| Currently, Peer-to-Peer technology is becoming more and more widely used in the design of systems and networks. The Internet traffic which Peer-to-Peer applications account for had already beaten that which the Web accesses account for and became the largest traffic volume in Internet. Among the various Peer-to-Peer applications, Peer-to-Peer file-sharing system is the most significant and popular Peer-to-Peer applications. Therefore, the study on Peer-to-Peer file-sharing system becomes an important task for the academia, which will not only help us know more details about Peer-to-Peer file-sharing system, but also can guide us in designing new Peer-to-Peer file-sharing system with better efficiency and performance.In this paper, we conduct a measurement study of a widely deployed Peer-to-Peer file-sharing system—Maze, based on its two-month-long log files. We focus on some basic metrics in the Peer-to-Peer file-sharing system, such as the user's behavior, characteristics of shared files and content distribution, following with a quantitive analysis. We find that the index traffic is much heavier than the query traffic. Besides, the index scheme plays an important role in the load balance of system, the accuracy of query and the performance of following content distribution. However, surprisingly the research on the index scheme is very limited and the state-of-art index scheme is not deliberately designed. Our paper conducts a comprehensive study on various designs of the index scheme. We concentrate on the load balance issue and the index quality issue. Afterwards, we propose a DHT based index scheme, which can achieve an even load balance and a high quality index at the same time.Moreover, we further study the index's impact on content distributions. We conduct a real trace driven simulation on there different content distribution strategies. Our result demonstrates a well-designed index scheme contributes a lot to the performance of content distribution. |
PDF Full Text Request