Comparison of Robust Cooperation Strategies for P2P Content Distribution Networks with Multiple Source Download

D. Schlosser, T. Hoßfeld, K. Tutschku
Research Report 385

Abstract

The performance of Peer-to-Peer (P2P) content distribution networks depends highly on the coordination of the peers. Sophisticated cooperation strategies, such as the multiple source download, are the foundation for efficient file exchange. The detailed performance of the strategies are determined by the actual peer characteristics and the peer behavior, such as the number of parallel upload connections, the selfishness, or the altruistic re-distribution of data. The purpose of this work is to evaluate and investigate different cooperation strategies for multiple source download and select the best one for a scenario for even leeching peers, i.e. peers which depart as soon as they have finished their download. The question arises whether the cooperation strategy can smoothen the overall performance degradation caused by a selfish peer behavior. As performance indicator the evolution of the numbers of copies of a chunk and the experienced download times of files is applied. The considered scenarios comprise best-case (altruistic peers) and worst-case scenarios (selfish peers). We further propose a new cooperation strategy to improve the file transfer even when mainly selfish peers are present, the CygPriM (cyclic priority masking) strategy. The strategy allows an efficient P2P based content distribution using ordered chunk delivery with only local information available at a peer.