Handbook of Research on Wireless Multimedia: Quality of Service and Solutions

Handbook of Research on Wireless Multimedia: Quality of Service and Solutions Nicola Cranley Dublin Institute of Technology, Ireland Liam Murphy University College Dublin, Ireland

InformatIon scIence reference Hershey • New York

Director of Editorial Content: Senior Managing Editor: Managing Editor: Assistant Managing Editor: Managing Development Editor: Assistant Development Editor: Editorial Assistant: Copy Editor: Typesetter: Cover Design: Printed at:

Kristin Klinger Jennifer Neidig Jamie Snavely Carole Coulson Kristin M. Roth Deborah Yahnke Rebecca Beistline Larissa Vinci Sean Woznicki Lisa Tosheff Yurchak Printing Inc.

Published in the United States of America by Information Science Reference (an imprint of IGI Global) 701 E. Chocolate Avenue, Suite 200 Hershey PA 17033 Tel: 717-533-8845 Fax: 717-533-8661 E-mail: [email protected] Web site: http://www.igi-global.com and in the United Kingdom by Information Science Reference (an imprint of IGI Global) 3 Henrietta Street Covent Garden London WC2E 8LU Tel: 44 20 7240 0856 Fax: 44 20 7379 0609 Web site: http://www.eurospanbookstore.com Copyright © 2009 by IGI Global. All rights reserved. No part of this publication may be reproduced, stored or distributed in any form or by any means, electronic or mechanical, including photocopying, without written permission from the publisher. Product or company names used in this set are for identification purposes only. Inclusion of the names of the products or companies does not indicate a claim of ownership by IGI Global of the trademark or registered trademark. Library of Congress Cataloging-in-Publication Data Handbook of research on wireless multimedia : quality of service and solutions / Nicola Cranley and Liam Murphy, editors. p. cm. Summary: “This book highlights and discusses the underlying QoS issues that arise in the delivery of real-time multimedia services over wireless networks”--Provided by publisher. Includes bibliographical references and index. ISBN 978-1-59904-820-8 (hardcover) -- ISBN 978-1-59904-823-9 (ebook) 1. Multimedia communications--Quality control--Handbooks, manuals, etc. 2. Wireless communication systems--Handbooks, manuals, etc. I. Cranley, Nicola. II. Murphy, Liam. TK5105.15.H36 2009 621.384--dc22 2008009105

British Cataloguing in Publication Data A Cataloguing in Publication record for this book is available from the British Library. All work contributed to this book set is original material. The views expressed in this book are those of the authors, but not necessarily of the publisher.

If a library purchased a print copy of this publication, please go to http://www.igi-global.com/agreement for information on activating the library's complimentary electronic access to this publication.

List of Reviewers

Janet Adams Dublin City University, Ireland

Giovanni Giambene Università degli Studi di Siena, Italy

Fatih Alagöz Boðaziçi University, Turkey

Gürkan Gür Boðaziçi University, Turkey

Michael Barry University of Limerick, Ireland

Snezana Hadzic Università degli Studi di Siena

Suzan Bayhan Boðaziçi University, Turkey

Begoña Blanco Jauregi University of the Basque Country, Spain

Tarek Bejaoui University of Carthage, Tunisia

Jose Luis Jodra University of the Basque Country, Spain

Boris Bellalta Universitat Pompeu Fabra, Spain

Scott Jordan University of California, Irvine, USA

Janez Bester University of Ljubljana, Slovenia

William Kent University of Limerick, Ireland

Graça Bressan Escola Politécnica da Universidade de São Paulo, Brazil

Andrej Kos University of Ljubljana, Slovenia

Paolo Chini Università degli Studi di Siena, Italy Imrich Chlamtac Create-Net, Italy Nikki Cranley University College Dublin, Ireland Francesco De Pellegrini Create-Net, Italy Gheorghita Ghinea Brunel University, UK

Harilaos G. Koumaras N.C.S.R “Demokritos”, Greece Ming Li California State University, Fresno, USA Fidel Liberal University of the Basque Country, Spain Carlos Macian Universitat Pompeu Fabra, Spain Andreas Mäder University of Würzburg, Germany

Michael M. Markou University of Cyprus, Nicosia, Cyprus

Tacha Serif Brunel University, UK

Sean Mc Grath University of Limerick, Ireland

Anna Sfairopoulou Universitat Pompeu Fabra, Spain

Gabriel-Miro Muntean Dublin City University, Ireland

Ronan Skehill University of Limerick, Ireland

Liam Murphy University College Dublin, Ireland

Dirk Staehle University of Würzburg, Germany

Nidal Nasser University of Guelph, Canada

Lingfen Sun University of Plymouth, UK

John Nelson University of Limerick, Ireland

Vassilis Tsaoussidis Democritus University of Thrace, Greece

Christos G. Panayiotou University of Cyprus, Nicosia, Cyprus

Mojca Volk University of Ljubljana, Slovenia

Panagiotis Papadimitriou Democritus University of Thrace, Greece

Marcio Nieblas Zapater Escola Politécnica da Universidade de São Paulo, Brazil

Dorel Picovici University of Limerick, Ireland Roberto Riggio Create-Net, Italy

Peifang Zhang University of California, Irvine, USA

Table of Contents

Foreword ............................................................................................................................................ xvi Preface ..............................................................................................................................................xviii

Section I Network Quality of Service Chapter I Evaluating QoS in a Multi-Access Wireless Network ........................................................................... 1 Ronan Skehill, University of Limerick, Ireland William Kent, University of Limerick, Ireland Dorel Picovici, University of Limerick, Ireland Michael Barry, University of Limerick, Ireland Sean McGrath, University of Limerick, Ireland Chapter II Radio Resource Management Strategies for HSDPA-Enhanced UMTS Networks ............................ 31 Dirk Staehle, University of Würzburg, Germany Andreas Mäder, University of Würzburg, Germany Chapter III Handoff Management in Next Generation Wireless Networks ........................................................... 55 Nidal Nasser, University of Guelph, Canada Tarek Bejaoui, University of Carthage, Tunisia Chapter IV Resource Management in IEEE 802.11 Based Wireless Networks ..................................................... 77 Ming Li, California State University, Fresno, USA Roberto Riggio, Create-Net, Italy Francesco De Pellegrini, Create-Net, Italy Imrich Chlamtac, Create-Net, Italy

Section II Application Quality of Service Chapter V Adaptive Codec Selection for VoIP in Multi-Rate WLANs .............................................................. 122 Anna Sfairopoulou, Network Technologies and Strategies (NeTS) Research Group, Universitat Pompeu Fabra, Spain Carlos Macián, Network Technologies and Strategies (NeTS) Research Group, Universitat Pompeu Fabra, Spain Boris Bellalta, Network Technologies and Strategies (NeTS) Research Group, Universitat Pompeu Fabra, Spain Chapter VI Buffer Control Techniques for QoS Provisioning in Wireless Networks .......................................... 157 Michael M. Markou, University of Cyprus, Cyprus Christos G. Panayiotou, University of Cyprus, Cyprus Chapter VII Power Saving in Wireless Multimedia Streaming to Mobile Devices ............................................... 183 Gabriel-Miro Muntean, Dublin City University, Ireland Janet Adams, Dublin City University, Ireland Chapter VIII Multimedia Services Provision in MANETs ..................................................................................... 203 Jose Luis Jodra, University of the Basque Country, Spain Fidel Liberal, University of the Basque Country, Spain Begoña Blanco Jauregi, University of the Basque Country, Spain

Section III End-User Quality of Service Chapter IX Quality Assurance in the IMS-Based NGN Environment ................................................................. 240 Andrej Kos, University of Ljubljana, Slovenia Mojca Volk, University of Ljubljana, Slovenia Janez Bester, University of Ljubljana, Slovenia Chapter X Quality of Experience for Video Services .......................................................................................... 258 Marcio Nieblas Zapater, Escola Politécnica da Universidade de São Paulo, Brazil Graça Bressan, Escola Politécnica da Universidade de São Paulo, Brazil

Chapter XI Perceptual Voice Quality Measurements for Wireless Networks ...................................................... 274 Dorel Picovici, Institute of Technology Carlow, Ireland John Nelson, University of Limerick, Ireland Chapter XII Enhancing the Multimedia Tour Guide Experience: Transmission Tailoring Based on Content, Location, and Device Type .................................................................................. 296 Tacha Serif, Brunel University, UK Gheorghita Ghinea, Brunel University, UK Chapter XIII PQoS Assessment Methods for Multimedia Services ........................................................................ 316 Harilaos Koumaras, National Center of Scientific Research, “Demokritos,” Greece Fidel Liberal, University of the Basque Country, Spain Lingfen Sun, University of Plymouth, UK

Section IV Cross-Layered Solutions Chapter XIV Scheduling and Access Control for Wireless Connections with Throughput Guarantees ................. 353 Peifang Zhang, University of California, Irvine, USA Scott Jordan, University of California, Irvine, USA Chapter XV Broadband Satellite Multimedia Networks ........................................................................................ 377 Paolo Chini, Università degli Studi di Siena, Italy Giovanni Giambene, Università degli Studi di Siena, Italy Snezana Hadzic, Università degli Studi di Siena, Italy Chapter XVI End-to-End Support for Multimedia QoS in the Internet .................................................................. 398 Panagiotis Papadimitriou, Democritus University of Thrace, Greece Vassilis Tsaoussidis, Democritus University of Thrace, Greece Chapter XVII Cross-Layer Radio Resource Management Protocols for QoS Provisioning in Multimedia Wireless Networks ..................................................................................................... 417 Tarek Bejaoui, University of Carthage, Tunisia Nidal Nasser, University of Guelph, Canada

Chapter XVIII Transport Protocols and QoS for Wireless Multimedia ..................................................................... 442 Gürkan Gür, Satellite Networks Research Laboratory (SATLAB), Boğaziçi University, Turkey Suzan Bayhan, Satellite Networks Research Laboratory (SATLAB), Boğaziçi University, Turkey Fatih Alagöz, Satellite Networks Research Laboratory (SATLAB), Boğaziçi University, Turkey

Compilation of References .............................................................................................................. 465 About the Contributors ................................................................................................................... 505 Index ................................................................................................................................................ 515

Detailed Table of Contents

Foreword ............................................................................................................................................ xvi Preface ..............................................................................................................................................xviii

Section I Network Quality of Service In this section Network Layer QoS solutions are presented. Critical factors affecting QoS for realtime multimedia streaming applications include reliability, efficient delivery, and end-to-end latencies. The tremendous growth and development of wireless networking technology has brought about fresh challenges in the provision of QoS for such real-time multimedia applications. This section focuses on key wireless networking technologies and presents novel solutions that have been designed to address these challenges. State of the art wireless technologies presented in this section include UMTS, HSDPA-Enhanced UMTS, 4G, and WLAN. Each of these wireless networking technologies has differing mechanisms for QoS support. Resource management schemes and admission control schemes can be used to prevent the network becoming overloaded such that poor network performance begins to negatively affect multimedia applications. Chapter I Evaluating QoS in a Multi-Access Wireless Network ........................................................................... 1 Ronan Skehill, University of Limerick, Ireland William Kent, University of Limerick, Ireland Dorel Picovici, University of Limerick, Ireland Michael Barry, University of Limerick, Ireland Sean McGrath, University of Limerick, Ireland This chapter introduces quality of service (QoS) in multi-access wireless networks, and demonstrates how QoS is implemented in IEEE 802.11 and UMTS. The chapter explains how these complementary technologies, when coupled together, provide a network that is greater than its individual parts. However, combining these networks brings new network management challenges, and so the impact of joint admission control strategies on QoS is evaluated. The evaluation results show that when UMTS and WLAN are tightly coupled, the end user enjoys a higher level of QoS.

Chapter II Radio Resource Management Strategies for HSDPA-Enhanced UMTS Networks ............................ 31 Dirk Staehle, University of Würzburg, Germany Andreas Mäder, University of Würzburg, Germany This chapter gives an overview of the background and functionality of the high-speed downlink packet access (HSDPA), and provides insights into the radio resource management of integrated UMTS/HSDPA networks. The authors introduce aspects of radio resource management specific to the HSDPA like channel-aware scheduling and radio resource sharing strategies. Furthermore, the impact of radio resource management on the quality of service is analyzed and it is shown that the selection of an RRM strategy is an integral part of the network planning and deployment process. Chapter III Handoff Management in Next Generation Wireless Networks ........................................................... 55 Nidal Nasser, University of Guelph, Canada Tarek Bejaoui, University of Carthage, Tunisia Major research challenges in the next generation of wireless networks include the provisioning of worldwide seamless mobility across heterogeneous wireless networks, the improvement of end-to-end quality of service (QoS), supporting multimedia services over wide area and enabling users to specify their personal preferences. The integration and interoperability of this multitude of available networks will lead to the emergence of the fourth generation (4G) of wireless technologies. One of the 4G challenges is the user’s ability to control and manage handoffs across heterogeneous wireless networks. This chapter proposes a solution to this problem using artificial neural networks. Chapter IV Resource Management in IEEE 802.11 Based Wireless Networks ..................................................... 77 Ming Li, California State University, Fresno, USA Roberto Riggio, Create-Net, Italy Francesco De Pellegrini, Create-Net, Italy Imrich Chlamtac, Create-Net, Italy This chapter provides a comprehensive review of the architectures, algorithms, and protocols for resource management in IEEE 802.11 based wireless networks such as wireless LANs, heterogeneous wired/wireless networks, mobile ad hoc networks, and wireless mesh networks. The focus is on the approaches for bandwidth allocation in different networks, as well as how these strategies are incorporated in each specific protocol.

