Optimizing continuous media delivery by multiple distributed servers to multiple clients using a genetic algorithm

Abstract

In this paper we explore the potential of a VoD system that is based on a paradigm that has been recently proposed: that of combining many distributed servers to handle the delivery of each requested media. When faced with a system comprised of N servers and M clients, the problems to be addressed are: (i) how to split the delivery task among different servers and (ii) how to pair clients and servers, the objective being to minimize the access time of the clients. To this end, we present an analytical framework that enables the division of the delivery process for each client in a distributed manner. This framework is coupled with a genetic algorithm that enables an optimal or near-optimal solution to the problem of pairing clients and servers, in a small number of generations. The paper is concluded by a rigorous study of a N-servers, M-clients system that answers a number of important questions like, what is the quality of service achieved and how our proposed system behaves under increased load. © IFIP International Federation for Information Processing 2004.