Performance Evaluation of Mobile Agents for E-Commerce Applications

Abstract. Mobile agents are emerging as a promising paradigm for the design and implementation of distributed applications. While mobile agents have generated considerable excitement in the research community, they have not translated into a significant number of real-world applications. One of the main reasons for this is the lack of work that quantitatively evaluates (i) the effectiveness of one mobile agent framework over another, and (ii) the effectiveness of mobile agents versus traditional approaches. This paper contributes towards such an evaluation. We identify the underlying mobility patterns of e-commerce applications and discuss possible client-server (CS) and mobile agent (MA) based implementation strategies for each of these patterns. We have compared the performance of three mobile agent frameworks, viz., Aglets, Concordia and Voyager, for their suitability for an e-commerce application development. We have performed experiments to quantitatively evaluate the effectiveness of CS versus MA strategies, for the identified mobility patterns. We used Java sockets for the CS implementation and the ObjectSpace VoyagerTM (Voyager) framework for the MA implementation. In this paper, we present our observations and discuss their implications.