A Mixture Distribution Based System in BitTorrent-Like P2P Networks

In this paper, we develop a novel file sharing system based on a mixture distribution model working in the BitTorrentlike p2p networks. The BitTorrent's built-in “tit-for-tat” unchoking mechanism delays the initial file sharing process for newly joined peers as well as brings the problem of free-riding that peers only download from others but never contribute to the network. We demonstrate a file sharing mechanism which allows peers to share pieces according to different mixture distributions. The mechanism utilizes the historical contributions of peers in the network to inspire cooperation among peers, Along with the mixture distribution model, the peers can only obtain the whole file by contributing to the network continuously which deters the free-riding behaviors. We theoretically prove that the peers take the truthful revealing as their dominant strategy and our system can speed up the initial process of file sharing. The experiments show that the proposed system performs well and has good scalability, as well as prevents the free-riding problem elegantly.