Efficient file search in non-DHT P2P networks

Abstract

Unstructured P2P networks dominate in practice due to their small maintenance overhead. However, the high volume of search traffic threatens its continued growth. The focus of this paper is to study how to improve the search efficiency in a non-DHT P2P network without a distributed indexing structure. We identify possible performance problems in KaZaa and Gnutella, and propose a flexible two-phase ticket-based search algorithm (TBS). In particular, the first phase is designed to reduce the search overhead and lookup delay in searching popular (highly replicated) files, and the second phase is designed to reduce the excessive duplicate messages in searching unpopular (rare) files. In addition, a random sampling solution is proposed to estimate timeout between consecutive search rounds, and a moving anchor solution is proposed to reduce duplicate visits of the same node. Moreover, we propose a ticket-based broadcast algorithm (TBA) by slightly modifying TBS, which has the significance in supporting various network functionalities. We evaluate the performance of the new techniques by both analysis and simulations, which demonstrate that the proposed solutions outperform the existing alternatives.