Finding an optimal mirror site

Abstract

Mirror sites help to reduce the load at individual servers and enable client requests to be serviced by any of the servers. Typically, a client requests service from a single mirror site. When there are more servers (mirror sites) to provide the same services, the client has to choose one randomly to get served. Without prior knowledge about the reliability and performance of mirror sites, the client might not be successful in accessing the resources on their first attempt. As an extension of our previous research (Y. Kim and M. Swain, "The role of Intelligent Model for Data Source Selection in Distributed Server Environment", submitted to ACMSE, GA, 2005), we propose a system enabling a client to access files from the mirror sites with confidence using intelligent agents in a multiagent environment. In this research, we have considered factors like bandwidth, availability and capacity of servers for our agent model in choosing a more probable data source in an intelligent way. Also discussions on the agent communications based on web services have been given.