Section II Application Quality of Service In this section, application layer techniques are presented to facilitate QoS for multimedia streaming services. There are two facets to Application-layer QoS: server-side and client-side techniques. Serverside techniques cover the encoding configuration, adaptive encoding algorithms, error resilience, and

preparation of the multimedia stream for delivery over the network. Client-side techniques include buffering, power saving, and decoding techniques such as error concealment and error recovery techniques. Through optimization of server-side and client-side encoding and decoding options, the application can compensate for and overcome limitations in the network and provide QoS for real-time multimedia streaming applications. This section presents novel server-side and client-side application layer solutions for voice and video streaming applications. Chapter V Adaptive Codec Selection for VoIP in Multi-Rate WLANs .............................................................. 122 Anna Sfairopoulou, Network Technologies and Strategies (NeTS) Research Group, Universitat Pompeu Fabra, Spain Carlos Macián, Network Technologies and Strategies (NeTS) Research Group, Universitat Pompeu Fabra, Spain Boris Bellalta, Network Technologies and Strategies (NeTS) Research Group, Universitat Pompeu Fabra, Spain This chapter introduces the problems caused to voice over IP calls on 802.11 networks due to the link adaptation mechanism. It provides an overview of all the components participating in this study, with special emphasis on the multi-rate anomaly. The authors propose a codec selection mechanism as a solution to the multi-rate problem which, by changing the codec of some of the calls at the moment of the rate change, tries to maintain delay and packet loss values at acceptable levels and provide the desired QoS for the voice flows. Chapter VI Buffer Control Techniques for QoS Provisioning in Wireless Networks .......................................... 157 Michael M. Markou, University of Cyprus, Cyprus Christos G. Panayiotou, University of Cyprus, Cyprus This chapter introduces network buffer control techniques as a means to provide QoS. This problem has been extensively studied in the context of wired networks; however, the proliferation of wireless networks and the introduction of multimedia applications has significantly changed the characteristics of the traffic mix that flows on the network. The objective of this chapter is to create a new methodology for automatically adapting the various buffer thresholds such that the network exhibits optimal or near optimal performance even as network conditions change. Chapter VII Power Saving in Wireless Multimedia Streaming to Mobile Devices ............................................... 183 Gabriel-Miro Muntean, Dublin City University, Ireland Janet Adams, Dublin City University, Ireland When a mobile device has wireless LAN capability, multimedia content can be streamed over a wireless network to that device. However, a major disadvantage of all mobile devices is their limited battery lifetime. Multimedia streaming puts extra pressure on the battery, causing it to discharge faster. This chapter describes adaptive buffer power save mechanism (AB-PSM), a novel power saving wireless communication solution that enables an increase in battery lifetime during mobile multimedia streaming.

Chapter VIII Multimedia Services Provision in MANETs ..................................................................................... 203 Jose Luis Jodra, University of the Basque Country, Spain Fidel Liberal, University of the Basque Country, Spain Begoña Blanco Jauregi, University of the Basque Country, Spain This chapter introduces the principal characteristics of MANETs (mobile ad hoc networks) and shows how these may affect both QoS conditions and QoS management/provisioning systems, and therefore the capabilities of MANETs for properly providing multimedia services. The authors claim that QoS management cannot be handled only at the network layer or by applying some QoS-aware routing protocols. In fact, any end-to-end QoS provision architecture will demand QoS control mechanisms and information exchange among all the layers.

Section III End-User Quality of Service The ultimate goal of optimization techniques at the network and application layer is to ensure enduser perceived QoS. Often end-user perceived QoS measurement techniques are incorporated into network and application layer QoS solutions in order to provide greater insight into the users’ QoS experience. It is only through accurate measurement of end-user perceived quality that QoS schemes can be developed and optimised. End-user QoS can be assessed using either objective or subjective methodologies. However, typically subjective testing results form the knowledge base and enhance the accuracy of objective metrics, so that objective metrics can then be incorporated into network and application layer QoS schemes. This section presents subjective results, targeted objective metrics and adaptive with integrated objective metrics. Moreover, since end-user perceived QoS is application and media dependent, targeted objective metrics are presented for voice, video and multimedia. Chapter IX Quality Assurance in the IMS-Based NGN Environment ................................................................. 240 Andrej Kos, University of Ljubljana, Slovenia Mojca Volk, University of Ljubljana, Slovenia Janez Bester, University of Ljubljana, Slovenia Commonly understood as the next generation networks (NGN), a composite environment of proven telecommunications and Internet-oriented mechanisms has become generally recognized as the telecommunications environment of the future. However, the nature of the NGN environment presents several complex issues regarding quality assurance that have not existed in the legacy environments. In this chapter, a service-aware policy based approach to NGN quality assurance is presented, taking into account both perceptual quality of experience and technology-dependant quality of service issues. Chapter X Quality of Experience for Video Services .......................................................................................... 258 Marcio Nieblas Zapater, Escola Politécnica da Universidade de São Paulo, Brazil Graça Bressan, Escola Politécnica da Universidade de São Paulo, Brazil

This chapter discusses the quality assurance of multimedia services over IP networks from the end user standpoint and introduces the concept of quality of experience (QoE). The discussion of quality assurance includes aspects that range from the network and application layers to the end user perspective. This chapter presents quality requirements for video and TV services and performance measures that focuses on the quality perceived by the end user. This approach is broader than that oriented to quality of service (QoS), which focuses on the performance measures from the network perspective. Chapter XI Perceptual Voice Quality Measurements for Wireless Networks ...................................................... 274 Dorel Picovici, Institute of Technology Carlow, Ireland John Nelson, University of Limerick, Ireland Perceptual voice quality measurement can be defined as an objective quantification of an overall impression of the perceived stimulus. An alternative to laborious subjective testing is objective predictive modeling, which employs a perceptual model of the human auditory and cognitive system to predict the human response to a voice signal in terms of its quality. This chapter describes subjective and automated objective testing methods, and provides a test case scenario for measuring voice quality. Chapter XII Enhancing the Multimedia Tour Guide Experience: Transmission Tailoring Based on Content, Location, and Device Type .................................................................................. 296 Tacha Serif, Brunel University, UK Gheorghita Ghinea, Brunel University, UK This chapter describes an investigation into user experiences of accessing streamed multimedia content when that content is tailored according to perceptual, device, and location characteristics. We propose that multimedia transmission to mobile and wireless devices should be made based on pre-defined profiles, which contain a combination of static (perceptual, device type, CPU speed, and display specifications) and dynamic information (streamed content type location of the device/user, context of the device/user). Furthermore, we believe that using profiling technology, mobile service providers can effectively manage local network traffic and cut down their bandwidth costs considerably. Chapter XIII PQoS Assessment Methods for Multimedia Services ........................................................................ 316 Harilaos Koumaras, National Center of Scientific Research, “Demokritos,” Greece Fidel Liberal, University of the Basque Country, Spain Lingfen Sun, University of Plymouth, UK The concept of PQoS (end-to-end perceived QoS), although in general it deals with the user satisfaction with a specific delivered/requested service, is in practice significantly differentiated by the nature of each delivered service. This chapter reviews various existing PQoS assessment methods for video, VoIP, on-line games and web services that have been published in the literature. It then moves beyond the current PQoS assessment methods and presents novel techniques for predicting the PQoS of a multimedia service.

Section IV Cross-Layered Solutions Cross-layered solutions make use of QoS-enabling features at various layers of the wireless multimedia service chain, including aspects of the application layer and network layer. Although cross-layered solutions can greatly increase the degree of optimisation of the end-user QoS, cross-layered solutions incur higher orders of complexity since they have a larger number of axes of optimisation. This section presents a number of cross-layered solutions that include algorithms at the network, application and end-user layers. Chapter XIV Scheduling and Access Control for Wireless Connections with Throughput Guarantees ................. 353 Peifang Zhang, University of California, Irvine, USA Scott Jordan, University of California, Irvine, USA Emerging wideband code division multiple access (WCDMA) data services will likely require resource allocation to ensure that throughput targets are met. Scheduling and access control can both be key components in this task. In this chapter, we introduce a two layer scheduler and connection access controller that attempts to balance efficiency with fairness. Through numerical analysis, the proposed joint scheduler and connection access controller is shown to achieve this. Chapter XV Broadband Satellite Multimedia Networks ........................................................................................ 377 Paolo Chini, Università degli Studi di Siena, Italy Giovanni Giambene, Università degli Studi di Siena, Italy Snezana Hadzic, Università degli Studi di Siena, Italy Nowadays there is an increasing need of broadband communication anytime, anywhere for users that expect to receive multimedia services with QoS. In such a scenario, the aim of this chapter is to present the satellite option to bridge the digital divide in those areas where terrestrial solutions are infeasible or too expensive. We provide a survey of the ETSI standardization framework for satellite networks, and then describe resource management schemes for both forward and return link. Finally a case study is provided for the integration of a DVB-S/DVB-RCS satellite system interconnected with a WiFi segment for local coverage. Chapter XVI End-to-End Support for Multimedia QoS in the Internet .................................................................. 398 Panagiotis Papadimitriou, Democritus University of Thrace, Greece Vassilis Tsaoussidis, Democritus University of Thrace, Greece An increasing demand for multimedia data delivery coupled with reliance on best-effort networks, such as the Internet, has spurred interest on effective quality of service (QoS) management for multimedia streams. Since today’s multimedia applications are expected to run in physically heterogeneous envi-

ronments composed of both wired and wireless components, we assess the efficiency of transport-layer solutions for multimedia traffic in heterogeneous networks. This chapter also describes the perceptual QoS assessment of voice and video streams. Chapter XVII Cross-Layer Radio Resource Management Protocols for QoS Provisioning in Multimedia Wireless Networks ..................................................................................................... 417 Tarek Bejaoui, University of Carthage, Tunisia Nidal Nasser, University of Guelph, Canada This chapter introduces the cross layer design for resource allocation over multimedia wireless networks. Conventional layered packet scheduling and call admission control schemes are presented and a number of cross-layered protocols that are recently proposed are investigated. The chapter highlights the QoS improvement and the performance gain obtained while considering the interlayer dependencies concept for various real-time and non-real-time applications. Chapter XVIII Transport Protocols and QoS for Wireless Multimedia ..................................................................... 442 Gürkan Gür, Satellite Networks Research Laboratory (SATLAB), Boğaziçi University, Turkey Suzan Bayhan, Satellite Networks Research Laboratory (SATLAB), Boğaziçi University, Turkey Fatih Alagöz, Satellite Networks Research Laboratory (SATLAB), Boğaziçi University, Turkey This chapter introduces the QoS issues and support in transport protocols for wireless multimedia transmission. Some of the proposed modifications to the TCP and UDP protocols in order to improve multimedia transmission quality in wireless networks are also summarized. In particular, UDP Lite, TCP friendly rate control protocol (TFRC), and real-time transport protocol (RTP)--Real-time transport control protocol (RTCP) are discussed. Finally, we conclude with some discussions on the current trends in transport protocols for wireless multimedia transmission and on some of the ongoing research issues.

Compilation of References .............................................................................................................. 465 About the Contributors ................................................................................................................... 505 Index ................................................................................................................................................ 515

xvi

Foreword

In the past two decades, we have witnessed tremendous advances in wireless technologies, in particular those aimed at personal and mobile communications using cellular and ad hoc configurations. Cellular mobile communication, considered a luxury in the early 1990s, has become one of the everyday necessities for hundreds of millions of people all around the world in less than 20 years. Applications have changed dramatically from simple voice telephony to a wide range of multimedia applications. Exponential advancement in VLSI technology and liquid crystal display at the same time provided telecommunications engineers with an unbelievable electronic gadget in people’s pockets. Ubiquitous communications has therefore become a reality and an all-in-one device is no longer a dream. Television broadcasting has found a new direction through the digital era, from large wall-mounted displays to the smaller and more private displays of mobile phones. Wireless communications has had to develop at the same pace as its hardware and software counterparts in mobile devices so that they can be connected to content providers over the Internet and the telecommunications backbones. New wireless technologies have been added to the single cellular air interface mobile phones. These days we see smart phones with several air interfaces, all built on a tiny chip. They can connect simultaneously to wireless local area networks; second-generation networks such as GSM and GPRS; third generation networks like UMTS; and Bluetooth, in multiple frequency spectra. Those devices sometimes even come with their own satellite navigation system, which can locate the device and provide further information to users. With the inclusion of Windows-based operating systems on mobile devices, the user device is no longer just a phone but a handy personal computer with the usual myriad applications. With all these advances, mobile multimedia is in our hands and the important issue is how the service quality can be maintained at a level similar to what we had in the past and which users have come to expect. The topic of mobile multimedia quality of service therefore remains the most important issue to be dealt with by telecommunications engineers. In the past ten years, we have seen many works in the literature on the topic of quality of service in mobile environment. Dr. Nicola Cranley and Prof. Liam Murphy have put together an excellent edition of chapters, carefully chosen, reviewed, and edited in their book covering the technical solutions to this problem. They break down the problem nicely into three parts: network layer, application layer, and end-user layer, which can serve as the main elements in providing end-to-end quality of service to mobile multimedia applications. As quality of service provisioning requires good cooperation among communication layers and is not achievable by individual layer’s attempts, the last part of the book addresses cross-layer solutions to the problem. Nicola and Liam have selected a wide range of experts from all over the world to detail the problems and possible solutions in this harmonized edition. The book, while written by many authors, is read as a single piece of work with a focused and understandable theme right throughout the entire edition. I

xvii

believe Wireless Multimedia: Quality of Service and Solutions will stand out as a long lasting reference book in the field of mobile multimedia for many years to come. I am confident that the tutorials and research works presented in this book will further seed new research topics in the field for a better and more efficient use of hardware and software advancements to achieve mobile multimedia communications into the future.

