McGRAW-HILL PROFESSIONAL ENGINEERING Pf Fundamentals of Mobile and Pervasive ee v Embedded wireless sensor networking V Wireless security a ws neon V Ubiquitous computing yer ae Ca Penta 4 v Middleware Frank Adelstein Sandeep K.S. Gupta Golden G. Richard III man Loren SchwiebertInformation contained in this work has been obtained by Tata McGraw-Hill, from sources believed to be ‘reliable. However, neither Tata McGraw-Hill nor its authors guarantee the accuracy or completeness of any information published herein, and neither Tata McGraw-Hill nor its authors shall be responsible for any errors. omissions, or damages arising out of use of this information. This work is published with the understanding that Tata McGraw-Hill and its authors are suet 2 information but are not attempting to render engineering or other . If such services are required, the assistance of an | appropriate professional should be sought. WA INS G1 Tata McGraw-Hill Copyright © 2005 by The McGraw-Hill Companies, Inc. All rights reserved. No part of this publication may be reproduced or distributed in any form or by any means, or stored in a database or retrieval system, without the prior written permission of the publisher Tata McGraw-Hill Edition 2005 Fourth reprint 2008 RLXCRRYXRCYXC Reprinted in India by arrangement with The McGraw-Hill Companies, Inc., New York Sales territories: India, Pakistan, Nepal, Bangladesh, Sri Lanka and Bhutan CIP Data is on file with the Library of Congress ISBN-13: 978-0-07-060364-6 ISBN-10: 0-07-060364-2 Published by Tata McGraw-Hill Publishing Company Limited, 7 West Patel Nagar, New Delhi 110 008, and printed at Gopaljee Enterprises, Delhi 110 053Contents, Preface xv Acknowledgments xix Chapter 1, Mobile Adaptive Computing 1.1_What Is Mobile Computing? 1.2 _Adaptability—The Key to Mobile Computing 1.2.1 Transparency 1.22 Constraints of mobile computing environments. 1.2.3 Application-aware adaptation - 1.3 Mechanisms for Adaptation 1.3.1 Adapting functionality 1.32 Adapting data 1.4 How to Develop or Incorporate Adaptations in Applications 1.4.1 Where can adaptations be performed? 12 1.5 Support for Building Adaptive Mobile Applications 1.5.1 Odyssey 1.5.2 Rover 1.6_Summary I= [eo loo leo lovlonlanleo lo low RE Risis Chapter 2. Mobility Management 2.1 Mobility Management 2.2 Location Management Principles and Techniques 2.2.4 Registration area—based location management 2.3 Location Management Case Studies 2.3.1_ PCS location management scheme. 2.32 Mobile IP 24 Summary Bssisisish & len Io Chapter 3. Data Dissemination and Management 3.1_Challenges 3. ta mi 3.2.1 Bandwidth allocation for publishing 3.2.2 Broadcast disk scheduling Ree isviii Contents 3.3. Mobile Data Caching or 3.3.1 Caching in traditional distributed systems 68 3.3.2 Cache consistency maintenance 69 3.4 Mobile Cache Maintenance Schemes 72 3.4.1 A taxonomy of cache maintenance schemes _ 72 3.42 Cache maintenance for push-based information dissemination 74 3.4.3 Broadcasting invalidation reports 75 Disconnected operation 77 isynchronous stateful (AS) scheme 78 prone : 2 _Mobile Web Caching 86 3.5.1 Handling disconnections 87 3.5.2 Achieving energy and bandwidth efficiency 87 3.6 Summary 88 Chapter 4. Context-Aware Computing 4.1_Ubiquitous or Pervasive Computing 91 92 4.2, What Is a Context? Various Definitions and Types of Contexts 94 4.2.1 Enumeration-based 94 4.2.2 Role-based ee : -) 4.3 Context-Aware Computing and Applications 96 4.3.1_Core capabilities for context awareness Ed 4.3.2 Types of context-aware applications 98 4.3.3 Developing context-aware applications 100 4.4. Middleware Support 102, 4.4.1 Contextual services 103 4.4.2 Actuator service 104 4.4.3 An example: context toolkit 404 4.44 Providing location context 105 4.5 Summary 106 Chapter 5. Introduction to Mobile Middleware 109 5.1_What Is Mobile Middleware? 