Research And Implementation Of Technologies For Distributed Video Server

Distributed Hierarchical Video Server is a key component in a large-scale VOD (Video-On-Demand) System. Designing and implementing a high-performance video server is remarkably valuable for research and market. In the paper, we discuss media streaming and stream scheduling in detail. We also introduce our distributed VOD system.Video server is an important research subject in networked multimedia. It is not only the core module of a VOD system, but also useful in lots of fields such as digital library and remote education. We analyze the researching trend of the VOD system, and then give the definition of the video server. After introducing several kinds of server architecture, we present the key problem in the research of video server. As an ideal solution, distributed hierarchical video server can be used in a large-scale VOD system, so it is focused on nowadays. In a video server, media streaming is related with the following problems: how to transform the file-based media to stream-based media, how to effectively transfer streaming media in the intranet or the internet. After confirming the feasibility of deploying the VOD service on PC platform, we compare the adaptability of the server-pushing architecture with the one of the user-pulling architecture. We also propose an extended architecture of distributed hierarchical VOD system. Based on this architecture, we design an extensible and cost-effective prototype. In the following sections, we will describe the control flow and the data flow among all the entities of the prototype in detail and introduce the key algorithms of the management server, the video server and the client.An appropriate streaming-scheduling algorithm can notably improve the capacity of the video server. After analyzing the existing streaming-scheduling algorithms, we propose a novel Patching-First algorithm. To effectively evaluate the streaming-scheduling algorithms, we design and implement an experimental platform based on the mathematical model of the Patching-First algorithm. Our simulations show that the Patching-First algorithm outperforms the other streaming-scheduling ones.