A locality-based LFH cluster strategy for overlay network

With the popularity of P2P-based file distribution, P2P technology has been successfully applied into media streaming system. Both in Mesh-based 'swarm' method and in Tree-based 'push' method, selecting suitable peers to get or relay data is a critical factor for the system quality, scalability and cost. In this paper, we firstly propose a lightweight locality based Leader-Follower(LF) clustering algorithm to group close node together. Then we construct a hybrid(H) virtual tree/mesh topology based on this clustering procedure. We construct virtual tree topology to push data among clusters and use mesh topology to swarm data in the cluster. Using simulations, we show that our method is robust to dynamic behavior such as peer join and departure, and fundamentally reduce the Inter-AS traffic. Compared with existing topology-aware method, our LF clustering method requires neither dedicated landmarks nor much proactive probe. And our hybrid framework has the advantage of efficiencies associate with Tree-based method and robustness, scalability of Mesh-based method.