A probability caching model in hybrid P2P information dissemination

P2P (peer-to-peer), which means resource transferring between clients, is a new technology applied in file sharing, instant message and media streaming. Nowadays it´s more and more popular for it supplies a low cost method for people to gain resources. Since its born P2P makes progress towards fully distributed model or hybrid model from central model. Hybrid model P2P integrates distributed and central features together, containing two kinds of peers: super-peers and leaf-peers. Super-peers act as servers and leaf-peers act as clients. Only super-peers construct a fully distributed P2P overlay. One problem is how to distribute resources in hybrid P2P overlay. Uneven distribution of resources will cause loads focus on a few super-peers and thus resulting in hot-spots, lowering the performance of P2P. It´s necessary to find a new policy to cache data evenly in hybrid P2P overlay. In this paper, we propose a probability caching policy for hybrid P2P information dissemination. The policy determines the target caching peer during forwarding process in order to gain even probability for data distribution. Thus the P2P overlay can achieve great performance in load balancing and simulations results support our assumption.