Adaptive multiversion data broadcast organizations

Abstract

Broadcasting provides an efficient means for disseminating information in both wired and wireless setting. In this paper, we propose a suite of broadcast organization schemes for multiversion data broadcast, i.e., data broadcast in which more than one value is broadcast per data item. Besides increasing the concurrency of client transactions, multiversion broadcast provides clients with the possibility of accessing multiple server states. For example, such a functionality is essential to support applications that require access to data sequences and have limited local memory to store the previous versions, such as in the case of sensor networks. We identify two basic multiversion organizations, namely vertical and horizontal broadcasts and propose an efficient compression scheme applicable to both. We also consider a multiversion client data cache and introduce appropriate cache replacement techniques. Finally, we propose an adaptive scheme that dynamically selects the appropriate broadcast organization based on the client access pattern. We provide performance evaluation results for both flat and broadcast disk organizations and for a variety of client query patterns.