109 5.2 Adaptation 110 .3_ Agents 41 : 5.4 Service Discovery 114 Chapter 6. Middleware for Application Developme! Adaptation and Agents 113 61 Acaptation 113 The spectrum of adaptation 114 "Resource monitoring 114 Characterizing adaptation strategies 115 An application-aware adaptation architecture: Odyssey 117 A sample Odyssey application 119 ‘More adaptation middleware 120, Copyrighted6.2 Mobile Agents _ 123 6.2.1 Why mobile agents? And why not? 125 62.2 Agentarchitectures CC‘“‘“‘*«s 62.3 Migration strategies 130 6.2.4 Communication strategies 73t 6.3 Summary 133 Chapter 7. Service Discovery Middleware: Finding Needed Services 137 7.1 Common Ground 140 arvic 142 7.2.1__Universally unique identitiers 142 2.2.2 Standardization 144 7.2.3 Textual descriptions _ 145 7.2.4 Using interfaces for standardization 149 More on Discovery and Advertisement Protocols 150 .1_Unicast discovery 150 Multicast discovery and advertisement 151 7.3.3 Service catalogs | 155 7.4 Garbage Collection 156 7.4.1 Leasing 156 7.4.2 Advertised expirations 158 7.5_ Eventing 159 7.6 Security 163 7264_Jink 7.6.2 Service location protocol 164 7.6.3 Ninja 165 7.7 Interoperability 166 7.7.1 Interoperability success stories 167 7.8_ Summary 167 Chapter 8. Introduction to Ad Hoc and Sensor Networks 171 8.1 Overview 171 8.1.1 Outline of chapter 172 8.2 Properties of an Ad Hoc Network 175 8.2.1 No preexisting infrastructure 175 c b ion 175 8.2.3 Power-limited devices 177 8.3_Unique Features of Sensor Networks 178 “8.3.1 Direct interaction with the physical world «SV. “Very limited 180 8.3.4 Operate without a human interface. 181 8.3.5 Specialized routing patterns 181 8.4 Proposed Applications 182 8.4.1 Military applications 182 8.4.2 Medical applications 785 8.4.3 Industrial applications 186 8.44 Environmental applications CST 8.4.5 Other application domains _ 188 Copyrighted materialx Contents Chapter 9. Challenges 9.1 Rooesaines Resources -1_No centralized authority Limited power “Wireless communication Storage constraints Limited input and output options 9.2 Security 9.2.1 Small keys 9.2.2_Limited computation 9.2.3 Changing network membership 3.2.4 Arbitrary topology 9.3 Mobility 9.31 Mobility requirements §.3.2_ Loss of connectivity 3.3.3 _Data loss 3.3.4 Group communication ‘Maintaining consistent views 9.4 Summary ‘Chapter 10. Protocols 10.1__Autoconfiguration 0.1.1_ Neighborhood dis Limited computation and storage & BRE BERBREB REESE EERE BB ia 10.1.2 Topology discovery 10.1.3 Medium access control schedule construction 10.1.4 Security protocol configuration 10.2 Energy-Efficient Communication 10.2.1 Multihop routing 10.2.2. Communication scheduling 10.2.3 Duplicate message suppression 10.2.4 Message fon 10.2.5 Duatradio scheduling 10.2.6 Sleep-mode scheduling 10.2.7 Clustering Mobility Requirements 10.3.1 Movement detection 10.3.2 Patterns of movement 70.3.3 Changing group dynamics 10.3.4 Resynchronization 10.4 Summary 10. @ Chapter 11. Approaches and Solutions 11.1_Deployment and Configuration T1.1.1_Random deployment 1.1.2 Scalability 11.1.3 _Seif-organization 11.1.4 Security protocol configuration BSREBE RRERESRR NERContents 11.1.5 _Reconfiguration/redeployment 249 11.1.6 Location determination 249 11.2 Routing 252 11.2.1, Event-driven routing 253 11.22 Periodic sensor readings 254 11.2.3 Diffusion routing 260 11.2.4 Directional routing 285 11.25 Group communication 268 11.2.6 Synchronization 271 11.3. Fault Tolerance and Rellability 273 11.31 FEC and ARQ 273 11.3.2 Agreement among sensor nodes (Reliability of measurements) 274 11.3.3 Dealing with dead or faulty nodes 278 11.4 Energy Efficiency 279 11.4.1 Uniform power dissipation 280 11.42 Sensor component power management 281 11.4.3 MAC layer protocols 282 11.4.4 Tradeoffs between performance and energy efficiency 283 11.8 Summary 284 Chapter 12. Wireless Security 287 12.1 Traditional Security Issues 287 12.1.1 Integrity 287 12.1.2 Confidentiality 288 12.1.3 Nonrepudiation 288 12.1.4 Availability 288 12.2 Mobile and Wireless Security issues 200 12.2.1 Detectability 200 12.2.2 Resource depletion/exhaustion 291 12.2.3 Physical intercept problems 291 12.2.4 Theft of service 201 12.2.5 War driving/walking/chalking 292 12.3 Mobility 209 12.4 Problems in Ad Hoc Networks 293 12.4.1 Routing 294 12.4.2_Prekeying 294 72.4.3 Reconfiguring 295 12.4.4 Hostile environment 295 12.5 Additional Issues: Commerce 295 125.1 Liability 296 12.5.2 Fear, uncertainty, and doubt 296 125.3 Fraud 296 12.5.4 Big bucks at stake 297 12.6 Additional Types of Attacks 297 12.6.1 “Man in the middie” attacks 297 12.6.2 Traffic analysis 298 426.3 Replay attacks 298 12.6.4 Buffer-overfiow attacks 298 12.7 Summary 299xii Contents Chapter 13. Approaches to Security 301 13.1_Limit the Signal 301 13.1.1 Wire integrity and tapping 301 13.1.2 Physical limitation 301 43.2_ Encryption 302 13.2.1 Public and private key encryption 302 13.2.2 Computational and data overhead 303 13.3 Integrity Codes 304 13.3.1 Checksum versus cryptographic hash 304 13.3.2 Message authentication code (MAC) 305 193.8 Paylosdversusheader 088 13.3.4 Traffic analysis 13.4 IPSec 307 13.4.1 Authentication header (AH) 307 13.4.2 Encapsulating