Binary Information Press
Characterizing the topology of large scale P2P network is crucial in understanding the performance of the network and its impact on ISP traffic. However, due to the inherent lack of structure in large P2P networks, it is highly challenging in obtaining the pattern of relations between groups of P2P nodes, especially the clustering of hosts. The densely connected graph only gives the neighboring relationship of individual nodes, but hides the clustering information. In this paper, the authors propose a cluster mining and visualization technique based on spectral method and applied it to trace data from the popular PPLive online TV network.