keys are bitrates and whose
sizes. values are vectors of frame sizes.
The choice of a value for n_s is important, as it determines the The choice of a value for n_s is important, as it determines the
number of vectors of frame sizes stored in map Traces. The minimum number of vectors of frame sizes stored in the map Traces. The
value one can choose for n_s is 1, and its maximum value depends on minimum value one can choose for n_s is 1, and the maximum value
the amount of memory available for holding the map Traces. A depends on the amount of memory available for holding the map Traces.
reasonable value for n_s is one that makes the steps' length l = 200 A reasonable value for n_s is one that results in steps of length l =
kbps. We will further discuss step length l in the next section. 200 kbps. The next section will discuss further the choice of the
step length l.
Finally, note that, as mentioned in previous sections, R_min and Finally, note that, as mentioned in previous sections, R_min and
R_max may be modified after the initial sequences are encoded. R_max may be modified after the initial sequences are encoded.
Hence, the algorithm described in the next section also covers the Hence, the algorithm described in the next section also covers the
cases when the current target bitrate is less than R_min, or greater cases when the current target bitrate is less than R_min, or greater
than R_max. than R_max.
6.2. Using the traces in the syntethic codec 6.2. Using the traces in the synthetic codec
The main idea behind the trace-driven synthetic codec is that it The main idea behind the trace-driven synthetic codec is that it
mimics a real live codec's rate adaptation when the congestion mimics the rate adaptation behavior of a real live codec upon dynamic
controller updates the target rate R_v dynamically. It does so by updates of the target rate R_v by the congestion control module. It
switching to a different frame size vector stored in the map Traces does so by switching to a different frame size vector stored in the
when needed. map Traces when needed.
6.2.1. Main algorithm 6.2.1. Main algorithm
We maintain two variables r_current and t_current: The main algorithm for rate adaptation in the synthetic codec
maintains two variables: r_current and t_current.
* r_current points to one of the keys of map Traces. Upon a change o The variable r_current points to one of the keys of map Traces.
in the value of R_v, typically because the congestion controller Upon a change in the value of R_v, typically because the
detects that the network conditions have changed, r_current is congestion controller detects that the network conditions have
updated to the greatest key in Traces that is less than or equal to changed, r_current is updated to the greatest key in Traces that
the new value of R_v. For the moment, we assume the value of R_v to is less than or equal to the new value of R_v. It is assumed that
be clipped in the range [R_min, R_max]. the value of R_v is clipped within the range [R_min, R_max].
r_current = r r_current = r
such that such that
( r in keys(Traces) and ( r in keys(Traces) and
r <= R_v and r <= R_v and
(not(exists) r' in keys(Traces) such that r < r' <= R_v) ) (not(exists) r' in keys(Traces) such that r < r' <= R_v) )
* t_current is an index to the frame size vector stored in o The variable t_current is an index to the frame size vector stored
Traces[r_current]. It is updated every time a new frame is due. We in Traces[r_current]. It is updated every time a new frame is
assume all vectors stored in Traces to have the same size, denoted due. It is assumed that all vectors stored Traces to have the
size_traces. The following equation governs the update of t_current: same size, denoted as size_traces. The following equation governs
the update of t_current:
if t_current < SkipFrames then if t_current < SkipFrames then
t_current = t_current + 1 t_current = t_current + 1
else else
t_current = ((t_current+1-SkipFrames) % (size_traces- SkipFrames)) t_current = ((t_current+1-SkipFrames) % (size_traces-SkipFrames))
+ SkipFrames + SkipFrames
where operator % denotes modulo, and SkipFrames is a predefined where operator % denotes modulo, and SkipFrames is a predefined
constant that denotes the number of frames to be skipped at the constant that denotes the number of frames to be skipped at the
beginning of frame size vectors after t_current has wrapped around. beginning of frame size vectors after t_current has wrapped around.
The point of constant SkipFrames is avoiding the effect of The point of constant SkipFrames is avoiding the effect of
periodically sending a (big) I-frame followed by several smaller- periodically sending a large I-frame followed by several smaller-
than-normal P-frames. We typically set SkipFrames to 20, although it than-average P-frames. A typical value of SkipFrames is 20, although
could be set to 0 if we are interested in studying the effect of it could be set to 0 if one is interested in studying the effect of
sending I-frames periodically. sending I-frames periodically.
We initialize r_current to R_min, and t_current to 0. The initial value of r_current is set to R_min, and the initial value
of t_current set to 0.
