An Optimal Load-Balancing Policy of a Distributed Multimedia-on-Demand System

• Main Authors: Tien-Lung Wen, 溫天龍
• Other Authors: Ren-Hung Hwang
• Format: Others
• Language: en_US
• Published: 2001
• Online Access: http://ndltd.ncl.edu.tw/handle/01647490554071430691

碩士 === 國立中正大學 === 資訊工程研究所 === 89 === Multimedia on Demand (MOD) is a promising network service for next generation Internet. However, scalability is always a major issue for a single-server MOD system. Therefore, distributed MOD systems have been proposed to achieve high scalability. In a distributed MOD system, a user’s request is either served by a local server, a remote server, or rejected. The policy to select the most appropriate remote server, referred to as the remote server selection policy, becomes a very important for research. The factors affect the selection decision includes the reward of serving a request and the communication cost for serving a request remotely. In this thesis, based on the Markov Decision Process (MDP), we propose a remote server selection policy that maximizes the total revenue for a distributed MOD system. Our numerical results show that the proposed policy outperforms the policies in the literature in various network configurations and traffic loads.