首页 | 本学科首页   官方微博 | 高级检索  
     

基于选择性马尔可夫模型的缓存预取策略
引用本文:蔡伟鸿,肖水,韦岗,熊智,黄敏华. 基于选择性马尔可夫模型的缓存预取策略[J]. 通信学报, 2010, 31(2): 58-66
作者姓名:蔡伟鸿  肖水  韦岗  熊智  黄敏华
作者单位:1. 华南理工大学,电子与信息学院,广东,广州,510006;汕头大学,计算机系,广东,汕头,515063
2. 汕头大学,计算机系,广东,汕头,515063
3. 华南理工大学,电子与信息学院,广东,广州,510006
基金项目:广东省现代信息服务业发展专项基金资助项目,省部产学研合作基金资助项目,广东省自然科学基金博士启动基金资助项目 
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.

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

Cache prefetching strategy based on selective Markov model
CAI Wei-hong,XIAO Shui,WEI Gang,XIONG Zhi,HUANG Min-hua. Cache prefetching strategy based on selective Markov model[J]. Journal on Communications, 2010, 31(2): 58-66
Authors:CAI Wei-hong  XIAO Shui  WEI Gang  XIONG Zhi  HUANG Min-hua
Affiliation:CAI Wei-hong1,2,XIAO Shui2,WEI Gang1,XIONG Zhi2,HUANG Min-hua2(1.School of Electronic & Information Engineering,South China University of Technology,Guangzhou 510006,China,2.Department of Computer,Shantou University,Shantou 515063,China)
Abstract: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.
Keywords:selective Markov model  proxy server  ,latency reduction ratio  cache prefetching
本文献已被 CNKI 万方数据 等数据库收录!
点击此处可从《通信学报》浏览原始摘要信息
点击此处可从《通信学报》下载全文
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号