security payload (ESP) 308 13.5 Other Security-Related Mechanisms 308 135.1 Authentication protocols 308 135.2 AAA 313 13.5.3 Special hardware 315 13.6 Summary 315 Chapter 14. Security in Wireless Personal Area Networks 317 444 Basicidea 14.1.1 Bluetooth specifications 317 4441.2 Bluetooth network terms SCSC~ST 14.1.3 Bluetooth securitymechanisms 820 14.2 Bluetooth SecuriyModes 8D 14.3 Basic Security Mechanisms 321 14.3.1 Initialization key 322, 143.2 _Unitkey 322 14.3.3 Combination key 322 143.4 Masterkey 0 8B 14.4 Encryption 323 14.5 Authentication 325 14.6 Limitations and Problems 325 14.7 Summary 327 Chapter 15. Security in Wireless Local Area Networks 329 454 Basicidea 15.2 Wireless AlphabetSoup 88. 15.3 Wired-Eduivalent Privacy (WEP) 333 15.3.1 WEP goals 333 15.3.2 WEP data frame 334 15.3.3 WEP encryption 334 15.3.4 WEP decryption 335 15.3.5 WEP authentication 8 15.3.6 WEP flaws ‘336 152.7 WEPfiyees 8BContents — xiii 15.4 WPA 340 168 soniye 8 15.5.1 Encryption protocols 340 15.5.2 Accesscontrolvia8021x 15.6 Fixes and “Best Practices” 344 15.6.1 Anything is better than nothing 344 15.6.2, Know thine enemy aM 15.6.3 Use whatever wireless security mechanisms are present 345 15.6.4 End-to-end VPN 345 15.6.5 Firewall protection 346 15.6.6 Use whatever else Is available 346 15.7 Summary 348 Chapter 16. Security in Wireless Metropolitan Area Networks (802.16) 349 16.1 BroadbandWirelessAccess 8D 462 IEEE@O216 16.3_802.16 Security 350 16.3.1_Key management a5 16.3.2 Security associations 351 16.3.3 Keying material lifetime 352 16.3.4 Subscriber station (SS) authorization 353 16.3.5 Encryption 353 18.4 Problems and Limitations 354 16.5 Summary 354 Chapter 17. Security in Wide Area Networks 357 Wi Basicides 8 12.2_CDMA 359 12 GSM 360 17.3.1_GSM authentication 360 17.3.2 GSM encryption 360 17.4 Problems with GSM Security 361 12.4.1 Session life 362 17.4.2 Weak encryption algorithm 362 17.4.3 Encryption between mobile host and base station only 363 17.4.4 Limits to the secret key 363 17.4.5 Other problems 364 15 The Four Generations of Wireless: 1G-4G_ 8 126_3G 364 12.7 Limitations 366 17.8 Summary 366 Appendix A_Brief Introduction to Wireless Communication and Networking 369 Appendix B Questions 375 Index _389Preface Mary sits with her PDA in a coffee shop in Woods Hole reviewing her appointments for the day and notes that the location for her lunch meet- ing with a CEO is marked TBD. After reviewing menus for a few nice restaurants, she selects an appropriate location. As she gets into her car, with notes for the lunch meeting she printed at the coffee shop, she quickly checks her credit card limit with her PDA, and then sends the CEO a note on the suggested location. Suddenly, 9000 miles to the west, asperm whale surfaces in the Pacific Ocean. The whale carries a small, low-power radio transmitter that has been recording its vital signs for the last 2 months. Now above the water, the transmitter establishes con- tact with a low-earth-orbit satellite and uploads its data. It also uploads vital signs for other similarly equipped whales it passed since the radio last contacted the satellite. The data indicate a growth in the pod pop- ulation, Scientists halfway around the world are thrilled with the new data and news filters up the chain to the company CEO, who can now approve a research grant for the chief scientist of a marine biology insti- tute he is meeting for lunch. As she leaves the restaurant, Mary shares the good news with members of her research group using her cell phone. Acommon thread in the above scenario is the use of mobile and per- vasive computing. Mobile computing is characterized and driven by portable, lightweight hardware, wireless communication, and innova- tions in application and system software. The technology drives new applications which in turn fuel the demand for the new technology. Pervasive computing uses small, battery-powered, wireless computing and sensing devices embedded in our environment to provide contextual information to new types of applications. The trend of technology get- ting smaller and more portable is likely to continue for the foreseeable future. Bandwidth available to a home user through a broadband Internet connection exceeds the bandwidth available to all but the largest organizations several years ago. The bandwidth in low-priced network connections continues to increase at a similar rate. Wireless and cellular technologies allow users to connect to networks from practicallyxvi Preface any location