hTracker: Towards a service provider oriented Peer-to-Peer architecture

Peer-to-Peer (P2P) applications such as BitTorrent occupy actually a dominant part of the Internet traffic. However, they don't take into consideration the importance of the large traffic volumes exchanged between Autonomous Systems (ASs). This constrains Internet Service Providers (ISPs) to restrict BitTorrent connections in a way of decreasing their peering costs. The difficulty in implementing such a mechanism is that most of the P2P applications are distributed and without any real control. We propose an approach which is based on the introduction of a control entity called hTracker. The latter consists in adding a control level in BitTorrent with a new peer selection policy, in order to reduce the download time and as well as the inter-ASs traffic. The principle adopted here is to fraction the traffic by using locality information and to make each peer selects his neighbors relying on the AS he belongs to. We also propose a formal peerId specification calculated using an HMAC function to map each peer with its AS membership. The large-scale simulations show valuable results that validate our solution.