Abbas Jamalipour, Fellow IEEE Sydney, Australia Abbas Jamalipour holds a PhD from Nagoya University, Japan. He is the author of the first book on wireless IP and two other books, and has co-authored six books and over 180 technical papers, all in the field of mobile communications networks. He is a fellow of IEEE (for contributions to next generation networks for traffic control), a fellow of Institute of Engineers Australia; an IEEE distinguished lecturer; the editor-in-chief of the IEEE Wireless Communications; and a technical editor of several scholarly journals including IEEE Communications, Wiley International Journal of Communication Systems, Journal of Communication Network, etc. His areas of research are wireless data communication networks, wireless IP networks, next generation mobile networks, traffic control, network security and management, and satellite systems. He was one of the first researchers to disseminate the fundamental concepts of the next generation mobile networks and broadband convergence networks as well as the integration of wireless LAN and cellular networks; some of which are being gradually deployed by industry and included in the ITU-T standards. Prof. Jamalipour has authored several invited papers and been a keynote speaker in many prestigious conferences. He served as the chair of the Satellite and Space Communications Technical Committee (2004-06); and currently is the vice chair of Communications Switching and Routing TC; and chair of Chapters Coordinating Committee, Asia-Pacific Board, all from the IEEE Communications Society. He is a voting member of the IEEE GITC and IEEE WCNC Steering Committee. He has been a vice chair of IEEE WCNC2003 to 2006, program chair of SPECTS2004, chair of symposiums at IEEE GLOBECOM2005 to 2007 and IEEE ICC2005 to 2008, among many conference leadership roles. He has received several prestigious awards, such as 2006 IEEE Distinguished Contribution to Satellite Communications Award, 2006 IEEE Communications Society Best Tutorial Paper Award, and 2005 Telstra Award for Excellence in Teaching.

xviii

Preface

The rapid advances in wireless technologies have brought about a demand for high quality multimedia applications and services such as video telephony, multimedia streaming, video games, audio streaming (e.g., podcasting, IP HDTV broadcasting), and voice over IP. These advanced multimedia services bring a new set of challenges for providing quality of service (QoS) for delivering these services over wireless networks. Wireless technologies are becoming increasingly sophisticated and efficient enabling support for higher bit rates. However, high and variable error rates and delays in wireless systems are still significant obstacles for providing QoS support for multimedia applications, especially when such variations occur on short timescales with respect to the applications being supported. Multimedia applications, in particular, impose significant resource requirements on bandwidth constrained wireless networks. Under these conditions it is difficult to provide any QoS guarantees. In particular the delay constraints associated with real-time multimedia pose the greatest challenge. Real-time multimedia is particularly sensitive to delay, as multimedia packets require a strict bounded end-to-end delay (i.e., every multimedia packet must arrive at the client device before its playout time with enough time to decode and display the contents of the packet). If the multimedia packet does not arrive on time, the packet is effectively lost and this affects the end-user perceived quality. QoS is a crucial part of wireless multimedia design and delivery. Poor QoS results in poor service uptake by users which will result in the potential offered by recent advances in wireless and multimedia technologies not fully utilized. There are many aspects to QoS provisioning. These include network-layer QoS, application-layer QoS and ultimately end-user QoS. Network-layer QoS is concerned with the reliable and fast delivery of multimedia data over the wireless technologies. Many new and emerging wireless technologies such as IEEE 802.11e have been designed and developed with integrated QoS enabling controls. However these controls need to be configured and optimized in order to provide Application-layer QoS. Application-layer QoS on the other hand is concerned with the quality of the multimedia encoding, delivery, adaptation, decoding and play out on the client device. End-user QoS is concerned with the end-user experience in terms of audio and visual quality. Typically these QoS layers are treated independently and in isolation yet the QoS schemes implemented at each of these layers have an effect on each other. It is essential for network managers, engineers and application developers to have an understanding of the QoS schemes that are in place at the network, application and end-user layer in order to be able to provide a fully end-to-end QoS solution. The ultimate goal of these QoS schemes is to maximize end-user QoS. With such diversity of QoS issues for multimedia and wireless technologies, there is an opportunity for novel QoS techniques to be developed at all layers.

xix

Objectives and structure Of this bOOk The objective of this book is to present state of the art research that tackles the challenges of providing QoS for multimedia services of wireless technologies. There are many aspects to QoS provisioning. We have identified three key layers for QoS provisioning. These include network-layer QoS, applicationlayer QoS, and ultimately end-user QoS. We have structured this book to represent the latest state of the art research in each of these three layers and have also included a section that covers Cross-layered solutions which touch two or more of these QoS layers. This book is intended to: • • •

Identify each of these different layers of QoS and highlight the underlying QoS issues that arise and affect the performance of multimedia applications over wireless networks. Present the QoS issues that arise with different types of multimedia applications and services over different types of wireless technologies. Present novel solutions and state of the art research that has been done to address QoS issues for different wireless multimedia applications.

The first section of this book deals with the network-layer QoS. The primary characteristic of next generation wireless and mobile communication systems is heterogeneity, for example, wireless cellular networks, wireless local area networks, wireless personal area networks. It is crucial for inter-operability and seamless roaming among these different networks. Wireless technologies have a finite bandwidth capacity and are error prone media, it is important for wireless networks employ radio resource management schemes and optimized QoS handling to satisfy and meet the needs of users of multimedia services. With the development of multimedia compression and coding technologies, more and more real-time applications, such as video and audio, and the proliferation of pervasive devices create a new demand for wireless multimedia communication systems. Multimedia services can be optimised and adapted to the challenges of wireless devices and service delivery. The second section deals with application-layer QoS. This encompasses quality-aware multimedia encoding, delivery, adaptation, decoding and buffering on the client device. The third section of this book deals with end-user QoS. End-user QoS is concerned with the end-user experience in terms of audio and visual quality. Traditionally, a reactive approach has been adopted for engineering QoS for multimedia streaming applications. The main problem with this approach is that a poorly designed system cannot be tuned to perform as well as a system that was well designed from the outset. By integrating end-user QoS into the design of multimedia streaming systems, a more proactive approach can be adopted for the development and delivery of multimedia services over wireless networks. Finally, the last section of this book presents cross-layer QoS solutions. Cross-layered solutions are extremely difficult since they rely on optimising QoS at two or more layers of the multimedia system, increasing the complexity of the optimisation; but also providing greater flexibility and potential to appropriately tune the network, application, and end-user layer to achieve the desired QoS.

network Qos With the recent advances in microelectronics, mobile devices now come equipped with a range of different wireless technologies on the one device. For example many PDA devices now come equipped with WLAN, 2.5G/3G cellular, Bluetooth, and Infrared. Often different wireless networks co-exist al-

xx

lowing users to switch between different wireless technologies. Switching between different wireless technologies is an important and difficult challenge since they have different physical layer, link layer and MAC layer schemes which result in a large variation in bandwidth and end-to-end delays making seamless transitions between different wireless technologies and continuous QoS difficult. There are many performance-related issues associated with the delivery multimedia applications over wireless networks. Among the most significant are finite bandwidth resources, high channel error rates, contention between users for access to bandwidth, collisions, signal attenuation with distance, signal interference, etc. There are a number of techniques that have been developed and integrated into wireless technologies in order to facilitate the provisioning of network-layer QoS. The most well-known mechanisms are the integrated services (IntServ) and the differentiated services (DiffServ). Different wireless technologies such as general packet radio service (GPRS)/universal mobile telecommunications system (UMTS) and IEEE 802.11e have very different mechanisms for QoS support. Resource management schemes and admission control schemes can be used to prevent the network becoming overloaded such that poor network performance begins to negatively affect the multimedia applications. In Chapter I, Skehill et al. describe how QoS can be provided in multi-access networks in particular UMTS and 802.11 networks. They demonstrate the new network management challenges and QoS provisioning problems posed by heterogeneous networks. To address this challenge they present their work on joint admission control strategies that can be employed to provide QoS. They demonstrate their system on an advanced test platform that replicates an integrated Release 4 UMTS network and standard IEEE 802.11b network. Their results show that by tightly coupling UMTS and WLAN technologies, the end user enjoys a higher level of quality of service. In Chapter II, Staehle and Mäder describe high speed downlink packet access (HSDPA), and provide some insights into the radio resource management of integrated UMTS/HSDPA networks. The development of HSDPA was initiated as response to an increasing demand for high-speed mobile Internet access. HSDPA enables data rates of several megabits per seconds with packet latencies under 100ms. This chapter covers aspects of radio resource management specific to HSDPA such as channel-aware scheduling and radio resource sharing strategies and analyzes the impact of radio resource management on the achievable quality of service. With pervasive and diverse wireless networks, users roam and move between heterogeneous networks. In Chapter III, Nasser and Bejaoui describe the challenges for next generation of wireless networks to provide seamless mobility across heterogeneous wireless networks and QoS provisioning to support multimedia services. They describe how 4G wireless technologies offer great potential to meet the challenges posed by multimedia services and change the way mobile devices are used and can be adapted to provide a wide variety of new multimedia applications. One of the key challenges in heterogeneous network environments is the ability to control and manage handoffs. Nasser et al. describe their solution using artificial neural networks (ANNs) to identify the best existing wireless network that matches predefined user preferences set on a mobile device when performing a vertical handoff. Multimedia applications are bandwidth hungry and resource demanding services. Providing multimedia services over bandwidth constrained wireless networks is challenging. In Chapter IV, Li et al. provide a rich and comprehensive overview of the architectures, algorithms, and protocols that are employed as radio resource management in IEEE 802.11 based wireless networks, mobile ad hoc networks, and wireless mesh networks. Li and Riggio demonstrate that with a successful resource management strategy, bandwidth usage can be minimized while providing maximized QoS and end-user QoS for multimedia applications.

xxi

application Qos There are several aspects to application-layer QoS that deal with all stages of the applications lifecycle including encoding, delivery, adaptation, decoding, error correction, and error concealment. Before multimedia can be streamed over the network, the multimedia content is encoded and prepared for transmission. The choice of the right encoding settings is crucial for the performance of the delivery of the multimedia stream over the network. For audio, voice, and video content, there are a number of different encoding schemes and encoding configuration parameters that can be used to optimally encode the multimedia content, which has an effect on the network-layer QoS requirements of the multimedia stream. When the content is ready to be delivered over the network, the encoded multimedia data is packetized with an optimal number of samples contained in each packet. However when the network is overloaded and congested, its ability to reliably deliver the multimedia packets within the strict delay constraints is reduced. Such delays can result in packets being lost at the client device since they arrived too late to meet the network-layer QoS delay requirements. Packets can also be lost due to errors over the wireless channel. The loss of multimedia data in the network has a significant impact on the end-user perceived QoS. To overcome this, many multimedia streaming systems have adaptation capabilities whereby the offered multimedia content is adapted in a seamless and imperceptible manner in order to minimize the unwanted negative effects of poor network conditions on the multimedia stream. On the client device, the multimedia player application has a number of techniques to recover any errors or losses in the received stream. Error correction and error concealment algorithms interpolate the missing multimedia data from the received data in order and mask these errors to improve the end-user perceived quality in spite of these losses. With the proliferation of WLAN technology, and the explosive growth of VoIP services, Chapter V by Sfairopoulou et al. describes their work for delivering high quality VoIP services over 802.11 networks. They describe how voice codecs can be optimized and adapted to the network conditions in order to provide the best QoS, for example, through the use of multi-rate voice codecs and link adaptation. They present a novel solution to the multi-rate and codec selection mechanism, which maintains the end-to-end delay and packet loss values within acceptable levels and provide the desired QoS for the voice flows. Chapter VI by Markou and Panayiotou describes how network buffer control techniques can be used as a means to provide QoS. Although this solution has been applied to wired networks, with the explosive growth of wireless access technologies and the challenges posed by multimedia applications, buffer control schemes have had to significantly adapt to the characteristics wireless networks. This chapter presents a new methodology for automatically adapting the various buffer thresholds such that the network exhibits optimal or near optimal performance even as network conditions change making it ideal for wireless multimedia service delivery. With the large number of wireless devices and their growing capabilities, power consumption and battery lifetimes are still significant obstacles in the continuous seamless play out of multimedia services on such devices. In Chapter VII, Muntean and Adams consider the power consumption problems presented by WLAN enabled mobile devices. Decoding and playout of multimedia services adds to the processing overhead causing it to discharge faster. In this chapter, Adams and Muntean describe an adaptive buffer power save mechanism (AB-PSM) that enables an increase in battery lifetime during mobile multimedia streaming which in turn increases the end-user QoS. In Chapter VIII, Jodra et al. describe the new application scenarios for mobile ad hoc networks, including multimedia services and/or online games. They discuss how the demand for such wireless connectivity requires certain levels of quality of service. For this reason, a large effort has been carried

xxii

out by both the industry and research community toward QoS provisioning in MANETs. However, when analyzing QoS aspects in MANETs, the first thing to notice is that MANETs’ special characteristics prevent the use of traditional QoS mechanisms in these environments. These particularities are related not only to the wireless transmission medium itself (which is shared with other kinds of wireless networks) but also to the uncertainty derived from the mobility of the nodes and subsequent network topology variations. Hence, major challenges to be faced while defining QoS mechanisms for MANETs include the wireless channel, multihop nature of communications, node mobility, lack of centralized control, dynamic network topology and limited device resources. These challenges result into a series of constraints to be added to QoS traditional ones (namely bandwidth, delay or jitter), such as battery power, CPU usage and stability of the routes. Most of these constraints demand different QoS mechanisms in different protocol layers, ranging from physical layer and MAC contention mechanisms to application level. However, since MANETs particularities are related to network topologies most of the efforts analyzed aimed at developing routing strategies focused on dealing with dynamic topology problems.

