Designing Efficient Fault-Tolerant Systems on Wireless Networks

The falling cost of both communication and mobile computing devices (laptop computers, hand-held computers, etc.) is making mobile computing affordable to both business users and private consumers. People on the road want to use their laptops to read Email, login on remote machines, buy or sell stocks, and so on; rescue workers at disaster sites (fires, floods, earthquakes, etc.) would like to send messages, keep records, and communicate with each other by mobile devices; a general needs to gather information from his soldiers and send commands to them. In these applications, a system failure can cause loss of opportunity, financial loss, or even loss of human lives. It is clear that in many applications, highly dependable systems are needed. Mobile wireless environments pose challenging problems in designing fault-tolerant systems because of the dynamics of mobility and limited bandwidth available on wireless links. Traditional fault-tolerance schemes cannot be directly applied to these systems. However, fault tolerance is much more important in mobile computing systems since mobile computing systems are more prone to failures. This is because (i) wireless networks have high error rates and more frequent disconnections and (ii) mobile devices are more prone to failure, physical damage, or loss. In this thesis, we address these problems at two levels: the network level and the operating system level. At the network level, we design fault-tolerant network protocols. In particular, we design efficient fault-tolerant channel allocation algorithms and fault-tolerant location management schemes. At the operating system level, we study checkpointing and recovery technique, which is an attractive approach for transparently adding fault-tolerance to distributed applications. Even though this technique can tolerate both mobile device and wireless network failures, we still study fault-tolerant network protocols, since wireless networks are not reliable and using operating system level approaches to tolerate network failures have high overhead compared to using network level approaches. Channel allocation has received considerable attention, but none of the existing schemes takes into consideration site failures, communication link failures, or even network congestion. Although the checkpointing and recovery approach has been extensively studied in traditional distributed systems, the new constraints posed by mobile computing systems, such as low bandwidth of wireless channels, high search cost, and limited battery life, make traditional checkpointing algorithms unsuitable to them. Mobile computing becomes an emerging trend. Due to the constraints of wireless networks, designing faulttolerant systems on wireless networks is a challenge. The author has many years experience of designing