MPEG-21-based Cross-Layer Optimization Techniques for enabling Quality of Experience

MPEG-21-based Cross-Layer Optimization Techniques for enabling Quality of Experience

Christian Timmerer

Klagenfurt University (UNIKLU) Faculty of Technical Sciences (TEWI)

Department of Information Technology (ITEC) Multimedia Communication (MMC)

http://research.timmerer.com http://blog.timmerer.com mailto:[email protected]

Acknowledgments: DANAE, ENTHRONE, P2P-Next, ALICANTE projects funded by EC, SCALIPTV/SCALNET funded by FFG, ASSSV funded by FWF and, in particular Michael Eberhard, Ingo Kofler, Robert Kuschnig, Michael Ransburg, Michael Sablatschan, Hermann Hellwagner

Outline

• Background / Introduction

– Cross-layer designs & optimizations– MPEG-21 Digital Item Adaptation

• How to exploit MPEG-21 for XL optimizations?

– Cross-Layer Model (XLM)– Instantiation of the XLM by utilizing MPEG-21

metadata– Cross-Layer Adaptation Decision-Taking Engine (XL-

ADTE)

• Conclusions

2010/01/20 2Christian Timmerer, Klagenfurt University, Austria

Background / Introduction

• Cross-layer designs– Aim: increase QoS/QoEby performing coordinated actions

across the network layers => violating the protocol hierarchy and isolation model

– Approaches: bottom-up or a top-down or jointly optimizing parameters at the different layers

– Common property: compromising interoperability in favor of performance

• Increasing the interoperability of cross-layer designs by adopting an open standard – MPEG-21 Digital Item Adaptation – for describing the functional dependencies across network layers

2010/01/20 Christian Timmerer, Klagenfurt University, Austria 3

Digital Item AdaptationDIA := syntax and semantics of tools that assist in the adaptation of Digital Items

Goals:• Satisfy transmission, storage and

consumption constraints as well asQuality of Service (QoS) management

• Enable transparent access to (distributed)advanced multimedia content by shieldingusers from network and terminal installationissues

• Codec Format-independent mechanisms that provide support for Digital Item Adaptation in terms of:– Resource adaptation– Description adaptation– Quality of Service management

• The adaptation engines themselves are non-normative tools2010/01/20 Christian Timmerer, Klagenfurt University, Austria 4

fundamental inputto any adaptation engine

2010/01/20 Christian Timmerer, Klagenfurt University, Austria

Usage Environment Description (UED)

Terminal Capabilities

• Codec Capabilities• Device Properties• Input-Output Characteristics

User Characteristics

• User Info• Usage Preference & History• Presentation Preferences• Accessibility• Location

Network Characteristics

• Capabilities• Conditions

Natural Environment Characteristics

• Location & Time• Audio-Visual

5

Context-related metadata describes the usage environment in terms of terminal capabilities; network characteristics; user characteristics; natural environment characteristics;

e.g., codec capabilities = mp2, ML@MP; available bandwidth=1500kbps; visually impaired; high-level ambient noise;

AdaptationQoS and Universal Constraints Description

• Content-related metadata – AdaptationQoS– describes the relationship between constraints; feasible adaptation operations satisfying these constraints; associated utilities (qualities);

e.g., available bandwidth is 384kbps, terminal display is CIF; reduce bit-rate; quality at QCIF/30fps/QP=10 versus CIF/10fps/QP=15e.g., bit-rate = 256kbps, frame-rate=30fps, resolution=CIF, etc.

• Universal Constraints Description (UCD): mathematical approach based on an optimization problem– find values for the variables representing adaptation parameters that do

not violate the limitation constraints (feasibility) and maximize the optimization constraint(optimality, objective function)


How to exploit MPEG-21 for XL optimizations?

Three-step approach1. Cross-Layer Model (XLM): describing the relationship between QoS

metrics at different levels – No specific notation (e.g., graphical)– For example:

2. Instantiation of the XLM by utilizing MPEG-21 metadata– AdaptationQoS (AQoS): describe the relationship between constraints,

feasible adaptation operations satisfying these constraints, and associated utilities (qualities)

– Usage Environment Description (UED): context information (network conditions, terminal capabilities, user preferences, etc.)