end-user Qos End-user QoS is the primary goal of application-layer QoS schemes and a somewhat secondary goal of network-layer QoS schemes. Measuring end-user QoS is an extremely complicated task which draws from many knowledge domains such as psychology, cognitive science and signal analysis. There are two key methods for assessing end-user QoS; the first is through subjective assessment and testing, while the second is through the use of objective metrics. Different objective metrics exist for audio, video and voice quality analysis. The main goal of objective metrics is to measure the perceived quality of a given audio or visual signal. There are many factors that affect how users perceive quality, such as audio loudness, lip synchronization, video content, viewing distance, display size, resolution, brightness, contrast, sharpness/fidelity, and color. However, it is only through the accurate measurement of end-user QoS that QoS schemes can be developed and optimized. In Chapter IX, Kos et al. present a service-aware policy based approach to next generation networks (NGN) quality assurance, considering both perceptual quality of experience and technology-dependent quality of service issues. The nature of the NGN environment presents several complex issues regarding quality assurance not faced in legacy environments, such as the multi-network, multi-vendor, and multioperator IP-based telecommunications environment, distributed intelligence, third-party provisioning, and fixed-wireless access. The existence of multiple separately operated and interconnected domains requires intelligent interconnection mechanisms. On the other hand, real-time personalized interactive multimedia NGN services require end-to-end quality assurance regardless of the traversed domains. Meeting these two requirements is a complex task and involves careful quality-related planning in each separate domain and coordination of these on a service-aware end-to-end basis. In Chapter X, Zapater and Bressan discuss the quality assurance of multimedia services over IP networks from the end user standpoint, and describe the concept of quality of experience (QoE). The focus is on video services that can be considered a significant evolution of services providers’ portfolio. Traditional quality management approaches adopted by service providers are mostly focused on the network perspective rather than the user perspective. The authors present quality requirements for video and TV services, and performance measures that focus on the quality perceived by the end user. This QoE approach is broader than one based on quality of service (QoS), since it takes into account how well a service meets customer goals and expectations rather than focusing only on network performance. Chapter XI by Picovici and Nelson describes the latest work for measuring perceptual voice quality and their application to wireless networks. They provide a review of various subjective testing meth-

xxiii

odologies and objective voice quality measures describing the target application of this metric and the performance limitations. They present the three main categories of objective voice quality measures, signal-based models, network-based (planning) models and both intrusive and single-ended monitoring based models. They have a devised a technique based on call history that can be used to predict end-user QoS for VoIP calls. In Chapter XII, Serif and Ghinea investigate user experiences of accessing streamed multimedia content. Through the creation of pre-defined QoS transmission profiles, the end-user experience is enhanced. These pre-defined profiles have built-in perceptual information and are based on both static (such as device type, CPU speed, and display specifications) and dynamic parameters (such as streamed content type location of the device/user, context of the device/user). From their work, end-user perceived QoS could be maintained while requiring fewer network resources. Chapter XIII by Koumaras et al. treats perceptual QoS (PQoS) assessment methods for multimedia applications and services. In this chapter, they describe subjective quality assessment methodologies and review the latest PQoS assessment models for multimedia services. Through their work, they present novel PQoS prediction models for Web, video, VoIP, and online gaming services. Their PQoS prediction model can be used to for the development and monitoring of PQoS-aware multimedia devices and networks for live multimedia services.

cross-Layered solutions Emerging wideband code division multiple access (WCDMA) data services will probably require resource allocation to ensure that throughput targets are met, and may employ scheduling and access control to achieve this. In Chapter XIV, Zhang and Jordan introduce a two-layer scheduler and connection access controller that attempts to balance efficiency with fairness. Their scheduler takes advantage of variations in the wireless channel, and they propose an algorithm that offers targeted throughput for interactive nomadic data streams, by integrating connection access control and resource allocation per connection request with rate scheduling on a per frame basis adaptive to slow fading. Upon the request of a data stream connection, a target throughput is negotiated between the user and the network/base station. The network attempts to achieve the throughput targets over the duration of each individual connection by maximizing a system objective based on users’ satisfaction as represented by a utility function. There is an increasing need for broadband communications anytime, anywhere for users that expect to receive multimedia services with support of quality of service. In Chapter XV, Chini et al. describe the satellite option to bridge the digital divide in those areas where terrestrial solutions are infeasible or too expensive. They provide a survey of the ETSI standardization framework for satellite networks, and describe resource management schemes for both the forward and return link. Finally they present a case study on the integration of a DVB-S/DVB-RCS satellite system interconnected with a WiFi segment for local coverage. Hybrid schemes make use the QoS-enabling features at various layers of the wireless multimedia service chain and optimise the QoS of the service using a cross-layer approach. Chapter XVI by Papadimitriou and Tsaoussidis presents a network and end-user QoS cross-layer approach for multimedia streaming services. In this chapter, they assess the efficiency of transport-layer solutions for multimedia traffic in heterogeneous networks. They compare the multimedia application requirements against the QoS features provided by the underlying network and present methods to measure the perceptual QoS assessment for the voice and video streams. In Chapter XVII, Bejaoui and Nasser provide a cross layer design for resource allocation over multimedia wireless networks. In this chapter they show how inter-layer dependencies can bring QoS

xxiv

improvements and the performance gains for real-time and non-real-time applications. This chapter concentrates on the packet scheduling and admission control schemes proposed for QoS provisioning for multimedia services over next generation wireless networks. In particular this chapter focuses on the radio channel conditions and explore the novel approaches based on cross-layered radio resource management protocol. The reliable transmission of multimedia services over bandwidth constrained error prone wireless networks is critical. In Chapter XVIII, Gur et al. describe the fundamental issues at the transport layer affecting the provision of QoS for wireless multimedia applications. They describe how the traditional transport layer protocols must be adapted to meet the challenges of delivering multimedia applications over best-effort wireless networks. They describe the latest trends and cross-layer solutions that rely on interaction between the different protocol layers. They show that cross layer protocol solution presents many challenges since such cross-layer schemes increase the inter-layer dependencies calling for a more complex protocol design, and present stability issues.

summary QoS is an important factor in the design and development of wireless multimedia applications and services. Different multimedia applications have very diverse QoS requirements in terms of bit rates, delay constraints, and loss tolerances. In a wireless environment, users are mobile and move between wireless technologies where the available resources are scarce and dynamically change over time. To complicate matters further, there is dramatic heterogeneity among end user devices in terms of latency, video visual quality, processing capabilities, power, and bandwidth. Providing QoS with both network and device heterogeneity to achieve efficiency in network bandwidth is a significant challenge. In this book we shall look at the major issues and challenges surrounding the provision of QoS for multimedia applications over wireless networks. Nikki Cranley and Liam Murphy, editors Dublin, January 2008

Section I

Network Quality of Service

In this section Network Layer QoS solutions are presented. Critical factors affecting QoS for real-time multimedia streaming applications include reliability, efficient delivery, and end-to-end latencies. The tremendous growth and development of wireless networking technology has brought about fresh challenges in the provision of QoS for such real-time multimedia applications. This section focuses on key wireless networking technologies and presents novel solutions that have been designed to address these challenges. State of the art wireless technologies presented in this section include UMTS, HSDPA-Enhanced UMTS, 4G, and WLAN. Each of these wireless networking technologies has differing mechanisms for QoS support. Resource management schemes and admission control schemes can be used to prevent the network becoming overloaded such that poor network performance begins to negatively affect multimedia applications.



Chapter I

Evaluating QoS in a Multi-Access Wireless Network Ronan Skehill University of Limerick, Ireland

Michael Barry University of Limerick, Ireland

William Kent University of Limerick, Ireland

Sean McGrath University of Limerick, Ireland

Dorel Picovici University of Limerick, Ireland

abstract This chapter introduces quality of service in multi-access wireless networks. Specifically it demonstrates how QoS is implemented in IEEE 802.11 and UMTS. The chapter explains how these complementary technologies, when coupled together, provide a network that is greater than its individual parts. Combining these networks brings new network management challenges. To this end, the impact of joint admission control strategies on quality of service is evaluated. The evaluation is performed on an advanced test platform that replicates an integrated Release 4 UMTS network and standard IEEE 802.11b network. The results show that when UMTS and WLAN are tightly coupled, the end user enjoys a higher level of quality of service.

intrOductiOn Mobile network configurations are becoming increasingly complex. Wireless communication networks are migrating from a set of insular competitive technologies toward a heterogeneous or converged wireless access topology comprising

a diverse range of radio interfaces. Cells from different radio technologies overlap in the same area resulting in co-existing layers of access technology. In this complicated environment, a multi-mode mobile can connect to different cells and unless there is knowledge about each cell it is difficult to optimise network performance and to manage resources efficiently.

Copyright © 2009, IGI Global, distributing in print or electronic forms without written permission of IGI Global is prohibited.

Evaluating QoS in a Multi-Access Wireless Network

Multi access or heterogeneous networks provide additional capacity for data traffic with the potential for load balancing of other services when the network becomes congested. This can be achieved through targeted admission of certain classes of traffic (e.g., background/interactive) when the session starts, or by forcing the handover of selected traffic. There has been great interest recently in the convergence of distributed, mobile networks and more localised wireless access technologies. As illustrated in Figure 1, universal mobile telecommunications system (UMTS) and IEEE 802.11 Wireless LAN (WLAN) represent two wireless technologies that show great promise in terms of interoperability and integration. UMTS was built from the ground up to support high levels of quality of service (QoS) for packet-based services in addition to providing voice in a macro-cellular environment. IEEE 802.11 WLAN is a contention based wireless access technology designed to provide high data rates with a micro-cellular

footprint. The interworking approach taken in the ARES testbed embeds WLAN into the UMTS radio access network (RAN). Since foundation level QoS was not part of the initial design of WLAN, it can be viewed as a complementary source of bandwidth for non-realtime critical services such as Web and e-mail in a UMTS system. Diverting background/interactive traffic to WLAN allows UMTS to support a higher number of voice and video calls, reducing call blocking rates. It is crucial for the pooled radio resources to be managed effectively to assure seamless interworking. Intrinsic radio resource management (RRM) relates to the management of a network’s local radio resources whereas joint radio resource management (JRRM) is concerned with the management of the environment where the mobile is capable of being connected to different cells. The mobile is connected to the cell where there are the most resources or where the technology best suits the connection type. Joint radio resource management strategies identify common compo-

Figure 1. Multiple wireless technologies converged can provide improved QoS to the end user



Evaluating QoS in a Multi-Access Wireless Network

nents, algorithms and parameters from intrinsic RRM to orchestrate cross-system radio resources in a managed, coordinated manner. JRRM aims to support intelligent interworking between different access technologies by balancing load across several radio access technologies (RATs), implementing admission control strategies, joint scheduling and congestion control and essentially exploiting increased system capacity, and thus improving quality of service. Deploying RRM strategies to marshal radio resources in multi-system networks is only part of the framework to capitalise on increased bandwidth sources. Key to the success of converged wireless networks is their ability to maintain QoS across all user sessions. In the most colloquial sense, QoS is measured by the satisfaction of the end user and how they perceive the network to be operating (i.e., can a user’s call be connected, do user sessions terminate prematurely?). In a more technical sense QoS is the ability to maintain a consistent level of bandwidth, amply sufficient to provide an uninterrupted and stable service to the end user. In addition to the generation of metrics to measure QoS (e.g., call blocking rates, bit error rate, etc.) one of the mandates of the heterogeneous testbed is to provide a perceptible user experience with regard to network performance. This is achieved by using a real user terminal that experiences a populated UMTS-WLAN system. The ‘reference user’ terminal enables real time feedback on the QoS performance of user applications such as video conferencing, Web browsing, e-mail, and network gaming, simply by providing an interactive experience for a real end user. In this chapter, the benefits of coupling different radio technologies are presented. Preceding the benefits is a review of different coupling strategies namely: loose, tight and very tight (ETSI BRAN TR 101 957, 2001). Following the coupling strategies review, details on how quality of service is implemented in UMTS and WLAN networks is presented. The QoS section illustrates how the different radio characteristics can be used to

complement each other in a multi-system environment. A very tightly coupled architecture is used to create the UMTS-WLAN test platform, named ARES, to evaluate the QoS benefits. Furthermore, a description of ARES and how it is used to evaluate interactive traffic over UMTS and WLAN is outlined. The results section quantifies the benefits of admitting interactive users to WLAN.

coupling strategies for uMts and WLan Although a highly competitive industry, network operators now face the same common denominator; meeting the bandwidth requirements of global user base and providing services that match fixed line systems and satisfy or indeed exceed end user expectation. For wireless access providers then, sources of bandwidth spanning multiple technologies need to be consolidated into a system comprising multiple points of wireless access (Agusti, 2006; Niebert et al., 2004; Sachs, Wiemann, Lundsjo, & Magnusson, 2004). Often referred to as the beyond 3G vision (Romero, Sallent, Agusti, & Diaz-Guerra, 2005), this represents a shift from competition to cooperation, where wireless access systems work together to service end user needs. A logical option then for the next wave of communications networks is the cooperative coupling of legacy systems and next generation networks, providing wireless access on top of an underlying Internet protocol (IP) infrastructure. At the heart of this development will be inter-working strategies between existing wireless networks to unlock greater pools of bandwidth and realise a heterogeneous network landscape. In this scenario wireless connectivity is provided across multiple RATs. UMTS and WLAN represent two wireless technologies that are complementary in terms of inter-operability. UMTS operates in a macro cellular environment to support a high level of QoS services across a geographically large area. WLAN provides Internet access in smaller cells



Evaluating QoS in a Multi-Access Wireless Network

that can be embedded as pockets of high-speed access with UMTS cells thus providing a complementary source of bandwidth to UMTS users. A diversity of solutions has been put forward as to how and at what level different systems are integrated. Bodies such as the 3GPP have studied the integration of WLAN into their system, to allow WLAN users access to services within the UMTS network. Others have developed working implementations of heterogeneous systems; and have demonstrated the feasibility of handover between WLAN and cdma2000 and WLAN and general packet radio service (GPRS). Key to which network integration strategy should occur is whether the systems can be integrated within the current standards or whether modifications will be necessary.

