对等用户参与P2P(peer-to-peer)文件共享应用的自由性,影响着该类系统的可用性,作为国内教育网上Maze系统的开发者,试图利用收集到的系统目志深入分析Maze用户特性,发现影响资源可用性的关键点,以指导Maze系统的演进.从用户需求的角度重新定义了P2P文件共享系统可用性的概念,并结合Maze系统目志,率先采用聚类技术对P2P文件共享系统的用户进行了量化分类,且深入研究了占用户总数大约0.77%的活跃型用户对Maze系统可用性的影响.发现活跃型用户具有服务器性质,可大幅提升系统的可用性,是改进P2P文件共享系统设计可利用的资源。
The availability of a P2P (peer-to-peer) file sharing system is heavily affected by the high churn rate oi users. When the largest P2P file sharing system Maze in CERNET is developed, the log collected in the system is used to get a better understanding of the users' characteristics, to find the key factor which influences the resource availability, and to instruct the future development of Maze system. In this paper, the concept of P2P file sharing systems' availability is redefined from users' perspective. With the log of Maze, it is the first study to use clustering technique to quantitatively categorize users in a P2P file sharing system. Based on the thorough study of behavior of the active users amounting to 0.77% of the total users and the impact on Maze's availability, this paper concludes that this kind of users plays a similar role to server so that they greatly increase the availability of system and are feasible resource to improve the performance of P2P file sharing systems.