and to remain connected in transit. This is the age of mobile and pervasive computing! This book provides a focused look at important topics in Mobile and Pervasive Computing. Although the underlying technology is essential, we avoid being completely bound to it. Technology will change and evolve, and in fact has changed even during the course of writing this book. Data rates increase. Protocols evolve. More efficient algorithms are created. But the underlying techniques, the essentials of how computers share, protect, and use data in an environment in which the location is not fixed, remain relatively stable. ‘I'his book is intended to address these fundamental points, the ones that continue to be significant long after this book is published. Audience This book has been designed for two audiences. The first group con- tains professionals seeking to get a solid understanding of mobile computing—the problems and solutions, the protocols, and the appli- cations. The second group comprises students in a senior-undergraduate or graduate-level course. To help the second group, Appendix B con- tains both short questions, to help review the chapters, and longer ques- tions and project ideas that instructors may select as starting points for lab assignments or for further research. Book Organization Conceptually, the book is organized into four parts: Part One, which comprises the first four chapters, covers issues related to mobile and pervasive computing applications. This includes not only disseminating and caching data, but also routing and loca- tion management to determine where to send data in the first place. Context aware computing is the final chapter of the first part. Part Two, which comprises Chapters 5 to 7, focuses on middleware, the layer that bridges mobile computing applications to the underly- ing systems that support this mobility. Middleware topics include agents and service discovery, as well as techniques and methods. Part Three, which comprises Chapters 8 to 11, covers an important enabling networking technology for pervasive computing: adhoc and wireless sensor networks. Adhoc networks form when needed, on-the- fly, without central management or infrastructure. Wireless sensors record some type of data, such as temperature and humidity, and form an adhoc network to efficiently disseminate these readings.Preface — xvii Part Three discusses the applications, problems, approaches, and protocols for wireless sensor networking. Part Four, which comprises Chapters 12 to 17, covers security in wireless environments. After describing common security problems, these chapters discuss how security is handled in personal, local, met- ropolitan, and wide area networks, as well as current research work and future trends. How to Use This Book Since this book provides the coverage of almost all the important aspects of mobile and pervasive computing, it can be used in several ways. Two of the authors have used this book in semester-long courses. Since each part of the book is almost self-contained, an instructor can easily select the parts most suitable for his or her course, permitting the book to be used in quarter-long courses or even short introductory courses. The book provides two appendices to assist in teaching this course. Appendix A provides a brief introduction to wireless communication and net- working that can be used by instructors to quickly cover the essentials of wireless communication and networking. Appendix B provides ques- tions for each chapter of the book. These questions can be used by an instructor to encourage students to further explore the topics covered in this book.Acknowledgments Over the entire process of the development of this book, many people have helped us; without them, this book would not have been possible. First we thank the many people who reviewed drafts of this book: John Bodily, Michel Gertraide, Manish Kochhal, Fernando Martincic, John Quintero, Georgios Varsamapoulos, and Yong Xi. We are grateful to the National Science Foundation (NSF) and the Center for Embedded Networking Technologies (CEINT) for supporting our research in the areas related to this book. We thank our employers, the Arizona State University, the University of New Orleans, the Wayne State University, and the Odyssey Research Associates/ATC-NY, for providing the working environment that made this book possible. We also thank the Ohio State University, where all of us obtained our doctoral degrees, and the Buckeye Spirit. Go