DEVELOPING AN ARCHITECTURAL MODEL AND MAINTAINING TRANSACTIONAL CONSISTENCY FOR UNIVERSAL MOBILE BROADCAST.

Abstract

Managing data is very important in the mobile environment. Due to the limited bandwidth of the mobile systems, data should be organized and presented to the user based on its need. Moreover, when a number of users have the same data interest, data broadcast can be used to manage sending the needed data to the listening users. This eases the impact of limited bandwidth and improves the performance of the mobile system. In this research, we first develop a suitable architecture for broadcasting data in the mobile environment. Our aim is to allow the clients to get the needed data in an efficient way, and at the same time to allow the participation of clients in updating the content of the broadcast. Secondly, we examine the inconsistency problem encountered through the concurrent execution of read/write transactions and broadcasting data in the mobile environment. In the data broadcast environment, consistency occurs only if there is control over the interleaved broadcast and update of the same data items. While a client is listening to the data items through the broadcast to acquire the needed data, write transactions at the server may alter the contents of those data items being read by the clients. If there is no control over the broadcast/write transactions, the client may deal with inconsistent data. Therefore, we should find a mechanism to manage the concurrency of the broadcast and write transactions at the server. Our aim is v to allow a high degree of transaction concurrency while preserving the consistency issue at the SaIne time. We propose a new concurrency control algorithm that solves the inconsistency problem which may be observed by the client during the broadcast. The algorithm manages the overlap of broadcast and update of data items so that mobile clients always see consistent data items. We use Broadcasted and Updated Cycles (BUC) for the purpose of conflict checking. Our technique is simple, yet efficient in both space and overhead. The main objective of our approach is to improve throughput/response time and reduce the overhead of the update transactions. BUC is resilient to the problems caused by communication failures (e.g. disconnection). Moreover, our technique allows only schedules which are conflict serializable and makes the Information Server (IS) rely only on the information stored in its database. We present a broadcast classification to show the unique and inherent properties of the broadcast environment. In our work we provide a general broadcast classification which might be used to show the different characteristics of different broadcast models. This classification is important to select the optimal and appropriate model according to the very special characteristics of the targeted environment. In addition, we provide a comparison to several broadcast models. Finally, we show the performance of our technique BUC through not only qualitative arguments and analytical studies, but also through simulation models. Our main objective is to show the flexibility and generality of our technique in achieving a pervasive and ubiquitous computing.