结构化P2P网络访问热点预测与负载均衡机制研究  

Access hotspot prediction and load balancing mechanism based on structured P2P network

作　　者：杜超[1] 徐蕾[1] 

机构地区：[1]沈阳航空航天大学计算机学院,沈阳110136

出　　处：《沈阳航空航天大学学报》2013年第6期38-43,共6页Journal of Shenyang Aerospace University

摘　　要：针对结构化P2P网络的访问热点带来的负载不均衡问题。提出网络中文件访问请求量的预测模型,及热点文件的副本复制策略。实验对比了ARMA、BP和一次指数平滑三种预测模型对文件访问量的预测及进行副本复制获得负载均衡的效果;结果表明,三种模型能够降低接近1/3的文件响应时间,文件访问的成功率提高了10%,三种预测模型中ARMA获得了较好的结果。Aimed at the load imbalance caused by access hotspots in structured P2P network, this paper pro- poses a file access request prediction model and duplication replication strategy of hotspot files. Experiment contrasts the load balancing effects, which is obtained by file access prediction and duplication replication of ARMA, BP and single-exponent smoothness models. Results show that the three models can reduce about 1/ 3 of the response time of files, and improve 10% of the success rate of transaction. In addition, ARMA mod- el is the best of the three prediction models.

关 键 词：对等网络 负载均衡 热点预测 副本复制 CHORD协议 ARMA 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]