Autonomous transaction processing using data dependency in mobile environments

Abstract

Transaction processing in mobile database systems faces new challenges to accommodate the limitations of mobile environments, such as frequent disconnections and low bandwidth. We propose a transaction processing protocol that increases the autonomy of clients, based on the dependency relation among updated data items. Lists of dependents, sent by the server to the mobile clients along with requested data items, are used by each client to build partial serialization graphs. Utilizing the graphs, mobile clients can autonomously verify serializability of locally executed read-only transactions. This information can also help mobile clients in early detection of the necessity to abort update transactions. Simulations for various data access patterns initiated by mobile clients provide insights on performance of the proposed protocol. Performance is heavily dependent on the depth of the dependency information for each data item.