An Efficient Method of Multimedia Materials Retrieval

Some of the traditional application on the Internet is experiencing the phase of scale expansion and sharp increase in traffic. In addition to expanding the capability of the server and building server clusters, making full use of P2P to solve the problems of bottlenecks, single point of failure and other issues is also a relatively new method. Based on the careful study of multimedia materials retrieval behaviors, we proposed a method of catalog information retrieval with relevant feedback. By constructing P2P overlay network, physical neighboring nodes are partitioned in one region. According to properties of the node, such as usage of CPU, free memory, available bandwidth and online time, those nodes with good capacity and high reputation are selected to be supernodes. Supernode is responsible for regional management and consistency maintenance of catalog file. Most of the user´s search request can be met in local catalog file directly, thereby our method reduces the retrieval server traffic greatly and improves the efficiency, it also has good scalability.