Transaction processing in mobile database system

Mobile Computing has become a reality thanks to the convergence of two technologies: appearance of powerful portable computers and development of fast reliable wireless networks. Among, applications that are finding their way to the market of Mobile Computing, those that involve data management, hold a prominent position. In past few years there has been tremendous surge of research in the area of data management in mobile computing. However mobile computing as exists today is not fully capable of processing database transactions. This research concentrates on providing transactional support appropriate for mobile computing. Mobile Database System (MDS) is a distributed system that supports mobility during information processing. A new transaction model for MDS is first presented. The scope of traditional transaction properties is expanded by introducing new location property to transaction. Transaction execution could be distributed over multiple components. Some sub transactions would be executed on MU and rest on the servers on wired network. For consistency preserving execution either all subtransactions should commit or all should abort. We propose TCOT (Transaction Commit on Timeout protocol) protocol for MDS. TCOT protocol is based on timeouts and uses lesser number of messages than 2-Phase Commit (2PC), which is commonly used in database systems for fixed networks. We show using detailed simulation study that TCOT performs better than 2PC. Next we propose mechanisms for processing transactions locally at the mobile unit in presence of data broadcast. Broadcast-based data dissemination is likely to be a major mode of information transfer in mobile computing and wireless environment. As these systems evolve, they will be used to run sophisticated applications many of which will involve data whose consistency must be maintained and data updates may originate at mobile client. In pull-based broadcast environment we propose integrated approach to broadcast data scheduling and transaction processing. Our approach aims at reducing tuning time, commit time and number of aborts for the transaction. For push-based environment a new concurrency control mechanism, which uses requires low bandwidth, ensures availability, accommodates the disconnection problem, and is scalable. We did detailed performance study of transaction processing in broadcast environment.