Research And Design Of P2P Video-on-Demand System Using Cache Strategy Based On Association Rule

With the high-speed development of computer networks, digital compression and storage technology, IP based real-time multimedia communication services grow rapidly. Currently, the IP based stream-media techniques gradually mature, the Internet media streaming services such as Video on Demand (VOD) have grown rapidly in recent years. But the service quality is not good enough, because the Server in Server/Client system model would be limited both on the hard disk I/O speed and internet I/O speed. This kind of model can not support huge capacity to provide media streaming service.To solve the bottlenecks of C/S model, the P2P network technology which uses unstructured structure, and Gossip protocol to search resources is applied in Video-on-Demand system. In the system, each node acts as not only the server which can provide services to other nodes, but also as client that receives services from other nodes. The nodes in the system connect and exchange data to each other directly, this mechanism can greatly reduce the work load of the server, improve the resource use efficiency, such as storage space, network bandwidth, etc.. Taking into account the fact that the process of data transmitting using Gossip protocol is always blindness, which is not suitable for the large scale of streaming media system, therefore, an architecture which uses a tracker server is adopted. In the client, a list which contains the information of member nodes and partner nodes is used, which focuses on controlling the transmitting of Gossip to make it more accurate and efficient.Against the disadvantages of cache strategy in the existing system, with the user behaviors in different programs and the same program,the video program is divided into segments and slices, and the request of each segment is recorded.Through analyzing the statistical result using data mining techniques, association rules can be obtained. When designing the cache strategy, the association rules can be used to make full use of memory and disk storage space. The memory cache is designed for sequence play mode, while the disk cache is for the browsing model to ensure that the data prefeched by the user can be play with the greatest possibility. This method can improve the hit rate of prefetching, decrease the waiting time of users in the process of browsing and provide a good watching experience for the user.