Figure 2. Loose coupling



Loose coupling Loose coupling (Buddhikot et al., 2003; Jaseemuddin, 2003) is the first step towards UMTS-WLAN integration. As shown in Figure 2, loosely coupled mobile networks do not have a connection at the RAN layer. Instead, the networks are connected via the core network. In these scenarios, WLAN and cellular networks are two separate access networks. The WLAN access network is attached to the Internet backbone, and the cellular network into the cellular core network. A new component required for this coupling architecture is a WLAN gateway or a gateway GPRS serving node (GGSN) emulator. This particular option allows ownership of the WLAN network to be separate from the UMTS network. Under this type of network deployment the WLAN element has access to the authorisation, authentication

Evaluating QoS in a Multi-Access Wireless Network

and billing mechanisms of the UMTS network (Ahmavaara, Haverinen, & Pichna, 2003). This enables the potential for common billing (3GPP TR 22.934, 2002). The WLAN network may have access to the packet switched or even the circuit switched services offered by the UMTS core network. The main benefit of this network architecture is the reuse of core network elements-one core network and two radio networks. The radio networks are, however, quite separate and distinct. Handovers from one RAT to another can be very costly in terms of signalling. Mobility may be provided by layer 3 mechanisms such as Mobile-IP (Tsao & Lin, 2002). As such, this type of network deployment does not lend itself towards the development of transparent seamless connections across multiple access technologies.

tight coupling The rationale behind the tightly coupled architecture is to make the 802.11 network appear as another 3G access network, as shown in Figure 3.

Tight coupling integrates the new radio network subsystem (RNS) at the serving GPRS support node, reusing the Iu interface. By coupling at the SGSN, WLAN resources are managed separately from UMTS resources. As in the loosely coupled architecture the different networks share the same authentication, signalling, transport and billing infrastructures, independent of the physical layer protocols. This coupling architecture allows for vertical handovers, which allow the application services to transfer connections between the different networks and access technologies. However, vertical handovers can be challenging to end-to-end transport protocols, as packets often get lost, delayed or reordered during a handover. Achieving seamless handover then is a difficult task. Furthermore, path characteristics such as bandwidth, latency and the buffer size can change instantly (Korhonen & Gurtov, 2004).

very tight coupling Very tight coupling (Cristache, David, & Hildebrand, 2003) involves establishing the WLAN

Figure 3. Tight coupling



Evaluating QoS in a Multi-Access Wireless Network

network as a second RNS, integrated at the radio network controller (RNC) as shown in Figure 4. The access point connects to the RNC through the Iub interface, presenting itself as a Node B. This mimics, quite closely, a hierarchical cellular layout. In this case, however, the micro cell RAT is of a different type to the macro cell. The benefit of this is that resource control for WLAN is colocated with the resource control for UMTS. In this way, it is possible to reuse the RRM located in the UTRAN and manage the WLAN HotSpot as a cell, or as part of a cell. The changes required in signalling are confined to the UTRAN. From a paging point of view the same RNC can retain control over locating the device in whatever RAT it is located. The underlying RAT is transparent to the core network (CN). Very tight coupling allows for the creation of joint radio resource management as information about the usage of both networks is available locally. Very tightly coupled networks embed microcell RATs in macro-cells. In terms of UMTS and WLAN a true heterogeneous architecture

Figure 4. Very tight coupling



can be realised using tight coupling. Using tight coupling resource control for both systems can be co-located enabling WLAN HotSpots to be managed as a cell or part of a cell in the UMTS system. This is the motivation behind adopting a very tight coupling approach in the ARES integrated testbed. Using very tight coupling enables the collection of information relating to air interface conditions and network capacity, providing a logical framework for JRRM and the best approach to maintaining awareness of RRM across multiple systems. The tight coupling framework is realised in the testbed by embedding a WLAN access point in the RAN.

trade-Off with coupling strategies Inevitably there is a trade-off with each of the coupling strategies when a network operator decides to integrate or couple networks together. There are several important factors, which must be considered. Table 1 illustrates some of the key trade-offs.

Evaluating QoS in a Multi-Access Wireless Network

In essence, with loose coupling there is no direct link between these two networks and they are completely disjoint in operation (both data and signalling paths). There is no change in the existing architectures for cellular and WLAN networks. The only interaction between the cellular and WLAN networks is at the billing system. With tight coupling there is a direct link between cellular and WLAN networks. The WLAN network appears to be one of the access networks from the cellular core network viewpoint. A WLAN gateway is required in the WLAN network to hide the WLAN functionality from the cellular core network. With very tight coupling there is a direct link between cellular and WLAN networks. The WLAN access point becomes one of the Node B from the cellular radio network controller viewpoint. In this case, enhancements

of cellular RNC are required such that the RNC manages the radio resource for both cellular Node Bs and WLAN access points. In terms of quality of service, tight and very tight coupling provide a means of reusing cellular QoS techniques.

QuaLity Of service in MuLtiaccess netWOrks The fundamental purpose of all communication networks is to provide a connection between end users. As networks have evolved, a multitude of services have been offered to users in the form of SMS, video conferencing, e-mail and Internet browsing. The success or failure of that network is dependent on how the user perceives performance of the service on offer. For a user to be satisfied

Table 1. Coupling trade-offs Loose coupling

Tight coupling

Security

Medium

High (reuse cellular security architecture)

Impact on existing cellular network architecture

None

High

Handoff speed

Slower

Faster

Session persistence

Yes

Yes

Service persistence (VOIP)

Possible

Yes

WLAN gateway required

Yes

Yes

WLAN traffic injected into cellular core network

No

Yes

QOS provisioning

Over-provisioning

Reuse cellular QOS architecture

Provisioning of the same administrative domain for WLAN and cellular

No

Possible

Mobile-initiated handoff or network initiated handoff

Mobile-initiated handoff is preferred

Both mechanisms are feasible

Standards development time

Short

Long



Evaluating QoS in a Multi-Access Wireless Network

with a network, a number of criteria must be met. The user should be able to connect easily, the call or service should last for the intended duration without prematurely ending and with respect to interactive data services (Web browsing), speed is a factor. QoS then is a measure of the perceived quality of the service being used and an indication of user satisfaction. Rather than being a single measure of a user’s satisfaction, QoS is the collective term for the quantifying of network performance from an end user point of view, and for the measures taken to ensure reliable services. To ensure that networks are performing at the desired level, one or more measurements of desired performance and priorities of a communications system are considered as part of QoS. QoS measures may include service availability, maximum bit error rate (BER), minimum committed bit rate (CBR) and other measurements that are used to ensure quality communications service. In heterogeneous environments (i.e., UMTSWLAN), the ability to offer end-to-end QoS while making the different RANs interoperate with each other is an important issue. QoS management mechanisms deal with such aspects. These mechanisms include call admission control (CAC), resource reservation, policing etc. The correspondence between the QoS offered by each sub-system of the end-to-end path is also required for a seamless service. In heterogeneous networks, end-to-end path may include different RAN environments (including multi-operator scenarios) as well as the Internet environment. Services must be apportioned to RATs that best support the service and can deliver on QoS guarantees to the end user. By examining the QoS models of each constituent RAT in a multi access network, a mapping of services to technology can be defined that maintains the QoS integrity of that service. UMTS was built from the ground up to provide a high level of QoS for the services it supports. This was done firstly by classifying the services on



offer in terms of performance related criteria, and then ensuring that the network performance was guaranteed on a per service basis. UMTS provides QoS service classifications that clearly define how each service is characterised in terms of QoS constraints and how bandwidth is provisioned to cater for that service. Time critical services such as voice and video conferencing can be delivered successfully over dedicated radio channels. By contrast, IEEE 802.11 Wireless LAN was not designed to support a high level of QoS. IEEE 802.11 WLAN was designed as a bridging technology to provide a wireless point of connection, over a short range, to a wired backbone. The design ethos did not encompass mechanisms to provide reliable delivery of realtime services such as Voice over IP (VoIP), video conferencing or multimedia streaming. The approach to delivering packets is known as best effort i.e. no guarantees of successful packet delivery. This makes WLAN suited to cater for less time critical services such as e-mail or Internet browsing, where QoS bounds are more relaxed than those of voice or video.

uMts supported services From the outset, UMTS was designed with a diverse service environment in mind, to realise a true mobile multimedia platform. Emphasis is placed on the delivery of these services with the specification of service classes. UMTS defines four service classes: • • • •

Conversational Streaming Interactive Background

The classes group services in terms of defined performance parameters such as delay and jitter, as illustrated in Table 2 (3GPP TS 23.107, 2003). The requirements of the QoS classes are met by negotiating appropriate QoS attribute values for each established or modified UMTS bearer.

Evaluating QoS in a Multi-Access Wireless Network

Table 2. UMTS traffic classes Traffic Class

Fundamental Characteristics

Examples of Service

Data Rates

Delay

Loss (FER)

Conversational

Preserve time variation between information entities of the stream. Conversational Pattern (stringent and low delay).

Voice

4 -25 Kbps

< 150 ms

70). Having as input the result of this comparison and the node’s current codec a new codec is chosen, using the following procedure (see also Algorithm 1): a.

In the case that the average value of the parameter is out of threshold then check its current value; if the current value is also out of threshold propose a codec of α steps lower in the codec ranking else propose a β steps lower codec. This way, even if the average performance of the call was not satisfactory during the fast RTCP monitoring time, if some other call has meanwhile changed codec and the system is beginning to recover (so the current value is above the threshold) then the call will suffer a smaller codec drop.

Adaptive Codec Selection for VoIP in Multi-Rate WLANs

b.

If on the other hand, both the average and the current value of the parameter are above the thresholds then there is no need to change codec.

This check is performed for each one of the three parameters (delay, loss, R) used in the evaluation and an average of the proposed codec of the three is chosen. Note that α > β, with α = 2 and β = 1 in the simulations, and that the codecs are ordered based on their bit rates, as shown in Table 1. Algorithm 1: Adaptation phase. while timer ≠ 0 do fastRTCP monitoring NowValue(param) ⇐ current values of delay, loss, R for param ⇐ delay, loss, R do AvgValue(param) ⇐ calculate average of parameter end for adaptTimer = adaptTimer - 1 end while for param ⇐ delay, loss, R do if AvgValue(param) > paramThreshold then if NowValue(param) > paramThreshold then change(param) ⇐ α else change(param) ⇐ β end if else change(param) ⇐ 0 {No drop} end if end for changeTotal ⇐ ∑(change(param))/3 newCodec ⇐ drop(currentCodec, changeTotal)

Recovery So far, the algorithm has analyzed the situation and taking into consideration all the feedback from the lower layers it has decided the most suitable codec to meet the needs of the current network

conditions. Here, at the recovery phase, is where the negotiations for the new codec agreement are performed at the application layer. This can be easily done using SIP, the signalling protocol for control of the call session parameters. More specifically, the SIP re-INVITE method is used, with a structure almost identical to the initial INVITE message and with only difference the new codec proposal in the SDP audio codec negotiation field. Hence, during the recovery phase the wireless node is asked to issue and send a SIP re-Invite message to the other end node, and re-negotiate through this the new codec. Depending on whether the other node accepts or not the new codec, the call continues normally or otherwise is dropped. In the case that the codec chosen as the most appropriate is lower than the lowest codec that a node can support, the easiest approach would be to drop the slow call in order for the others to continue with no problem. However, and depending on whether talking about a centralized or distributed implementation, there are other solutions in order to “save” the call. One idea, applying at the distributed implementation case, would be for the call to continue as it is during some stand-by time, without any codec change; If during this time some other node changes codec and the quality metrics show that the problem is solved then the call can continue successfully, otherwise the call will be then dropped. On the same line, for the centralized implementation, the AP could choose another call for codec adaptation if the call originally chosen cannot change any further. These adjustments can vary highly, as they depend on the specific needs of the implemented scenario in each case and on the trade off between capacity and quality/fast reaction and recovery. The details of the different variations of this solution are not considered here. After the negotiations are over, the algorithm returns to the adaptation phase and continues to monitor the system using the fast RTCP messages and evaluate its performance after the change; if



Adaptive Codec Selection for VoIP in Multi-Rate WLANs

Figure 10. Algorithm flow chart

the parameters are higher than the upper thresholds then it can return to the normal monitoring phase, else it needs to perform another codec change until reaching acceptable QoS levels.



