村上研究室 大阪大学大学院情報科学研究科 情報ネットワーク学専攻 インテリジェントネットワーキング講座(工学部 電子情報工学科 情報通信工学科目)

english

Master Thesis - 2003

A Study On Hybrid Content Delivery Scheduling Method
ZhengYu Xie

At present, high-speed transmission technology has spread rapidly. And the broadband internet has permeated into people's daily life. With various way of people's life style, the demand to Video on Demand (VoD) technology is increasing because of the increasing demand to the video or TV program through network.

The existing systems normally delivery contents with the technology of unicast, broadcast or multicast. In the unicast systems, point-to-point connections are built from the server to clients; the bandwidth is needed in proportion to the number of clients. Thus, the unicast systems are not spread widely because of the low system efficiency. In the broadcast systems, the response time is stable even the system load is high, but it wastes the bandwidth very much in principle when it transfers the contents which are not popular. In the multicast systems, with the on-demand and multicast transfer scheduling, the efficiency of response time and request blocking rate is decreased; however when the request rate is high, the efficiency of response time and request blocking rate increase rapidly. There are strong points and weak points respectively with the existing content delivery systems.

In this research, we propose a hybrid system which deliveries the contents with plural methods instead of a single method. In the hybrid system, the server deliveries the popular contents with broadcast, the other contents with multicast by using multi-channels adaptively. The contents are decided to transfer on broadcast channels or on multicast channels according to the contents' popularity degree. And the allocated multi-channels can be changed from broadcast to multicast or from multicast to broadcast if the contents' popularity degree changes. By this way, in the hybrid system, the weak points of broadcast and multicast are reduced; the strong points are increased. In this paper, we also prove that the new algorithm has higher efficiency by comparing the proposed system to conventional systems.

PDF file(local access only)
pagetop