When a new frame is due, we need to calculate its size. There are When a new frame is due, its size can be calculated following one of
three cases: the three cases below:
a) R_min <= R_v < Rmax: In this case we use linear interpolation of a) R_min <= R_v < Rmax: the output frame size is calculated via
the frame sizes appearing in Traces[r_current] and linear interpolation of the frame sizes appearing in
Traces[r_current + l]. The interpolation is done as follows: Traces[r_current] and Traces[r_current + l]. The interpolation is
done as follows:
size_lo = Traces[r_current][t_current] size_lo = Traces[r_current][t_current]
size_hi = Traces[r_current + l][t_current] size_hi = Traces[r_current + l][t_current]
distance_lo = ( R_v - r_current ) / l distance_lo = ( R_v - r_current ) / l
framesize = size_hi * distance_lo + size_lo * (1 - distance_lo) framesize = size_hi * distance_lo + size_lo * (1 - distance_lo)
b) R_v < R_min: In this case, we scale the trace sequence with the b) R_v < R_min: the output frame size is calculated via scaling with
lowest bitrate, in the following way: respect to the lowest bitrate R_min, as follows:
factor = R_v / R_min factor = R_v / R_min
framesize = max(1, factor * Traces[R_min][t_current]) framesize = max(1, factor * Traces[R_min][t_current])
c) R_v >= R_max: We also use scaling for this case. We use the c) R_v >= R_max: the output frame size is calculated by scaling with
trace sequence with the greatest bitrate: respect to the highest bitrate R_max:
factor = R_v / R_max factor = R_v / R_max
framesize = factor * Traces[R_max][t_current] framesize = factor * Traces[R_max][t_current]
In case b), we set the minimum to 1 byte, since the value of factor In case b), we set the minimum output size to 1 byte, since the value
can be arbitrarily close to 0. of factor can be arbitrarily close to 0.
6.2.2. Notes to the main algorithm 6.2.2. Notes to the main algorithm
* Reacting to changes in target bitrate. Similarly to the Note that main algorithm as described above can be further extended
statistical model presented in Section 5, the trace-driven synthetic to mimic some additional typical behaviors of a live encoder. Two
codec can have a time bound, tau_v, to reacting to target bitrate examples are given below:
changes. If the codec has reacted to an update in R_v at time t, it
will delay any further update to R_v to time t + tau_v. Note that,
in any case, the value of tau_v cannot be chosen shorter than the
time between frames, i.e. the inverse of the frame rate.
* I-frames on demand. The synthetic codec could be extended to o I-frames on demand: The synthetic codec can be extended to
simulate the sending of I-frames on demand, e.g., as a reaction to simulate the sending of I-frames on demand, e.g., as a reaction to
losses. To implement this extension, the codec's API is augmented losses. To implement this extension, the codec's incoming
with a new function to request a new I-frame. Upon calling such interface (see (a) in Figure 1) is augmented with a new function
function, t_current is reset to 0. to request a new I-frame. Upon calling such function, t_current
is reset to 0.
* Variable length l of steps defined between R_min and R_max. In the o Variable step length l between R_min and R_max: In the main
main algorithm's description, the step length l is fixed. However, algorithm, the step length l is fixed for ease of explanation.
if the range [R_min, R_max] is very wide, it is also possible to However, if the range [R_min, R_max] is very wide, it is also
define a set of steps with a non-constant length. The idea behind possible to define a set of intermediate encoding rates with
this modification is that the difference between 400 kbps and 600 variable step length. The rationale behind this modification is
kbps as bitrate is much more important than the difference between that the difference between 400 kbps and 600 kbps as target
4400 kbps and 4600 kbps. For example, one could define steps of bitrate is much more significant than the difference between 4400
length 200 Kbps under 1 Mbps, then steps of length 300 kbps between 1 kbps and 4600 kbps. For example, one could define steps of length
Mbps and 2 Mbps; 400 kbps between 2 Mbps and 3 Mbps, and so on. 200 Kbps under 1 Mbps, then steps of length 300 Kbps between 1
Mbps and 2 Mbps; 400 Kbps between 2 Mbps and 3 Mbps, and so on.
6.3. Varying frame rate and resolution 6.3. Varying frame rate and resolution
The trace-driven synthetic codec model explained in this section is The trace-driven synthetic codec model explained in this section is
relatively simple because we have fixed the frame rate and the frame relatively simple due to fixed frame rate and frame resolution. The
resolution. The model could be extended to have variable frame rate, model can extended further to accommodate variable frame rate and/or
variable spatial resolution, or both. variable spatial resolution.