Implementation Issues (Delay) An important parameter of any of the solutions focusing on VoIP traffic is the delay that they suppose and if this delay affects the call and is noticeable

Adaptive Codec Selection for VoIP in Multi-Rate WLANs

by the user. The delay during a call, translated into interruptions of the speech flow, should not be more than some hundred milliseconds so as not to be practically noticed by the user. When higher than this, the user will start to notice there is a communication problem. If this problem is solved fast enough from a human’s perspective (in less than a few seconds), then it can be considered as an incidental interruption and will not affect much the rest of the communication. Otherwise, the user will most probably end up hanging up and terminating the call. Therefore, the delay of the process mentioned above - and of any similar solution - is an important consideration when it comes to implementation. Although this delay is not as critical as in other environments, since the call is not interrupted during the process, it is essential to be able to recover from the network changes and their negative effects as fast as possible in order to avoid the user hanging up himself. The total delay, from the moment the algorithm receives the first alarm signals until the moment that the system recovers can be represented from the following equation: DelayTotal = Da + Db + Dc where Da is the duration of the adaptation phase, Db the time needed for the SIP re-Invite/OK/ACK messages for the codec renegotiation procedure and Dc the general processing time of the algorithm. While Db, being 1.5 times of the round trip time, depends on variable factors, like the internet delay when working on a wired to wireless scenario, it is usually in the order of milliseconds, as is the Dc. So both these delays are negligible in comparison with the duration of the adaptation phase, and in extension the RTCP interval delay, which is what really increases the total process delay to the order of seconds. Remember that the adaptation random timer is also set depending on the frequency of RTCP packets arrival, which was

the reason for choosing the fast RTCP extension at the first place. With the adaptation timer normally set to a value between 3 and 5 seconds, the delay should not be more than a few seconds (e.g., less than 5), which is acceptable from the point of view of human perception and taking into account that during this time the call is not dropped. From the simulations below and choosing a frequency of 1 second for RTCP interval, it can be seen that the delay is in fact not more than 3-4 seconds. Clearly, this delay can be minimized even more choosing a lower RTCP interval.

distributed vs. centralized architecture One of the main advantages of the algorithm previously presented is its flexibility when it comes to implementation. Since it is entirely based on the feedback from packets already existing and circulating in the network, like RTCP reports and MAC layer information, it does not need any specific modifications on the MAC layer, the Access Point or any of the nodes involved in the cellular architecture. In addition to this, it can be implemented both in a distributed and in a centralized mode, with minimum changes between the two versions and permitting higher flexibility depending on the specific characteristics of the working scenario. The basic difference between the two implementations is the location of the core adaptation algorithm. In the distributed scenario, the algorithm is implemented on each node and each node is made responsible for monitoring and adapting its own state. The node, based on the information of its rate changes (from MAC) and its QoS feedback (from RTCP), should determine whether or not to change codec. On the contrary, in the centralized case, the AP is in charge of monitoring all calls, including the transmission rate and the codec used by each one. When a call passes from fast to slow then the AP must determine



Adaptive Codec Selection for VoIP in Multi-Rate WLANs

if there is a need for a codec change, how many calls must change codec and which ones of them in particular must change, so as to reach network stability again, based on the RTCP information exchange between the clients. Therefore, the complexity and processing work is higher in the centralized version since the AP has to intercept all the RTCP packets in their way from one end to the other and calculate the parameters needed for the threshold comparison for all calls. Then the AP chooses the calls with the worst performance and decides according to the adaptation procedure, which calls to change and to which codec, giving more weight and priority in changing the slow calls first. The next step is to inform the nodes that there is the need to change codec, which includes suggesting to them to issue a SIP re-invite message for re-negotiating the codec with the other end. This is not a trivial process and is certainly more complex than in the distributed version, although some ideas on the interfacing issues of the AP with the nodes are given in section 3. However, the centralized view of the problem as a whole gives much better results and more possibilities of achieving an optimal codec combination among the nodes. As it has been proved in the authors’ own work (Sfairopoulou, Macián, & Bellalta, 2006), there is no need for all calls to change codec at the same time and the results are better when changing slow-rate calls than when changing fast-rate ones. This is due to slow-calls being the ones actually causing the problem, as seen in the problem statement. This priority on slow over fast calls is not possible in the distributed implementation, where each node will decide the action to be taken depending on its own limited view of the system, the call to detect first the QoS decrease will be the first to react, without knowing if there are other calls in the cell and the state (codec/rate) of each one. Additionally, since the control of the adaptation phase timer is not centralized, more nodes can coincide and change simultaneously codec, while



in the centralized implementation the AP can be set to wait during a random time between each codec change, which permits that less number of calls will have to change. But on the other hand, while the distributed approach may not be the globally optimal solution is easy to implement and it distributes the processing load of the algorithm. Simulation results show that there is an improvement in the performance of the algorithm when used in its centralized version; less calls are changing codec, the packet loss percentage is almost zero and the overall MOS achieved is higher than in the distributed implementation. These results will be reviewed in the following section.

Performance results In order to test the performance of the codec adaptation solution explained above, extensive simulations was performed using the network simulator tool NS-2 (NS2, 2005). The description of the testing scenario as also the performance results are provided next.

Scenario Description The results presented next are obtained using a hot-spot multi-rate scenario (Figure 11), with the network composed by one 802.11e (Std 802.11e, 2005) basic service set including N=9 wireless nodes and one Access Point connected to the wired network, acting also as a Proxy Server. A total number of nine bi-directional calls, established between one wired and one wireless client, are considered active during the simulations. The nodes start with a date rate of 11Mbps (fast-rate calls) and at predefined instants some flows change to 1Mbps data rate (slow-rate calls). It is assumed that all calls start with the G.711 codec, have the same duration, and change when needed to one of the lower bitrate codecs seen in Table 1.

Adaptive Codec Selection for VoIP in Multi-Rate WLANs

Table 5. System parameters of the IEEE 802.11b specification paRaMeteR

ValUe

paRaMeteR

ValUe

Rdata

{11, 5.5, 2, 1} Mbps

Rbasic

{11, 5.5, 2, 1} Mbps

Rphy

{1} Mbps

DIFS

50 μs

CWmin

32

SIFS

10 μs

CWmax

1024

SLOT (σ)

20 μs

m

5

EIFS

364 μs

ACK

112 bits @ Rbasic

RTS

160 bits @ Rbasic

CTS

112 bits @ Rbasic

MAC payload

[0, 18496] bits @ Rbasic

-

MAC header

240 bits @ Rdata

MAC FCS

32 bits @ Rdata

PLCP preamble

144 bits @ Rphy

PLCP header

48 bits @ Rphy

Retry Limit (R)

RS = 4, RL = 7

K (Queue length)

20 packets

Figure 11. Multi-rate WLAN scenario The AP coverage area A and radius R is composed of two regions: FAST and SLOW areas. The FAST area, AF is defined by a circle of radius R1 and the rest is defined as the SLOW area, AS (A=AF + AS). Mobile Nodes are uniformly distributed through the coverage area, so with probability PF=AF/A the STA is in the FAST region and with PS=1-PF the SLOW region.

Result Analysis

Monitoring frequency for the MAC monitor is f = 5 seconds, equal to the normal RTCP monitoring frequency. The fast RTCP transmission interval is one of the tunable parameters when using the extended RTCP version proposed by Ott et al. (2004) and is set here at δ = 1 seconds for a faster algorithm reaction time. It is assumed that all users support all codecs and there is no other traffic or other interferences in the wireless network. Each STA has a queue length of K = 50 packets. The values of the parameters set used can be found in Table 5.

As explained previously in section 1.3.2, in a scenario with 9 calls, all of them using G.711 codec, when 2 calls drop to 1 Mbps rate the new state is not feasible. This can be translated in voice quality metrics as high delays and high packet loss ratio. The effect can be observed in the example presented here. No Codec Adaptation Algorithm When there is no codec control algorithm implemented in the network, observe the effect when two nodes start transmitting at a lower rate changing from 11 Mbps to 1 Mbps (at instants t = 95 sec and t = 105 sec on Figures 12, 13, 14 and 15. As McGovern et al. (2006a) mention “the congestion in 802.11 is not gradual; the system has the a tendency to transition from an uncon-



Adaptive Codec Selection for VoIP in Multi-Rate WLANs

gested state delivering good performance to a congested state delivering very poor performance with the addition of little extra traffic” (p. 1). Due to this characteristic of the 802.11 networks, the observed packet loss percentage during the simulation increases almost instantly after the rate change happens to values reaching 90%. In fact, this result can be translated as a call drop since almost all packets are lost during a big part of the call. Moreover, the packet delay reaches very high values (of approx. 1 sec), as the queue

length of the AP becomes saturated (Figure 13). The congestion of the system, both in terms of loss and delay, is much more obvious in the AP, since it aggregates the traffic of all calls, which is the reason of the big difference observed on the results between uplink and downlink (as explained in section 1.3.1). In this case the AP acts as a bottleneck dropping queue packets and provoking a significant increase in packet loss ratio and delay. The same saturation can be also observed in the very low throughput obtained in Figure 15

Figure 12. Average aggregated packet loss percentage of VoIP flows in (a) Downlink (b) Uplink

Figure 13. Average aggregated delay of VoIP flows in (a) Downlink (b) Uplink

0

Adaptive Codec Selection for VoIP in Multi-Rate WLANs

and the low quality perceived by the user in terms of MOS in Figure 14. This MOS, as calculated in real-time using the E-model, drops to values as low as 1, meaning communication breakdown according to the MOS standard definition. The situation is corrected only when one of the two nodes that previously dropped to a lower rate changes again to a higher rate (11 Mbps) at simulation instant t = 145 sec. After this point, a decrease on delay and packet loss is observed, although they still remain higher than the desired for a correct VoIP transmission, with delay above

100ms and packet loss percentage of 10% in the downlink. Distributed Implementation of the Codec Adaptation Algorithm With the implementation of the codec adaptation algorithm in either of its two modalities (centralized and distributed), and since the codec of some of the calls is adjusted, the congestion level of the AP is significantly reduced, and as a result the effects of the multi-rate are barely noticed by the users. Looking at the distributed implementation

Figure 14. Average aggregated MOS obtained for VoIP flows in (a) Downlink (b) Uplink

Figure 15. Average aggregated throughput of VoIP flows in (a) Downlink (b) Uplink



Adaptive Codec Selection for VoIP in Multi-Rate WLANs

to begin with, almost instantly as the rate changes happen at t = 95 sec and t = 105 sec of the simulation, the algorithm takes action by first changing proactively the codec of the nodes that suffered the rate drop. The MAC information arrives as soon as the rate change takes place and the proactive codec change is very fast. However, this change is not enough and the RTCP packets announce that the QoS alarm situation continues. The nodes that receive this alarm (in fact this could be all nodes) enter the adaptation phase and after the adaptation timer expires they decide whether to change or not the codec they use. Indeed, the throughput values of the Figure 15, indicate that only some nodes had finally changed codec; the average aggregated throughput is about 40Kbps (with overheads included) which is higher than the result would be if all nodes had changed, according to the bitrates used by each codec on Table 1. This is because, as the codec change of other nodes lowers the congestion levels bit-by-bit, after the adaptation timer the node may encounter the problem already solved. So while the total throughput may be lower than before the rate changes, since some calls now use codecs that require less bandwidth, the system is no longer saturated. This can be verified in the packet loss and delay figures, where there is just one peak of high loss percentage reaching 20% and high delay of around 200 ms at the moment of the transmission rate changes and are corrected efficiently in less than 4 sec. The effect of the codec change observed in the packet loss ratio (Figure 12) agrees with the one expected; When using a lower bitrate codec, the offered load on the queue decreases and therefore less packet losses due to buffer overflow in the Access Point are observed. The results of MOS come to justify that the user perceived quality is maintained at very high levels, with only an instant drop at the moment of the rate change and until the nodes start reacting.



Centralized Implementation of the Codec Adaptation Algorithm Even more impressive are the results of the centralized implementation. At the moment, the MAC monitoring receives the rate change signal, it lowers by one the codec of the affected nodes. Along with this proactive codec change, only one more codec change of a fast node’s call was in fact needed during simulation, provoked from the RTCP monitoring, and the system recovers without noticing any of the negative effects mentioned above. Again, as it can be observed from the packet loss and delay figures, the peaks of high loss percentage and delay at the moment of the transmission rate change are corrected very fast. During the rest of the time packet loss is practically 0 and delay is no more than a few milliseconds. The average MOS value, indicating the user perceived quality is maintained in very high values around 4.3, as can be seen in Figure 14, with only an instant drop at the moment of the rate change and until the nodes start reacting. This shows a huge gain compared to the MOS with value less than 1.5 achieved when no algorithm is present. By having overall control on the timing and the order of the codec changes and prioritizing more efficiently the change of the slow calls before the fast ones, it results that the total number of calls that need to change codec is lower than on the distributed implementation. Observe the total throughput obtained (Figure 15) that is higher than in the distributed mode, which is translated in more calls transmitting with higher bit rate codec. This is because since the slow calls are the ones blocking the others, it is more efficient to lower more the codec of these calls, apart from also being the most fair solution. Again, both delay and packet loss results adjust to the expected performance as in the distributed implementation and even slightly better. To conclude, the performance of both the previous implementations, centralized and distributed, is satisfactory: no calls are dropped,

Adaptive Codec Selection for VoIP in Multi-Rate WLANs

the reaction of the algorithm and correction of the quality problems is quite fast and the packet loss is minimized which leads to high average MOS values. The difference in the performance between the two implementations is as expected, with the centralized giving better results since the Access Point has an overall control of the nodes and provides a more efficient combination of codecs. However, this implies higher processing effort for the Access Point and since the results of the distributed method are quite satisfactory, it can provide an easier and also effective alternative to the centralized version. The results presented in this chapter were compared against an analytical model presented by Bellalta et al. (2005). The system behaviour was shown to match quite precisely the expected results, as foreseen by the analytical model.

OPen issues and future guideLines This chapter has presented the impact of the multi-rate effect on VoIP communications in an 802.11 environment. Several proposed solutions have also been pointed out, namely the use of adaptive codecs, variations on the Link Adaptation mechanism and a cross-layer algorithm using MAC, SIP and RTCP information for the re-negotiation of codecs. Along the line of this last option, a number of further enhancements can be drafted. On the one hand, an unaddressed issue regards the communication between the Access Point and/or SIP Proxy and the Mobile Station in order to interchange information about the cell state. Another open issue is which exact functionality should the Access Point take: Should every AP implement a mini-SIP proxy, in order to implement the algorithm and its associated workload? Should it intercept all traversing RTCP and SIP packets, in order to gain detailed information about current VoIP sessions, especially in the centralized case?

