Application Layer Traffic Optimization For BitTorrent-Like P2P File Sharing Systems

Recently, with the popularity of the internet, Peer-to-Peer network technology grows rapidly. The P2P systems are obtainning more and more internet users'welcome for their convenience, flexibility and low-cost.File sharing system is one of the most widely application of P2P technology. Its traffic represents about 70% of all internet traffic. The characteristics of P2P file sharing system is that each user not only downloads others' content, but also uploads to others. They serve other peers and other peers serve them too.BitTorrent is a representative P2P file sharing system. It is different from former P2P file sharing systems that each file is divided to many pieces in BitTorrent. A peer can download many pieces at same time. A peer can serve others just after finished only one piece, but not after finished the whole file.However, BitTorrent-like P2P file sharing systems don't consider locality and inter-ISP links on the underlying network topology. Their schemes of choosing neighbors and selecting pieces are based on overlay network toplogy, but not consider the traffic brought by inter-ISP data transferring on key links, which may lead to performance decrease of the whole network.Recently, more and more people begain to concern that localizing the traffic of P2P file sharing systems by using Application-Layer Traffic Optimization approach. The ALTO approach includes two points:how to best provide the topology of the underlying network and how to allow the requesting node to use such information to effectively reach the node on which the content resides.Therefore, for improving the performance of BitTorren-like P2P file sharing systems by using ALTO approach, firstly, this article proposes a novel approach to let BitTorrent obtains the topology of underlying autonomous system, then modifies the algorithms of choosing neighbors, selecting pieces and choking/unchoking based on the AS topology. Finally, this article proposes a cache approach based on the modified algorithms for furtherly optimizing application traffic of BitTorrent.Finally, we prove the correctness of our scheme above in some simple simulation scenarios by using P2P simulator GPS which support simulating underlying networks. Then we compare our scheme with tradional BitTorrent, and show its high performance and adaptation for the comparison. So it is absolutely correct of our proposition that to improve the performance of BitTorrent-like P2P file sharing systems by using ALTO approach.