Bucks, Go! Our heartfelt thanks go to our original sponsoring editor Marjorie Spencer; our editor Steve Chapman, who provided the needed level of support and encouragement to see this book through from conception to completion; Priyanka Negi, who oversaw the copyediting process; and the entire crew of McGraw-Hill. Last but not least, we thank our family members and friends who accepted the time away from them spent writing this book. Without their support, this book could never have been finished.Chapter Mobile Adaptive Computing 1.1 What ts Mobile Computing? What is mobile computing? What are its different aspects? Basically, mobile computing systems are distributed systems with a network to communicate between different machines. Wireless communication is needed to enable mobility of communicating devices. Several types of wireless networks are in use currently, and many books have been devoted to mobile communication. In this book we cover only the basic aspects of mobile communication. However, this book is not devoted to wireless communication and networks. Readers not familiar with wire- less communication and networking can refer to Appendix A for a short description. In this book, we are concerned mostly with the logical aspects of mobile communication. What is the difference between mobile computing and mobile commu- nication? Communication is necessary for distributed computing. Many mobile computing tasks require mobile communication. But that is not the end of it. Mobile communication does not solve all the problems. As we will see in this book, many issues need to be resolved from a higher- level perspective than just being able to exchange signals/packets. What interests us the most about mobile computing is what we can do with this new facility of mobile communication. What new applica- tions can be enabled? Think about the applications we currently use. Many of these applications do not adapt very well to our needs. For example, suppose that we are browsing the World Wide Web and are interested in going to an Italian restaurant for lunch. What information will we get when we do a search for “Italian restaurant” through search engines such as Google? Usually we get a lot of information that we do not require. The search engines of today do not care about who is invoking2 Chapter One the search or the location from where the search is being invoked. You can type “Italian restaurant in Tempe” or “Italian restaurant near Arizona State University.” You can make your query as specific as pos- sible. But wouldn't it be nice if the applications were aware of your loca- tion or, more generally, your context and if the response from the search engine were ordered according to your context? Surely, such context aware applications will make us more productive. Let us consider another application, such as video streaming over the Internet. Suppose that you are on the move, watching a movie. Wireless communication is different from wired communication. When you have a wire, you usually have a fixed bandwidth available to you and your application. Once the application is started and the movie starts, you can watch the movie at a good quality of service (QoS). (Currently, this is not really true if streaming is done across the Internet. Nevertheless, it is expected that this will become a reality in the near future.) In wireless communication, however, the bandwidth is mostly shared among several users in a dynamic fashion. This is to say that no dedicated bandwidth is available. Even if your application can reserve certain wireless band- width, the usable bandwidth fluctuates owing to the nature of the wire- less medium. There are both short- and long-term fluctuations. The question is, “How should the application respond to these fluctuations?” One way that the application can respond is in a uniform manner, irre- spective of what you are watching. The other approach is to respond based on the type of movie you are watching. For example, suppose that you are watching an action movie. The application can reduce the band- width requirement by switching from full-color video to black and white or by reducing the resolution. On the other hand, if you are watching an interview of Bill Gates, the application may switch simply to audio streaming. There are several important points to note here. These deci- sions are based on the content of the video, and the decision making may involve both the client and the server (the client needs to inform the server