基于选择性马尔可夫模型的缓存预取策略  

Cache prefetching strategy based on selective Markov model

作　　者：蔡伟鸿[1,2] 肖水[2] 韦岗[1] 熊智[2] 黄敏华[2] 

机构地区：[1]华南理工大学电子与信息学院,广东广州510006 [2]汕头大学计算机系,广东汕头515063

出　　处：《通信学报》2010年第2期58-66,共9页Journal on Communications

基　　金：广东省现代信息服务业发展专项基金资助项目(GDIID2008IS046);省部产学研合作基金资助项目(2009B090300345);广东省自然科学基金博士启动基金资助项目(07301411)~~

摘　　要：通过分析研究现有流媒体缓存管理算法和用户的访问行为特征,提出了一种新的基于选择性马尔可夫模型的缓存预取策略。该策略通过序列合并方法对用户访问拖曳行为进行建模,采用状态剪枝优化方法FP_Vlike得到选择性马尔可夫模型FPMM_Vlike,并在此之上结合替换算法LRU-2构建出一种流媒体代理服务器缓存预取机制FPVlike_LRU-2。仿真结果表明,在访问延时降低量方面,FPVlike_LRU-2要比FP_LRU-2、SP_LRU-2、LRU-2分别高出10%、12%、17%,且在最佳的情况下该值能够达到60%以上。Through analyzing the existing streaming media cache management algorithm and user＇s watching behavior characteristics, a new cache prefetching strategy based on selective Markov model was presented. The strategy, by mod- eling the user＇s VCR action of choosing the merging sequence method, applied the FP_Vlike method to get the selective Markov model FPMM_Vlike and built a streaming media proxy cache prefetching mechanism FP_Vlike-LRU-2 by com- bining the replacement algorithm LRU-2.The experimental results show that, FP_Vlike-LRU-2 is 10%, 12%, 17% higher than FP_LRU-2, SP_LRU-2 and LRU-2 respectively in reducing latency experienced by users, and this value is able to reach over 60% in the ideal situation.

关 键 词：选择性马尔可夫模型 代理服务器 延时降低量 缓存预取 

分 类 号：TN915[电子电信—通信与信息系统]