– Universal Constraints Description (UCD): limitation and optimization constraints

3. Cross-Layer Adaptation Decision-Taking Engine (XL-ADTE)– Software module solving an optimization problem adopting any algorithm


Example: Adaptive XL-based Streaming


TID, DID, QID

max. payload size, forward error correction

supported display resolution, frame-rate

packet loss, jitter

signal strength, physical rate

temporal id (TID)dependency id (DID)quality id (QID)packet size

vertical and horizontal resolutionbit rateframe rate

Basic Cross-Layer Model

Advanced Cross-Layer Model

• Desirable characteristics– TCP friendliness: long-term throughput similar to TCP– Responsiveness: time to act upon a certain event– Smoothness: variation experienced for a particular flow

• TCP-friendly Rate Control Protocol (TFRC)

– Throughput T in bytes/sec is modeled as a function of• Segment size sin bytes• RTT estimate r in seconds• Loss event rate pas a fraction between 0.0 and 1.0• TCP retransmission timeout value tRTOin seconds (simple tRTO= 4r)

– Adapts sending rate accordingly• If Tcurr>Tnewthen reduce rate else increase rate


Instantiation of XLM using MPEG-21

• AdaptationQoS (AQoS)– Parameters (TID, …) as IOPins– Basic XL model as Look-Up Table (LUT)– Advanced XL model as Stack Function (SF)

• Usage Environment Description (UED)– Display resolution as display capabilities– Max bit-rate of codec as codec capabilities– RTT as packetTwoWay– Loss event rate by using the packetLossRate

• Universal Constraints Description (UCD)– Limit constraints

• resulting bit-rate < TFRC transmit rate• resulting bit-rate < max bit-rate of codec• video resolution < display size

– Optimization constraint: max bit-rate


Cross-Layer Adaptation Decision-Taking Engine (XL-ADTE)

Example: Adaptation of Scalable Video

• MPEG/ITU-T Scalable Video Coding (SVC)– 3 dimensions of scalability: spatial, temporal, signal-

to-noise ratio (SNR)

• Spatial dimension [pixels]: 640x360, 1024x576, 1920x1080

• Temporal dimension [fps]: 15, 30• Step 1: Determine Variables


Variables Domain Variables Domain

SpatLayer {0, 1, 2} HorizRes [176, 1920]TempLayer {0,1} VertRes [144, 1080]

FrameRate [1, 30]BitRate [0.5, 12]

Table 1. (a) Adaptation Parameter Variables; (b) Content Property Variables.

Example: Adaptation of Scalable Video (cont’d)

• Step 2: Identify Functional Dependencies

• Step 3: Restrict Solution Space (Limit Constraints)

• Step 4: Define Objective Functionmaximize FrameRate


Function MappingHorizRes = f(SpatLayers) {0640, 11024, 21920}VertRes = f(HorizRes) VertRes = HorizRes / 16 * 9BitRate = f(SpatLayers, TempLayers) {(0,0)2, (1,0)5.5, (2,0)7.5,

(0,1)3, (1,1)8, (2,1)12}FrameRate = f(TempLayers) {015, 130}

Name SyntaxHorizontal Resolution HorizRes ≤ 1024Vertical Resolution VertRes ≤ 768Bitrate BitRate ≤ 7

Example: Adaptation of Scalable Video (cont’d)

• Possible adaptation parameters

• Feasible adaptation parameters

• Optimal adaptation parameters


BitRate 2 5.5 7.5 3 8 12

SpatLayers 0 1 2 0 1 2

TempLayers 0 0 0 1 1 1

FrameRate 15 30

TempLayers 0 1

HorizRes 640 1024 1920

SpatLayers 0 1 2

BitRate 2 5.5 7.5 3 8 12



HorizRes 640 1024 1920

SpatLayers 0 1 2

BitRate 2 5.5 7.5 3 8 12



FrameRate 15 30

TempLayers 0 1

HorizRes 640 1024 1920

SpatLayers 0 1 2

Conclusions

Three steps to cross-layer interoperability

• Cross-Layer Model (XLM): describing the relationship between QoS metrics at different levels

• Instantiation of the XLM by utilizing MPEG-21 metadata

• Cross-Layer Adaptation Decision-Taking Engine (XL-ADTE)


Thank you for your attention

... questions, comments, etc. are welcome …

Ass.-Prof. Dipl.-Ing. Dr. Christian TimmererKlagenfurt University, Department of Information Technology (ITEC)

Universitätsstrasse 65-67, A-9020 Klagenfurt, [email protected]

http://research.timmerer.com/Tel: +43/463/2700 3621 Fax: +43/463/2700 3699

© Copyright: Christian Timmerer

152010/01/20 Christian Timmerer, Klagenfurt University, Austria