A Dynamic Server Selection Method Using QoS Statistics, Vol.J86-B,No.3, pp.499-510

概要

論文の詳細を見る
Mirroring is a technology to improve service reliability and user QoS in the Internet application by placing copies of content or a file at the multiple servers at different sites in the Internet, where one of the selected servers will respond to the user request. Assuming file download in Web applications, this paper investigates server selection methods to minimize file download time. In particular, this paper focuses the client-side server selection method and proposes a dynamic server selection method based on QoS statistics data. In this method, QoS data, periodically collected, is classified by day of the week and time of the day and used as the basis of server selection. We discuss the optimal values for various parameters in the proposed server selection method based on the experiments in the Internet. We next compare performance of the proposed method and several conventional and popular server selection methods based on the experiments in the Internet. We use the objective performance measures such as the average and maximum download times as well as download give-up ratio and stress level, which may reflect user perception. We show that the proposed method outperforms other methods based on these evaluations.
社団法人電子情報通信学会の論文
2003-03-01