ISP Friendly Peer Selection in BitTorrent

Abstract

Peer-to-peer (P2P) applications provides services such as content sharing, videoon- demand, voice-over-IP which are very popular among Internet users and it accounts to a significant amount of global internet traffic. Since P2P systems uses overlay networks on existing Internet topology it poses significant new challenges to Internet Service Providers (ISPs). In order to reduce operational costs ISPs are throttling or blocking P2P traffic which is unfavourable for a majority of internet users. Research community has proposed several approaches but they breaks the existing protocol and software thus they have not been used widely. In this research, we design and evaluate an approach of reducing these costs using BitTorrent P2P application. Our approach involves of using knowledge of network paths that can be gathered in trackers and using it to bias the peer selection without any modification to any client software. Using the results of an evaluation of nearly 300 nodes in PlanetLab testbed the proposed approach will reduce cross- ISP traffic by 44% while improving in download speed (40%) and upload speed (39%).