No definitive answers have been provided so far, but some directions can be pointed out. First, as Camarillo, Kauppinen, Kuparinen, Ivars, and Res (2007) state, a possible mechanism for the interchange of information regarding codecs to be used for better overall session quality in the cell could be the SUBSCRIBE/NOTIFY mechanism of SIP (Roach, 2001). Although Camarillo et al. (2007) restricts its use to IMS, the basic idea is easily translated to the multi-rate scenario. Furthermore, by means of the same transactions the actual cell-wide quality could be communicated, in order for the STAs to take proactive actions in the distributed algorithm, if need be. In principle, every AP needs not be a SIP proxy, if the algorithm is distributed and the existing SIP proxies would be multi-rate aware. Otherwise, it could certainly prove very advantageous, indeed necessary, that either every AP is co-located with a SIP proxy, or that a communication protocol exists between the two. This second options seems less feasible, for it involves strong modifications in existing proxy architectures and functionalities. The first option, however, only involves a further modification of the AP software, which must be adapted in any case to make it multi-rate aware. The AP/SIP proxy certainly needs to be made aware of the session state in order for the algorithm to work. Hence, it must intercept all RTCP and SIP messages, or be made aware of their content (e.g., by means of SUBSCRIBE/NOTIFY messages). An alternative avenue for further work is to consider if private extensions to the RTCP protocol, foreseen in the standard, would be useful. For example, RTCP provides application-specific reports that can be freely defined to convey extra information, relevant for a certain application only. It could be considered if certain real-time applications running on multi-rate environments, like for example online gaming, may take advantage of this to enhance their overall user experience. However, one of the paramount characteristics of the algorithm presented in this chapter lies



Adaptive Codec Selection for VoIP in Multi-Rate WLANs

precisely in that it does not need any protocol extensions or modifications at any layer in order to work properly. This represents a huge practical advantage to achieve fast and effective deployment, as well as correct interworking with legacy equipment that is not multi-rate aware.

areas and how they can complement each other. As a matter of fact, any action taken by the algorithm could be communicated to the CAC so that it can immediately adapt to the new network conditions. However, the exact mechanism for the co-operation between the two modules are still to be determined.

cOncLusiOn references While VoIP will continue to grow in popularity and use, future wireless networks will have to come up with more efficient ways to handle traffic heterogeneity. Current WLANs introduce several performance impairments, which reduce the efficient usage of the wireless bandwidth. In this chapter, an extended analysis was performed on the problems encountered in VoIP flows due to the specific characteristics of the 802.11 WLANs, with special emphasis on the Link Adaptation mechanism and the multi-rate effect on the voice flows. The main causes of the problem as also various proposed solutions were examined and a codec adaptation algorithm was presented as an example implementation in more detail. Having a first approach on how such a solution could work with the simulation results presented in section 2.5, such a deployment can be highly beneficial for voice traffic over multi-rate WLANs, with relative small additional cost. Further testing of this kind of solutions as also some experimental testbed and/or an enhanced AP implementation including the modifications mentioned in section 3 is highly recommended in order to have more sustainable results on the viability of these proposals and the other implementation issues that arise from them. Adapting the parameters of the VoIP flows is the most discussed solution for coping with the multi-rate effects. The co-operation of such an algorithm with the Call Admission Control (CAC) mechanisms of 802.11 networks is the next important step from there. It has been pointed out how these two algorithms work on different



Bellalta, B., & Meo, M. (2006). Call admission control in WLANs. Resource, Mobility and Security Management in Wireless Networks and Mobile Communications. Auerbach Publications, CRC Press. Bellalta, B., Macian, C., Sfairopoulou, A., & Cano, C. (2007a). Evaluation of joint admission control and VoIP codec selection policies in generic multirate wireless networks. ITC/IEEE NEW2AN’07, St. Pertersburg, Russia. Bellalta, B., Meo, M., & Oliver, M. (2007b). VoIP call admission control in WLANs in presence of elastic traffic. IEEE Journal of Communications Software and Systems, 2(4). Bellalta, B., Oliver, M., Meo, M., & Guerrero, M. (2005). A simple model of the IEEE 802.11 MAC protocol with heterogeneous traffic flows. In IEEE Eurocon 2005, Belgrade, Serbia, and Montenegro. Bianchi, G. (2000). Performance analysis of the IEEE 802.11 distributed coordination function. IEEE Journal on Selected Areas in Communications, 18(3). Cai, L., Xiao, Y., Shen, X., & Mark, J. (2006). VoIP over WLAN: Voice capacity, admission control, QoS, and MAC. International. Journal of Communication Systems, 19, 491-508. Camarillo, G., & García-Martn, M. (2004). The 3G IP multimedia subsystem (IMS): Merging the Internet and the cellular worlds. John Wiley and Sons.

Adaptive Codec Selection for VoIP in Multi-Rate WLANs

Camarillo, G., Kauppinen, T., Kuparinen, M., Ivars, I., & Res, E. (2007). Towards an innovation oriented IP multimedia subsystem [IP multimedia systems (IMS) infrastructure and services]. IEEE Communications Magazine, 45(3), 130-136. Cano, C., Bellalta, B., & Oliver, M. (2007). A simple model of the IEEE 802.11 MAC protocol with heterogeneous traffic flows. In IEEE Personal, Indoor and Mobile Radio Communications (PIMRC’07), Athens, Greece. Chen, J., Lee, L., & Tseng, Y. (2006). Integrating SIP and IEEE 802.11e to support handoff and multi-grade QoS for VoIP applications. In Proceedings of the 2nd ACM International Workshop on Quality of Service & Security for Wireless and Mobile Networks (pp. 67-74). New York, NY, USA, ACM Press. Cisco Technical Document 7934. (2006). Voice over IP--Per call bandwidth consumption. Updated February 2006. Retrieved from http://www. cisco.com Cole, R., & Rosenbluth, J. (2001). Voice over IP performance monitoring. ACM SIGCOMM Computer Communication Review, 31(2), 9. ETSI. (1998). Digital cellular telecommunications system (Phase 2+): Adaptive multi-rate (AMR) speech transcoding (GSM 06.90). European Telecommunications Standards Institue (ETSI) G.107 Rec. (2000). The E-model, a computational model for use in transmission planning. International Telecomunications Union (ITU-T) Recommendation. G.113 Rec. (2001). Transmission impairments due to speech processing. International Teleccomunications Union (ITU-T) Recommendation. Gardner, M., Frost, V., & Petr, D. (2003). Using optimization to achieve efficient quality of service in voice over IP networks. IEEE Performance, Computing, and Communications Conference Proceedings (pp. 475-480).

Garg, S., & Kappes, M. (2003). Can I add a VoIP call? IEEE International Conference on Communications (ICC’03), Anchorage, Alaska, USA. H.323 Rec. (1998). H. 323, packet-based multimedia communications systems. International Teleccomunications Union (ITU-T) Recommendation. Haratcherev, I., Taal, J., Langendoen, K., Lagendijk, R., & Sips, H. (2005). Automatic IEEE 802.11 rate control for streaming applications. Wireless Communications and Mobile Computing, 5(4), 421-437. Hedge, N., Proutiére, A., & Roberts, J. (2005). Evaluating the voice capacity of 802.11 WLAN under distributed control. In IEEE LAN/MAN Conference, Megalo Arsenali, Chania, Greece. Heusse, M., Rousseau, F., Berger-Sabbatel, G., & Duda, A. (2003). Performance anomaly of 802.11b. In IEEE INFOCOM, San Francisco, USA. Hole, D. P., & Tobagi, F. A. (2004). Capacity of an IEEE 802.11b Wireless LAN supporting VoIP. In IEEE International Conference on Communications (ICC’04), Paris, France. Huang, C., Chindapol, A., Ritcey, J., & Hwang, J. (2006). Link layer packet loss classification for link adaptation. In WLAN. 2006 40th Annual Conference on Information Sciences and Systems (pp. 603-608). Kim, J. O., Tode, H., & Murakami, K. (2005). Service-based rate adaptation architecture for IEEE 802.11 e QoS networks. In IEEE Global Telecommunications Conference (GLOBECOM’05). Liu, J. S., & Lin, C. H. R. (2005). A relay-based MAC protocol for multi-rate and multi-range infrastructure wireless LANs. Wireless Personal Communications, 34, 7-28. Lundberg, T., de Bruin, P., Bruhn, S., Hakansson, S., & Craig, S. (2005). Adaptive thresholds for



Adaptive Codec Selection for VoIP in Multi-Rate WLANs

AMR codec mode selection. In IEEE 61st Vehicular Technology Conference (VTC 2005-Spring).

M., & Schooler, E. (1998). SIP: Session Initiation Protocol.

McGovern, P., Murphy, S., & Murphy, L. (2006a). Addressing the Link adaptation problem for VoWLAN using codec adaptation. In IEEE Global Telecommunications Conference (GLOBECOM’06).

Schulzrinne, H., Casner, S., Frederick, R., & Jacobson, V. (2003). RFC3550: RTP: A Transport Protocol for Real-Time Applications. Internet RFCs.

McGovern, P., Murphy, S., & Murphy, L. (2006b). Protection against link adaptation for VoWLAN. In Proceedings of the 15th IST Mobile and Wireless Communications Summit. NS2. (2005). Network Simulator, release 2.28. Retrieved from http://www.isi.edu/nsnam/ns/ Ott, J., Wenger, S., Sato, N., Burmeister, C., & Rey, J. (2004). Extended RTP profile for RTCP-based feedback (RTP/AVPF). Internet Engineering Task Force. Technical report, Internet Draft. P800 Rec. (1996). P. 800: Methods for subjective determination of transmission quality. International Telecommunications Union (ITU-T) Recommendation. Qiao, Z., Sun, L., Heilemann, N., & Ifeachor, E. (2004). A new method for VoIP quality of service control use combined adaptive sender rate and priority marking. In IEEE International Conference on Communications. Roach, A. (2001). SIP-specific event notification. draft-sip-events-00 (work in progress), July. Rosenberg, J., Schulzrinne, H., Camarillo, G., Johnston, A., Peterson, J., Sparks, R., Handley,



Servetti, A., & De Martin, J. (2003). Adaptive interactive speech transmission over 802.11 wireless LANs. In Proceedings of the IEEE International Workshop on DSP in mobile and Vehicular Systems. Sfairopoulou, A., Macián, C., & Bellalta, B. (2006). QoS adaptation in SIP-based VoIP calls in multi-rate 802.11 environments. In ISWCS 2006, Valencia, Spain. Std. 802.11 (1999). Wireless LAN medium access control (MAC) and physical layer (PHY) specifications. IEEE Std 802.11. Std. 802.11e (2005). Wireless LAN medium access control (MAC) and physical layer (PHY) specifications; Amendment: Medium access control(MAC) quality of service enhancements. IEEE Std 802.11e. Trad, A., Ni, Q., & Afifi, H. (2004). Adaptive VoIP transmission over heterogeneous wired/wireless networks. Lecture Notes in Computer Science, 3311 (pp. 25–36). Springer. Wu, P., Tseng, Y., & Lee, H. (2005). Design of QoS and admission control for VoIP Services over IEEE 802.11e WLANs. In National Computer Symposium.



Chapter VI

Buffer Control Techniques for QoS Provisioning in Wireless Networks Michael M. Markou University of Cyprus, Cyprus Christos G. Panayiotou University of Cyprus, Cyprus

abstract This chapter introduces the network buffer control techniques as a mean to provide QoS. This problem has been extensively studied in the context of wirelined networks; however, the proliferation of wireless networks and the introduction of multimedia applications has significantly changed the characteristics of the traffic mix that flows on the network. The objective of this chapter is to create a new methodology for automatically adapting the various buffer thresholds such that the network exhibits optimal or near optimal performance even as network conditions change. The behavior of the network (generally a discrete event system—DES) is approximated by that of a stochastic fluid model (SFM); then using infinitesimal perturbation analysis (IPA) we obtain sensitivity estimators of the performance measure(s) of interest with respect to the control parameter. These estimators are easy to compute using data observed from the DES’s sample path. Finally, the computed estimators are used in stochastic approximation algorithms to adjust the thresholds.

intrOductiOn The emergence of advanced multimedia and other real-time applications has increased the demand for better than best effort services thus

increasing the pressure on network providers to provide quality of service (QoS) guarantees. As a result, providers need to find ways to configure the parameters of their networks such that the application requirements are met. Over the past

Copyright © 2009, IGI Global, distributing in print or electronic forms without written permission of IGI Global is prohibited.

