Towards a faster BitTorrent

Abstract. BitTorrent is the most popular peer-to-peer system for file sharing. In this protocol, the file is divided into pieces and the clients upload and download the pieces to each other. The tit-for-tat rule of the protocol enforces the cooperation between selfish peers. In the original BitTorrent protocol, sometimes the neighbors of a peer have no piece that the peer does not already have, which inhibits the download. Rare pieces can cause long waiting times. An elegant way of solving these problems is the extension of the original protocol by source coding. This coding method increases the diversity of the pieces in the network which accelerates tit-for-tat piece exchange and leads to faster downloads.
We propose a novel deterministic source coding method. The main advantages of our method compared to state of the art random coding methods are the reduced traffic overhead and the guarantee of decodability of the original file after downloading \(d\) different coded pieces, where \(d\) is the number of pieces of the file. We analyze the deterministic method theoretically. The theoretical results are backed up by simulations.