Analyze the Impact of User Search Behavior on DHT-based P2P File Sharing System

DHT has been proposed in the literature as a general infrastructure for building large scale distributed system. While DHT supports exact key search inherently, several arguments against it contend that DHT cannot support the keyword search well, which is widely used in real systems. Through study of search logs from a live peer-to-peer file-sharing system, this paper presents the query characteristics in the real system. In contrast to previous studies, we focus on analyzing the potential impact of keyword-based search pattern on the system performance, by which we demonstrate the real potential performance problem. Though unbalanced distribution of query traffic, we find that it actually can be handled easily by a moderate PC. However, the index process for shared files will cause unbalanced large traffic in DHT-based systems, which should not be neglected by future designer of peer-to-peer file-sharing system