Agent Based Architecture for Parallel and Distributed Complex Information Processing

Abstract

Information processing is a rather multidisciplinary field of research involving various communities resulting in a wealth of applications and solutions. Many effective solutions have been proposed to solve several specific problems. Fortunately, these solutions are generally complementary and can cooperate together in order to solve more complex problems or to improve some specific solutions. Consequently, we propose in this paper a flexible open architecture able to integrate complementary solutions according to the simple rule of plug and play. The proposed architecture is based on agent technology which provides especially autonomy and intelligence. Such aspects generate important added values in building open architectures. The proposed architecture has the great advantage to offer to users a flexible infrastructure where they can easily govern the response time of their own applications to be processed. This flexibility is due to the polyvalence of the corresponding agents. An analytical study has been made to ascertain and evaluate the key performance parameters of this proposed architecture. Conducted experiments show first that this architecture is indeed an efficient framework to support the integration of complementary solutions. Second, we found that the number of agents to be used, the achievable speedup and efficiency factors depends mainly on the task granularity which can be easily governed by users and the capacity of the underlying network. Finally, we found that even a loosely coupled architecture such as a LAN or a grid computing can support effectively the proposed approach.
Copyright © 2015 Praise Worthy Prize - All rights reserved.