that it no longer wants the video frames, only the audio). In this chapter we will reexamine computer (software) system design to see how it needs to be changed to accommodate mobility. As we will see, many of the changes are concerned with providing mechanisms for adapting to changing environmental and system conditions, such as the location and the available resources. Mobile computing is about pro- viding information anytime anywhere or, more generally, computing anytime and anywhere. Mobile computing is also about dealing with limitations of mobile computing devices. For example, personal digital assistants (PDAs) and laptops have small interfaces and are powered by batteries. One of the major issues is how to do computation in an energy-efficient manner. Battery technology is not advancing at the same pace asMobile Adaptive Computing 3 processor technology, so it is not expected that battery capacity will double at a fixed rate, as is true about processor speeds, according to Moore's law. One does not have to deal with these issues when design- ing systems for stand-alone or distributed systems. One may have to deal with issues of fault tolerance in distributed systems, such as server crashes or network link failures. However, energy usually is not an issue. In mobile systems, energy becomes a resource like processing time or memory space. Therefore, one now has to design resource man- agement techniques for energy, just as traditional operating systems deal with process and memory management. Different computers in a mobile computing environment may have dif- ferent capabilities. Any collaborative activity between these devices needs an underlying software entity to deal with the heterogeneity of these devices. This software entity is called a middleware layer. A middleware layer also may allow a mobile device and a wired device to interact. When mobile clients move from one administrative domain to another, they may want to know what new services are available. How about security or privacy? Wireless communication takes place over an “open” wire and is relatively easy to tap. It may seem that tradi- tional techniques of cryptography can be used to secure wireless commu- nication. However, the main problem is that these secure techniques are designed for wired networks and are computation- and communication- intensive. Attempts to reduce these overheads lead to security schemes that are relatively easy to break. We will look into security schemes for mobile networks in later chapters. 1.2 Adaptability—The Key to Mobile Computing Humans have evolved to be on the top of the food chain. It is probably unde- niable that we have done so by being able to adapt quickly and effectively to varying situations. It is only because of our ability to adapt that we can be found from Arctic regions to the Sahara Desert. What are the techniques by which we are able to adapt to such diverse environments? Can we incorporate some of these techniques into our computing systems? Anyone who has used computers seriously would like them to be more resilient and adaptive to our needs and circumstances. Computing systems and appli- cations fail for various reasons. What is most frustrating is when they fail for no apparent reason. You install a new application, and some other apparently unrelated application stops functioning. And sometimes we just want our computers to learn from our past actions and act proactively and appropriately. Making systems resilient and adaptive is not a trivial task. It took us millions of years to rise to the top of the food chain. Computers as we know them today are not even a hundred years old.4 Chapter One The vision of mobile computing is to be able to roam seamlessly with your computing devices while continuing to perform computing and communication tasks uninterrupted. Many technological advances at various fronts such as security, privacy, resource allocation, charging, and billing have to be made to make this feasible. A quintessential characteristic of any solution to mobile computing problems is its abil- ity to adapt to dynamic changes in computing and communication envi- ronments. A system's agility to react to changes in the computing environment and continue its computing tasks uninterrupted is a new measure of performance in mobile computing environments. Consider a scenario in which you move from one coverage area of an access point to another while a video streaming application is running on your computer. To continue