Snapshots Segmentation And Dynamics Analysis For P2P Streaming Network Topology

With the development of the P2P and multimedia technology, P2P streaming media business has become one of the most popular business applications, which has aroused widespread research interests. Flows the users of streaming media network growing, people get a better performance of resource sharing, but the highly dynamic characteristics of users brings continuous dynamic network structure, which is a challenge to optimize network performance. Grasp the dynamic characteristics of the network topology is important to improve P2P streaming media applications.The first problem of analysis for dynamic topology is how to determine the instantaneous network topology-snapshot. The conventional method is to divide the topologies based on fixed time interval-fixed windows segmentation, the data within a time window is a snapshot. This method cannot follows the dynamical changes of network adaptively. In this paper, we use a variable windows partition method, the degree of topological change is used for a criterion of snapshot segmentation, we take the relatively stable topology as a snapshot-the length of time window is not fixed. Compare with the method of fixed windows segmentation, the snapshot can get a stable network vision using variable window segmentation. In addition, we set weight to each edge based on the appearance of time, setting different thresholds to extract stable topology, for the comparative analysis between different snapshots and stable topology.After snapshots segmentation and stable topology extraction, we carry out comparative analysis of topological features from degree distribution, clustering coefficient, average path length and betweenness, we found that different snapshots and stable topology accord with the power law degree distribution, but not the characteristics of a small world model. Then the relationship of links are analyzed, we find that many links are tend to connect to the new nodes; proportion of links with transitive relationship between all known nodes is large, but small in a signal snapshot, and it is related to nodes' reporting interval. Finally, the paper carried out a analysis of community division and a comparison of the Q value, and found that the stable topology has a higher Q value than snapshots, but they all are in the range of 0.3-0.7.