Transparent and Scalable Client-side Server Selection using Netlets

Replication of web content in the Internet has been found to improve service response time, performance and reliability offered by web services. When working with such distributed server systems, the location of servers with respect to client nodes is found to affect service response time perceived by clients in addition to server load conditions. This is due to the characteristics of the network path segments through which client requests get routed. Hence, a number of researchers have advocated making server selection decisions at the client-side of the network. In this paper, we present a transparent approach for clientside server selection in the Internet using Netlet services. Netlets are autonomous, nomadic mobile software components which persist and roam in the network independently, providing predefined network services. In this application, Netlet based services embedded with intelligence to support server selection are deployed by servers close to potential client communities to setup dynamic service decision points within the network. An anycast address is used to identify available distributed decision points in the network. Each service decision point transparently directs client requests to the best performing server based on its in-built intelligence supported by real-time measurements from probes sent by the Netlet to each server. It is shown that the resulting system provides a client-side server selection solution which is server-customisable, scalable and fault transparent.