Research On User-group Clustering Algorithm In CDN-P2P Hybrid Streaming System

CDN-P2 P hybrid distribution network combines the advantages of CDN and P2 P technology, has become the focus of current research. This paper focus on the user-group clustering algorithm in CDN-P2 P hybrid streaming system.Firstly, for the problem of cross-ISP and inter-domain traffic appears in the distribution system resulting in high communication costs, poor communication quality and the backbone pressure, considering the ISP and node location, a new clustering algorithm ClusterBayes based on unsupervised clustering algorithm and Bayesian classifier was proposed. Use the improved K-Medoids clustering and hierarchical clustering algorithm to group the selected representative nodes, and use the Bayesian classifier to group the remaining nodes and the newly added nodes. Then physical closer nodes with good communication performance were divided into the same group, so that the requesting node gets the nearest high-quality access to resources. Finally, digital simulation results show that the algorithm is effective.Then, we applied ClusterBayes algorithm to the server node selection, and for the problem of ignoring the global performance in current server selection algorithms, a global optimum server selection algorithm OptiSS was proposed. We get the final server selection scheme by solving the global traffic function. Finally, digital simulation results show that the algorithm improves the user experience and decreases the global traffic.Finally, the proposed algorithm was applied to streaming media distribution, and the streaming media system model based on the ClusterBayes algorithm and the OptiSS algorithm was constructed.