1 Performance Analysis of Mobile Agents in Heterogeneous Mobile Database Systems

Multidatabase systems are widely used as a mechanism to integrate distributed heterogeneous data sources. The Summary Schemas Model (SSM) as a multidatabase solution provides high performance, supports scalability, and can handle imprecise queries. A distinct characteristic of this model is that it utilizes a Thesaurus to capture the semantic relation between database contents and represents information in an abstract manner. However, in today's mobile computing environment, the SSM faces several new challenges including handling unreliable wireless communication and supporting mobile users that have limited resources. The advances in mobile agent technology have enabled multidatabase designers to address these obstacles. The execution autonomy of mobile agents relaxes the requirement on network connection while the conventional distributed system design paradigm, the client-server model, depends heavily on the reliability of the network. Moreover, the execution autonomy feature provides a great opportunity for conserving energy, a limited and critical resource of mobile devices. Because of the many assets mobile agents possess, in recent years, mobile agent based distributed information system development is thriving [11][20][21][23]. In order to take full advantage of the SSM and the agent technology, we proposed and implemented a novel multidatabase information retrieval system, namely MAMDAS, which represents an application of mobile agents within the Mobile Data Access System framework. It adopts the SSM as the underlying multidatabase organization model and uses mobile agents to handle queries. Compared to the client-server based SSM prototype, the MAMDAS is 6 times faster and demonstrates greater scalability and robustness. In addition, our past experience also showed that a centralized Thesaurus is a source of bottleneck in MAMDAS and hence, to 2 improve the performance one has to adopt a decentralized Thesaurus organization within the SSM topology. To our best knowledge, in spite of the large body of research that has been done in agent based system design, the energy efficiency issues remain unexplored. In this paper, we mainly address two issues: • Optimize the performance of MAMDAS by decentralizing the Thesaurus. • Analyze the effect of mobile agents on mobile device energy consumption by applying different power management schemes. Our experimental results show that the decentralized Thesaurus can significantly improve the scalability of the system. The results also show that by deploying appropriate power management policy, the mobile device energy savings can be as high as 51%. We believe that such a high performance, energy efficient, scalable, and robust global information retrieval system can …