流量统计窗周期缓存算法仿真研究  

Caching Algorithm Simulation Based on Flow Statistical Window

作　　者：胡忠[1] 李俊[1] 陈珍敬[1] 钱斌[1] 

机构地区：[1]中国科学技术大学网络传播系统与控制联合实验室自动化系,合肥230027

出　　处：《系统仿真学报》2010年第7期1741-1745,1751,共6页Journal of System Simulation

基　　金：国家科技支撑计划(2008BAH28B04);安徽省高校自然科学研究重点项目(KJ2008A106)

摘　　要：传统VOD流媒体服务器点播只考虑用户从电影文件起始位置的连续播放,没有考虑用户快进快退暂停的行为模式。在VCR操作的情况下,如何有效解决磁盘I/O的带宽瓶颈提高字节命中率是一个亟待解决的问题。将通过对USTC影视日志的分析跟踪每部影片内部流行度。同时引入决策统计窗的概念并且提出周期缓存算法。周期缓存算法采用固定的用户访问流量作为统计窗口,并且以间断缓存作为新的缓存接入策略。采用实际点播数据的仿真研究表明:在VCR操作下,统计窗缓存算法的性能优于传统的定长分段算法、指数分段算法。Traditional study of Video-on-Demand（VOD） streaming media servers supposed that all users watched videos continuously from the beginning of the files,but left out other users behaviors such as fast-forward,fast-backward,random access and pause.Actually,under the circumstance with VCR operations,how to effectively eliminate bottleneck of disk I/O bandwidth and raise the bytes hit rate has become an urgent problem.Based on the analysis of USTC video logs,the internal popularity value of each video is tracked,the concept of statistical window is introduced and a periodical caching algorithm is brought forward.This algorithm employed a fixed traffic flow as statistical window,and adopted intermittent caching as new caching access policy.Simulation results with practical data show that：performance of the statistical window caching algorithm is superior to that of the fixed segmentation algorithm and the exponential segmentation algorithm,especially under VCR operations.

关 键 词：VCR 统计窗 分段缓存 间断缓存 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]