内容分发网络中资源感知的服务节点选择问题  

Resource-Aware Server Selection in Content Delivery Networks

作　　者：惠雯[1,2] 林闯[2] 杨扬[1] 

机构地区：[1]北京科技大学计算机与通信工程学院,北京100083 [2]清华大学计算机科学与技术系,北京100084

出　　处：《北京邮电大学学报》2012年第3期103-107,共5页Journal of Beijing University of Posts and Telecommunications

基　　金：国家自然科学基金重点项目(60932003);国家重点基础研究发展计划项目(2010CB328105)

摘　　要：提出了一种在内容分发网络中选取最优服务节点的方法.该方法权衡了服务性能和服务开销,根据实时测量的网络状况和服务节点工作状态,利用网络坐标方法预测服务开销,同时在网络坐标空间中,利用分层聚类思想组织各服务节点,使用户请求能在保证服务性能的前提下,自适应地定位到开销最小的服务节点.实验结果表明,该方法可有效提高资源定位准确度,减少服务开销且具有较好的时间特性.In order to solve the problem of resource location in content delivery networks, a new algo- rithm for selecting optimal service nodes is proposed. The algorithm provides a good solution for both service performance and overhead. First, by taking advantage of the network coordinates approach, the computation and communication overhead of different service nodes are computed depending on network and server status change. Then, based on the hierarchical clustering algorithm, all the service nodes are reorganized in the coordinate space. Finally, the user requests are adaptively redirected to the service nodes with minimum overhead without sacrificing the service performance. Experiments verified that the precision of resource location is improved and the overhead is reduced with acceptable time performance by the new algorithm.

关 键 词：内容分发网络 资源定位 分层聚类 网络坐标 

分 类 号：TN393[电子电信—物理电子学]