receiving the video stream uninterrupted and without deterioration in video quality, the video stream packets now should be routed automatically through the new access point. In an Internet Protocol (IP)—based network this may involve the mobile client obtaining a new IP address in the new access point’s IP network and informing the server of the new address so that it can now send the pack- ets to the new address. Many more sophisticated solutions to this prob- lem have been developed. The point here is that the underlying system has to take many actions automatically to ensure continued connectiv- ity, in this case uninterrupted viewing of the video stream. In essence, the system has to adapt to the changes in the environment, such as the network configuration and the availability of communication and com- putation resources and services. However, is this enough? More specif- ically, the preceding adaptation scheme does not take into account the applications requirements, and the applications themselves did not play any part in the adaptation. Does this application-transparent way of adapting suffice to meet the goals of mobile computing? 1.2.1. Transparency Transparency is the ability of a system to hide some characteristics of its underlying implementation from users. Much of the research effort in distributed computing has been devoted to developing mechanisms for providing various forms of transparency. Examples of these include the following: ® Access transparency is the ability of a system to hide the differences in data representation on various machines and the mode of access of a particular resource. = Location transparency is the ability of a system to conceal the loca- tion of a resource. Related to location transparency are name trans- parency (which ensures that the name of a resource does not revealMobile Adaptive Computing 5 any hints as to the physical location of the resource) and user mobil- ity (which ensures that no matter which machine a user is logged onto, she should be able to access resources with the same name). = Failure transparency is the ability of the system to hide failure and recovery of a system component. Mobile computing systems can be viewed as a form of distributed system, and attempts can be made to provide “mobility transparency,” which would encompass the transparencies just mentioned. This would, in essence, support application-transparent adaptation. But is this an achievable, or even desirable, goal for building mobile computing sys- tems and applications? Let us look closely at the characteristics of the mobile computing environment and their implications in these regards. 1.2.2. Constraints of mobile computing environments: Mobile computing has many constraints that distinguish a mobile com- puting environment (MCE) from the traditional desktop workstation/ PC-based distributed computing environment. Notable among these are the following (Satyanarayanan, 1996a, 1996b): 1. Mobile computers can be expected to be more resource-poor than their static counterparts. With the continued rapid improvement of hard- ware technology, in accordance with Moore's law, it is almost certain that a laptop computer purchased today is more powerful that the desktop computer purchased just a year or even a few months ago. However, mobile computers require a source of electrical energy, which is usually provided by battery packs. Since batteries store a finite amount of energy, they need to be replaced or recharged. The first option costs money, and the second option, although cheaper in terms of money expended, requires plugging in the computer for recharging, restrict- ~ ing mobility. This has an impact on the design of mobile computers— all the hardware and software components in mobile computers are designed to reduce energy consumption and to increase the life- time of the batteries, For example, processors on mobile computers are designed to consume less energy and, consequently, achieve a lower computation performance. 2. Mobile computers are less secure and reliable. Since mobile computers accompany their users everywhere, they are much more likely to be lost or stolen. Furthermore, they are more likely to be subjected to hostile or unfriendly use, e.g., a child throwing his daddy’s PDAin a tantrum. 