Using Proximity Information between BitTorrent Peers: An Extensive Study of Effects on Internet Traffic Distribution

Peter Danielis, Jan Skodzik, Jens Rohrbeck, Vlado Altmann, Dirk Timmermann, Thomas Bahls, Daniel Duchow
2011 International Journal on Advances in Systems and Measurements   unpublished
Peer-to-Peer file sharing generates by far the most Internet traffic reaching up to 70 % in some regions of the world. These data volumes pose a significant challenge to In-ternet Service Providers regarding traffic engineering. Because Peer-to-Peer routing is usually agnostic of the underlying topol-ogy, traffic engineering abilities of Internet Service Providers are inhibited and their core networks are overburdened with Peer-to-Peer data. To disburden Internet Service Providers' core
more » ... , a new algorithm for the BitTorrent protocol is proposed in order to improve peer selection. BitTorrent users are provided with accurate information on the hop counts to other BitTorrent users to select physically proximate users. Thereby, the initial Time-To-Live value of outgoing IP packets is copied and inserted as part of the BitTorrent payload. At the packet's destination, the hop count is calculated as the difference between the copied Time-To-Live value and the Time-To-Live value of the IP header. Simulation results for standard and modified BitTorrent implementation are presented highlighting beneficial effects on both the traffic of Internet Service Providers' core networks and BitTorrent users' download performance. The extensive exploration of impacts on Internet traffic includes the observation of traffic loads and peaks in the core networks as well as the number of dropped packets due to congestions. Moreover, as realistic simulations require the consideration of the BitTorrent users' behavior of continuously leaving and entering the BitTorrent network, this article elaborates on the dynamic nature of peers.