一种基于聚类算法的P2P流媒体服务平台可视化监控系统的设计与实现  

Design and Implementation of Visualized Monitoring System Based on Clustering Algorithm for P2P Streaming Services

作　　者：戴瑾[1] 谭良良[2] 王钦辉[2] 袁征帆[1] 叶保留[2] 

机构地区：[1]南京大学金陵学院,江苏南京210089 [2]南京大学计算机软件新技术国家重点实验室,江苏南京210049

出　　处：《微电子学与计算机》2012年第12期184-188,共5页Microelectronics & Computer

基　　金：国家自然科学基金项目(60903025);江苏省科技支撑计划(BE2010179)

摘　　要：基于P2P技术的流媒体服务技术已得到广泛应用.然而P2P流媒体服务系统中拓扑结构的动态性与用户行为的自主性使得流媒体服务质量面临挑战.本文以南京大学研制的交互式P2P流媒体服务平台NJUVoD为背景,基于协同过滤机制及聚类算法,设计了一个面向P2P流媒体服务平台的可视化Web监控系统.实现了对用户行为、拓扑结构、系统状态的在线分析,为了解系统运行状况、优化系统性能提供了决策依据.With the advances of Peer-to-Peer （P2P） technology, streaming service was widely used over the Internet. However, the dynamics of overlay topology as well as the spontaneous behavior of users make it ehallengeable to improve the service quality. To address the above challenges, we design and implement a web monitoring system based on NJUVoD, an interactive P2P streaming service platform developed by Nanjing University. By using collaborative filtering mechanism and clustering algorithm, the web monitoring system could analyze the user behavior, topology structure and system status on-the-fly. In addition, it provides a tool for understanding the runtime status and optimizing the system performance.

关 键 词：P2P流媒体系统 聚类算法 缩放算法 系统性能分析 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]