3. Mobile connectivity can be highly variable in terms of its perform- ance (bandwidth and latency) and reliability. Disconnections, both6 Chapter One Acoourcarpaar More reliance on boo Infrastructure Less secure ‘e 26 Mobile Dynamic changes fs computing environment in environment &s Intermittent connectivity ° |__| Less reliance Eneray-constrained on infrastructure Figure 1.1. Need for dynamic adaptation in mobile computing environments. voluntary and involuntary, are common. The link bandwidth can vary by orders of magnitude over time and space. Thus, in general, resource availability and quality vary dynamically. The preceding characteristics of a mobile computing environment require rethinking about.how mobile applications and systems should be designed. Resource paucity and the lower reliability of mobile devices point toward designing systems in such a manner that more reliance is placed on the static infrastructure. On the other hand, the possibility of disconnections and poor connectivity point toward making systems less reliant on the static infrastructure. Further, as mobile devices are moved around (or even if they are not), their situations keep changing over time. Mobile devices should change their behavior to be either more or less reliant on static infrastructure depending on current con- ditions. Figure 1.1 illustrates this need for dynamic adaptation in a mobile computing environment. 1.2.3. Application-aware adeptation Who should be responsible for adaptation—the application, the system, or both? There are two extreme approaches to designing adaptive sys- tems: application-transparent (the system is fully responsible for adap- tation) and laissez-faire (the system provides no support at all) (Satyanarayanan, 1996a). Obviously, the laissez-faire approach is not desirable because it puts too much burden on the application developer. Further, no support from the underlying system restricts the types of adaptations that can be performed. However, as the following example points out, the application-transparent approach is not sufficient either. Consider two different multimedia applications. In one you are video- conferencing using a mobile device, and in the other you are watching a live video stream from a remote server on your mobile device. Now con- sider the following scenarios. In one, you move from an area with suf- ficient bandwidth for your application to an area where the amount ofMobile Adaptive Computing 7 bandwidth is less than that needed by your application. In the other, your laptop’s battery power level drops considerably. Both scenarios deal with changes in availability of resources. How would you like your system/application to behave under each scenario? In the application (user)-transparent approach, the system/applica- tion may behave the same irrespective of the application running. However, different responses may be suitable depending on the type of application that is running. For example, in the first scenario, a non- adaptive system may just do nothing and let the audio/video quality drop. In the second scenario, the system may just give a warning to the user without any assistance on how to deal with the situation. In an adaptive system, various behaviors can be envisioned. For example, the system may try to do its best in both situations. However, the system’s adaptation does not take into account the kind of application that is run- ning. For example, in the first scenario, the system may try to adapt by requesting that the server or other peers start to send lower-quality video, in effect requiring lower bandwidth. In the second scenario, the system may try to conserve energy by reducing the intensity of the back- light of the display (besides warning the user of the lower battery power level). A still more adaptive approach is possible in which the system interacts with the user/application in deciding how to adapt. In the application-transparent approach, the responsibility of adap- tation lies solely with the underlying system. On the other hand, in the application-aware approach, the application collaborates with the under- lying system software. Here the system provides status information about the available resources. The application uses this information to make decisions on how to adapt to changes in the resource availability. Each application can adapt in its own way. Figure 1.2 illustrates the spec- trum of adaptation strategies that are possible (Satyanarayanan, 1996a). Application-aware (Collaboration between system and application) T Laissez-faire Application-transparent (no system support) (no changes to. application) Figure 1.2 The spectrum of adaptation strategies (Satyanarayanan, 1996a).