Buffer Control Techniques for QoS Provisioning in Wireless Networks

few years, QoS provisioning has been an active area of research resulting in the standardization of several architectures and protocols. Most of the past research assumed wirelined networks and that the dominant network traffic is based on the TCP (transport control protocol). However, the situation with current networks is changing. Wireless networks are becoming more popular and there is an increase of the UDP (user datagram protocol) based traffic (e.g., real-time protocol (RTP) for voice over IP). Thus, there is a need for new protocols and architectures that adopt the characteristics of the wireless channels and of the new traffic mix to provide the required QoS guarantees. Management of different types of services such as video, voice, file transfer and email while provisioning at the same time the QoS level that each service demands is a challenging task; the analysis of large scale networks is excessively difficult and queuing theory is largely based in the Poisson assumption that does not capture the bursty nature of the realistic traffic. Furthermore, any proposed solution has to be scalable and easy to implement, adding the least possible overhead to the system’s operation. Integrated services (IntServ) is a proposed architecture for delivering QoS guarantees. In IntServ every application that requires some level of QoS guarantees has to make a resource reservation at each intermediate node along the path of the flow. The underlying protocol used for signaling to dynamically allocate resources in IntServ is the resource reservation protocol (RSVP), described by Braden et al. (1997). In this architecture, communication between a sender and a receiver is established only when every node (router) in the intermediate path between them has the necessary resources to support the QoS requirements of the new flow without affecting the QoS delivered to existing flows. A major drawback of this approach is that each router that supports a flow has to maintain information about it, making it difficult to keep track of all flows



when the network scales up. Furthermore, the overhead caused due to RSVP signaling reduces the utilization efficiency. Differentiated services architecture (DiffServ) solves the problem by providing a framework for classification of the traffic and differentiation between the levels of service that each class will receive. In DiffServ each data packet is classified as belonging to one of a finite number of traffic classes (Blake et al., 1998). Routers in the network treat each incoming packet according to its class, enabling the protection of higher priority traffic against the lower priority, providing a more efficient and scalable traffic management mechanism. The treatment of each packet is achieved by mapping its traffic class to a per-hop behavior (PHB), which defines how a packet will be forwarded. The four available standard PHBs are: default, class-selector (Nichols et al., 1998), assured forwarding (Heinanen, Baker et al. 1999), and expedited forwarding (Jacobson et al., 1999). However, DiffServ architecture has also some disadvantages. DiffServ mechanism cannot provide individual connection QoS guarantees. Moreover, there are no clear incentives for applications to voluntarily mark their packets with a priority other than the highest. Policing mechanisms that downgrade an application’s packets if it exceeds its allocated bandwidth exist however, for example, see (Heinanen & Guérin 1999; Heinanen & Guérin 1999). In the last years, some researchers tried to combine the advantages of the two architectures and achieve an improved mechanism. A framework to apply IntServ over DiffServ was proposed by Bernet et al. (2000), whereas Zhang and Mouftah (2001) developed a sender-initiated resource reservation mechanism over a DiffServ network to offer end-to-end QoS. In addition, other frameworks that are more appropriate for wireless ad-hoc networks have been proposed, for example, flexible QoS model for mobile ad hoc networks (FQMM). FQMM is a hybrid service model that takes advantage of the per-flow

Buffer Control Techniques for QoS Provisioning in Wireless Networks

granularity of IntServ and the services aggregation into a number of classes performed by DiffServ. Based on the assumption that the number of flows requiring per flow QoS guarantees is much smaller than the low priority flows, FQMM is designed to provide per flow QoS guarantees for the high priority flows while the lower priority flows are aggregated into a number of classes as in DiffServ. Although FQMM combines the advantages of both IntServ and DiffServ, it has also some unresolved issues such as the traffic classification policy, the amount of traffic that will be provided to each per flow service and the allotment of per flow or aggregated service for the given flow (Reddy et al., 2006). In this chapter, we consider data traffic that is categorized in different priorities according to its importance to the end user, in order to meet the predefined QoS requirements. In the case of IntServ data traffic can be a data flow whereas in DiffServ architecture a class may consists of all packets with the same PHB. The end user may be anyone that uses the network’s services at a higher network level (e.g., a client application that presents data sent from a video server or a taxi passenger that uses his or her cell phone to make a videocall while traveling on a highway). The importance of data traffic can be related directly to the application that generates it and thus the priority of each packet will be the same or it can vary between data generated from the same application. An example for the first case is voice over IP (VoIP) and file transfer applications; VoIP applications generate data with the same, strict requirements in delay whereas every FTP packet requires best effort with as low losses as possible. On the other hand, a single application may generate data with different levels of priority. For example, MPEG defines three different types of frames; the “I” frames carry more information than the “P” or “B” frames (Hoffman et al., 1998). Thus, a source may mark all packets that carry “I” frames with higher priority and the rational is that in case of congestion, a router may drop

packets that carry “P” or “B” frames increasing the probability that “I” frames will get through and thus the degradation of the quality of the received video will be minimized. Alternatively, the priorities may be set by a policing mechanism at the ingress gateway (Heinanen & Guérin 1999; Heinanen & Guérin 1999). Several techniques for providing statistical QoS guarantees to multimedia streams can be found in the literature, each one designed to be applied in a different level of the OSI structure. For example, some higher-level techniques reduce the quality of the transmitted multimedia content in order to minimize the amount of information sent and, as a result, their transmission rate. On the other hand, some network-level techniques implement packet admission control policies in the network’s buffers to control incoming traffic. The emphasis of this chapter is on the latter case (i.e., controlling the buffer of the intermediate nodes) even though similar concepts can be applied for controlling the transmission rate of the sender. This chapter presents several buffer control techniques that have been proposed for QoS provisioning in wirelined and wireless communications. Furthermore, the chapter presents a methodology for automatically adapting the various buffer thresholds in ways such that the network exhibits optimal or near optimal performance. The methodology adopted is similar to the one in Panayiotou et al. (2004) and references therein. Namely, the behavior of the network (generally a discrete event system (DES)) is approximated by that of a stochastic fluid model (SFM). Using the SFM one can use infinitesimal perturbation analysis (IPA) (Cassandras & Lafortune, 1999) and obtain sensitivity estimators of the performance measure of interest with respect to the control parameter of interest (e.g., the buffer thresholds). The simplicity of these estimators allows us to compute them using data observed from the sample path of the real system. Finally, at the end of every observation period, the computed sensitivities can be used together with stochastic approxima-



Buffer Control Techniques for QoS Provisioning in Wireless Networks

tion type algorithms to adjust the thresholds and achieve optimal performance even as network conditions change. This chapter extends the results of (Panayiotou et al., 2004) by allowing for packet expiration due to QoS delay violations as well as by allowing for user mobility (in the context of cellular networks). The chapter is organized as follows: The next section presents relevant work on techniques for buffer management for QoS, namely tail drop and threshold policies as well as active queue management. The subsequent section proposes a framework that utilizes stochastic fluid modeling and infinitesimal perturbation analysis (SFM/IPA framework) for developing buffer control algorithms in communication systems. Finally, the chapter concludes with an outlook of future trends.

backgrOund The degree of user satisfaction varies between the services that generate or receive the traffic and is reflected in their QoS requirements. Among the most common performance measures that need to be satisfied in a QoS agreement is the average packet delay, the delay jitter and the packet loss probability. A fundamental tradeoff in buffer management is the balance between packet delay and packet losses. On one hand, buffers are needed to limit packet losses and unnecessary retransmissions during bursts of traffic. On the other hand, large buffers may introduce excessive packet delays causing degradation of the provided QoS. Thus, providers need ways of managing their buffers to alleviate the effects of congestion and such that the best possible QoS is provided. Note that the buffer management problem is not static since it does depend on the current network conditions, which change over time. Thus dynamic mechanisms are in need that can adjust the buffer parameters such that the best possible QoS is always achieved. Due to the large size of the

0

current networks, the proposed solutions need to be scalable and distributed in nature. Most of the past research on buffer management assumed that the medium of communication is reliable and that the individual flows respond to congestion by reducing their transmission rate (TCP based). In current networks, several paths exist that consist of at least a wireless link where the probability of packet error is significant due to poor transmission, interference, weather phenomena etc. As a result, in wireless networks, a packet drop does not necessarily imply congestion, which is the case in wirelined networks. This fact is very important, since until recently, all proposed solutions for congestion control assumed that a packet loss directly implies congestion and thus these solutions could not be applied in wireless networks. Furthermore, until recently, the overwhelming majority of traffic was based on TCP, which includes control mechanisms that limit the transmission rates of sources when the network becomes congested. The emergence of multimedia applications that use UDP based protocols (such as real-time transport protocol (RTP) and real time streaming protocol (RTSP)) is changing the mix of traffic that flows in the network, thus in current and future networks, it is expected that a significant number of flows will not respond to network congestion. RTP (Schulzrinne et al., 2003) defines a standardized packet format for delivering multimedia content (audio and video) over the Internet whereas RTSP is used in streaming media systems enabling the client to control the traffic transmission procedure of the streaming media server (Schulzrinne et al., 1998). RSVP can be used in conjunction with RTP to enhance the provided service in multimedia applications.

buffer Management techniques A buffer management algorithm consists of two key mechanisms, the backlog controller that specifies when to drop a packet and the dropper that specifies which packet to drop.

Buffer Control Techniques for QoS Provisioning in Wireless Networks

The following paragraphs present some of the most popular queue (buffer) management and congestion control techniques. Most of these techniques that can be found in literature (e.g., (Labrador & Banerjee, 1999) discuss buffer management algorithms for IP and ATM networks) are similar to TCP in the sense that they use acknowledgments as a means to control the network traffic. The reason that the greatest portion of research work was targeted to improve TCP performance is because the overwhelming majority of the network traffic was TCP traffic. However, it is known that the share of multimedia traffic in the traffic mix grows continually, thus it is expected that both TCP and UDP traffic will coexists. Hence, more sophisticated strategies for buffer management are needed in order to improve the playback quality of the multimedia content, which is packetized in UDP-based packets.

Backlog Controllers Random Early Detection To overcome the reduced throughput problems that arise in TCP when network is highly loaded and improve system’s performance Floyd and Jacobson proposed Random Early Detection (RED) (1993), a form of active queue management (AQM). AQM is a proactive mechanism used to inform users in a network about incipient congestions enabling them to react and avoid it. The idea behind RED is to give indications about an oncoming congestion by probabilistically dropping (or marking) arriving packets. Users consider the packet loss (equivalently, the marked packet) as a result of congestion and reduce their sending rates; due to that, RED performs better with protocols that perceive packet losses as congestion indications (e.g., TCP). This technique can improve the end-to-end performance compared to other reactive mechanisms (which choose to drop packets only when congestion occurs) since many unnecessary packet drops are avoided. In addition, the synchronization problem is alleviated since

only a fraction of the flows drastically cut their sending rates. RED maintains an exponentially weighted moving average of the queue length: t +1 t Qavg = (1 − w) ⋅ Qavg + w ⋅ Qt

(1)

t

where Qavg is the weighted average of queue length at time instant t, Qt is the queue length at time t and w is a weight factor commonly set to 0.002 (Floyd, 1997). When Qavg exceeds a maximum threshold thmax all arriving packets are dropped. If Qavg is between a minimum thmin and a maximum threshold thmax , packets are randomly dropped with a probability p(Qavg) that increases linearly as a function of Qavg up to a maximum value pmax, as described in Eq. (2) and shown in Figure 1. pmin is usually set to zero. 0,   Q − th  avg min p (Qavg ) =  ,  thmax − thmin  1,

if 0 ≤ Qavg < thmin if thmin ≤ Qavg < thmax otherwise

(2) The decision which packet to drop is random in order to provide fairness between the connections which are candidates to slow down. The values of thmin, thmax, pmin, pmax and the weight w are parameters set by the network administrators and can be tuned as to avert short-lived conges-

Figure 1. RED drop function



Buffer Control Techniques for QoS Provisioning in Wireless Networks

tion to exist, maintaining high throughput and low delay. However, tuning RED’s parameters is not easily achievable; actually, the difficulty of its parameter tuning is one of the most important drawbacks of RED (Alemu & Jean-Marie 2004; Bonald et al., 2000; Christiansen et al., 2001). The most important factor that adds extra complication to the tuning problem is the dynamic nature of networks which may cause continuous changes of the optimal operating point for a RED enabled gateway, thus generating the need for dynamic tuning of RED parameters ( Alemu et al., 2004; Chrysostomou et al., 2003; Hollot et al., 2002; Misra et al., 2000; Sirisena et al., 2002; Verma et al., 2003). Furthermore, although for long lived sessions RED seems to improve the system performance, experiments with short lived traffic such as Web traffic, don’t show any clear benefits for using RED over a reactive mechanism and more specifically, tail drop (Christiansen et al., 2001). Other AQM techniques include explicit congestion notification (ECN), used to prevent congestion and improve TCP’s performance and a version of ECN that supports differentiated services (Floyd 1994; Floyd et al., 1993), random early marking (REM) that uses optimization techniques to maximize a utility function subject to the constraint that the output link has finite capacity (Athuraliya et al., 2000; Lapsley & Low 1999;) and BLUE that uses packet loss and link utilization history to manage congestion (Feng et al., 2002). RED with In/Out bit (RIO) is an approach that incorporates mechanisms for service differentiation in multiclass traffic. RIO extends RED considering two classes of traffic: assured (high priority) and best effort (low priority) and uses two parameter sets, one for each traffic in in and class: in-profile with parameters thmax, thmin out in p max and out-of-profile with parameters thmax , out out thmin and p max. It is usually more aggressive in dropping out-of-profile packets setting parameter in out in out > thmax , thmax and > thmin values such that thmin out in p max > p max . RIO also maintains the values of the



average queue length of in-profile packets (Qavg) and total packets (Qavg) and controls the packets rejection accordingly (Clark & Fang, 1998). As with RED, the out-of-profile thresholds can be dynamically adapted using estimates of the packet arrival rates (Chien & Liao, 2003). in

RED in Wireless Networks RED performance in wireless data networks is ineffective under self-similar traffic. As shown in a simulation study, self-similarity affects negatively the performance of the network. It is believed that this ineffectiveness comes from the channel errors and network’s varying resources (Gao et al., 2002). Goodput and delay performance of TCP over 3G networks can be improved if slope based discard (SBD) algorithm is used. SBD is an AQM mechanism with easier than RED implementation and configuration (Alcaraz & Cerdan, 2006). Multiclass RED in wireless networks uses traffic classification, RED and weighted first in first out (FIFO) scheduling, in which the scheduling rate is proportional to pmax. This scheme is shown to offer low loss to loss-sensitive applications and low delay to delay-sensitive applications (GyasiAgyei, 2002). Threshold Policies In this approach, the buffer space is partitioned in N partitions based on B1