When the encoded picture quality at a given bitrate is low, one can When the encoded picture quality at a given bitrate is low, one can
potentially decrease the frame rate (if the video sequence is potentially decrease either the frame rate (if the video sequence is
currently in low motion) or the spatial resolution in order to currently in low motion) or the spatial resolution in order to
improve quality-of-experince (QoE) in the overall encoded video. On improve quality-of-experince (QoE) in the overall encoded video. On
the other hand, if target bitrate increases to a point where there is the other hand, if target bitrate increases to a point where there is
no longer a perceptible improvement in the picture quality of no longer a perceptible improvement in the picture quality of
individual frames, then one might afford to increase the spatial individual frames, then one might afford to increase the spatial
resolution or the frame rate (useful if the video is currently in resolution or the frame rate (useful if the video is currently in
high motion). high motion).
Many techniques have been proposed to choose over time the best Many techniques have been proposed to choose over time the best
combination of encoder quatization step size, frame rate, and spatial combination of encoder quatization step size, frame rate, and spatial
skipping to change at page 14, line 12 skipping to change at page 14, line 22
[Ozer2011][Hu2010]. Future work may consider extending the trace- [Ozer2011][Hu2010]. Future work may consider extending the trace-
driven codec to accommodate variable frame rate and/or resolution. driven codec to accommodate variable frame rate and/or resolution.
From the perspective of congestion control, varying the spatial From the perspective of congestion control, varying the spatial
resolution typically requires a new intra-coded frame to be resolution typically requires a new intra-coded frame to be
generated, thereby incurring a temporary burst in the output traffic generated, thereby incurring a temporary burst in the output traffic
pattern. The impact of frame rate change tends to be more subtle: pattern. The impact of frame rate change tends to be more subtle:
reducing frame rate from high to low leads to sparsely spaced larger reducing frame rate from high to low leads to sparsely spaced larger
encoded packets instead of many densely spaced smaller packets. Such encoded packets instead of many densely spaced smaller packets. Such
difference in traffic profiles may still affect the performance of difference in traffic profiles may still affect the performance of
congestion control, especially when outgoing packets are not paced at congestion control, especially when outgoing packets are not paced by
the transport module. We leave the investigation of varying frame the media transport module. Investigation of varying frame rate and
rate to future work. resolution are left for future work.
7. Combining The Two Models 7. Combining The Two Models
It is worthwhile noting that the statistical and trace-driven models It is worthwhile noting that the statistical and trace-driven models
each has its own advantages and drawbacks. While both models are each has its own advantages and drawbacks. Both models are fairly
fairly simple to implement, it takes significantly greater effort to simple to implement. It takes significantly greater effort to fit
fit the parameters of a statistical model to actual encoder output the parameters of a statistical model to actual encoder output data
data whereas it is straightforward for a trace-driven model to obtain whereas it is straightforward for a trace-driven model to obtain
encoded frame size data. On the other hand, once validated, the encoded frame size data. On the other hand, once validated, the
statistical model is more flexible in mimicking a wide range of statistical model is more flexible in mimicking a wide range of
encoder/content behaviors by simply varying the correponding encoder/content behaviors by simply varying the correponding
parameters in the model. In this regard, a trace-driven model relies parameters in the model. In this regard, a trace-driven model relies
-- by definition -- on additional data collection efforts for -- by definition -- on additional data collection efforts for
accommodating new codecs or video contents. accommodating new codecs or video contents.
In general, the trace-driven model is more realistic for mimicking In general, the trace-driven model is more realistic for mimicking
ongoing, steady-state behavior of a video traffic source whereas the ongoing, steady-state behavior of a video traffic source whereas the
statistical model is more versatile for simulating transient events statistical model is more versatile for simulating transient events
skipping to change at page 15, line 25 skipping to change at page 15, line 25
+------>| frame from | +------>| frame from |
steady-state | trace | steady-state | trace |
+---------------+ +---------------+
Figure 3: Hybrid approach for modeling video traffic Figure 3: Hybrid approach for modeling video traffic
As shown in Figure 3, the video traffic model operates in transient As shown in Figure 3, the video traffic model operates in transient
state if the requested target rate R_v is substantially higher than state if the requested target rate R_v is substantially higher than
the previous target, or else it operates in steady state. During the previous target, or else it operates in steady state. During
transient state, a total of K_d frames are generated by the transient state, a total of K_d frames are generated by the
statistical model, resulting in 1 big burst frame with size K_B statistical model, resulting in one (1) big burst frame with size K_B
followed by K_d-1 smaller frames. When operating at steady-state, followed by K_d-1 smaller frames. When operating at steady-state,
the video traffic model simply generates a frame according to the the video traffic model simply generates a frame according to the
trace-driven model given the target rate, while modulating the frame trace-driven model given the target rate, while modulating the frame
interval according to the distribution specified by the statistical interval according to the distribution specified by the statistical
model. One example criterion for determining whether the traffic model. One example criterion for determining whether the traffic
model should operate in transient state is whether the rate increase model should operate in transient state is whether the rate increase
exceeds 10% of previous target rate. exceeds 10% of previous target rate. Finally, as this model follows
transient state behavior dictated by the statistical model, upon a
substantial rate change, the model will follow the time-damping
mechanism defined in Section 5.1, which is governed by parameter
tau_v.
