Broadcasting Algorithms in Peer-to-Peer Data Distribution Systems

Our focus is on the study and development of efficient algorithms for the task of broadcasting within peer-to-peer (P2P) data distribution systems. The purpose of broadcasting is to completely distribute a target set of data pieces from a set of initial sources to all peers within the distribution system. Thus, findings from the study of the problem of broadcasting have useful implications for the task of data delivery and file distribution within these systems. Our work was initially motivated by the desire to develop different data distribution algorithms than those currently employed by popular Internet-based P2P systems. Current P2P file distribution systems such as BitTorrent employ game theoretic motivated algorithms and replication heuristics in the face of distributing files over the Internet to participating peers that are autonomous, self-serving and possibly dishonest. We hope to instead develop efficient data distribution algorithms that would be employed in peer-to-peer networks such as internal or private networks, storage area networks and networks of parallel storage devices where a degree of centralization, coordination and dependability is available. We believe that the task of broadcasting is a fair theoretical abstraction of this problem. Past work in the field has studied certain basic versions of the broadcasting problem extensively. We will instead focus our efforts on Disjoint Multi-Source Broadcast and Arbitrary Multi-Source Broadcast, two variants of broadcasting that were previously studied very little.