8. Implementation Status 8. Implementation Status
The statistical model has been implemented as a traffic generator The statistical model has been implemented as a traffic generator
module within the [ns-2] network simulation platform. module within the [ns-2] network simulation platform.
More recently, both the statistical and trace-driven models have been More recently, the statistical, trace-driven, and hybrid models have
implemented as a stand-alone traffic source module. This can be been implemented as a stand-alone, platform-independent traffic
easily integrated into network simulation platforms such as [ns-2] source module. This can be easily integrated into network simulation
and [ns-3], as well as testbeds using a real network. The stand- platforms such as [ns-2] and [ns-3], as well as testbeds using a real
alone traffic source module is available as an open source network. The stand-alone traffic source module is available as an
implementation at [Syncodecs]. open source implementation at [Syncodecs].
9. IANA Considerations 9. IANA Considerations
There are no IANA impacts in this memo. There are no IANA impacts in this memo.
10. References 10. References
10.1. Normative References 10.1. Normative References
[H264] ITU-T Recommendation H.264, "Advanced video coding for
generic audiovisual services", 2003,
<http://www.itu.int/rec/T-REC-H.264-201304-I>.
[HEVC] ITU-T Recommendation H.265, "High efficiency video
coding", 2015.
[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, Requirement Levels", BCP 14, RFC 2119,
DOI 10.17487/RFC2119, March 1997, DOI 10.17487/RFC2119, March 1997,
<http://www.rfc-editor.org/info/rfc2119>. <https://www.rfc-editor.org/info/rfc2119>.
10.2. Informative References 10.2. Informative References
[H264] ITU-T Recommendation H.264, "Advanced video coding for
generic audiovisual services", May 2003,
<https://www.itu.int/rec/T-REC-H.264>.
[HEVC] ITU-T Recommendation H.265, "High efficiency video
coding", April 2013,
<https://www.itu.int/rec/T-REC-H.265>.
[Hu2010] Hu, H., Ma, Z., and Y. Wang, "Optimization of Spatial, [Hu2010] Hu, H., Ma, Z., and Y. Wang, "Optimization of Spatial,
Temporal and Amplitude Resolution for Rate-Constrained Temporal and Amplitude Resolution for Rate-Constrained
Video Coding and Scalable Video Adaptation", in Proc. 19th Video Coding and Scalable Video Adaptation", in Proc. 19th
IEEE International Conference on Image IEEE International Conference on Image
Processing, (ICIP'12), September 2012. Processing, (ICIP'12), September 2012.
[IETF-Interim] [IETF-Interim]
Zhu, X., Mena, S., and Z. Sarker, "Update on RMCAT Video Zhu, X., Mena, S., and Z. Sarker, "Update on RMCAT Video
Traffic Model: Trace Analysis and Model Update", April Traffic Model: Trace Analysis and Model Update", April
2017, <https://www.ietf.org/proceedings/interim-2017- 2017, <https://www.ietf.org/proceedings/
rmcat-01/slides/slides-interim-2017-rmcat-01-sessa-update- interim-2017-rmcat-01/slides/slides-interim-2017-rmcat-01-
on-video-traffic-model-draft-00.pdf>. sessa-update-on-video-traffic-model-draft-00.pdf>.
[ns-2] "The Network Simulator - ns-2", [ns-2] "The Network Simulator - ns-2",
<http://www.isi.edu/nsnam/ns/>. <http://www.isi.edu/nsnam/ns/>.
[ns-3] "The Network Simulator - ns-3", <https://www.nsnam.org/>. [ns-3] "The Network Simulator - ns-3", <https://www.nsnam.org/>.
[Ozer2011] [Ozer2011]
Ozer, J., "Video Compression for Flash, Apple Devices and Ozer, J., "Video Compression for Flash, Apple Devices and
HTML5", ISBN 13:978-0976259503, 2011. HTML5", ISBN 13:978-0976259503, 2011.
 End of changes. 68 change blocks. 
204 lines changed or deleted 220 lines changed or added

This html diff was produced by rfcdiff 1.46. The latest version is available from http://tools.ietf.org/tools/rfcdiff/