摘要
网络带宽和磁盘I/O带宽是视频点播系统的两大瓶颈.为了缓解磁盘I/O带宽问题,本文引入统计窗概念,提出自适应统计窗缓存算法,采用周期性缓存决策方法管理缓存的流媒体数据,同时采用区分优先级缓存策略和冗余释放策略提高算法性能.采用实际点播数据进行的仿真研究表明:自适应统计窗缓存算法的性能优于定长分段、指数分段和自适应分段算法,特别是在VCR操作时,本算法的性能更佳.
Network bandwidth and disk I/O are the two biggest bottlenecks in VOD system. Using memory to cache media data can reduce the dick traffic greatly. In the paper, we propose a novel algorithm： adaptive statistical window caching algorithm, which uses caching decision to manager the cached media data. Our algorithm also adopts two novel strategies： priority caching and redundant release, to achieve better performance. The proposed method is evaluated by simulations using traces from one actual VOD server. Simulation results indicate that our proposed method is better than the uniform segmentation algorithm, the exponential segmentation algorithm and the adaptive and lazy segmentation algorithm, especially in VCR situation.
出处
《小型微型计算机系统》
CSCD
北大核心
2009年第2期209-214,共6页
Journal of Chinese Computer Systems
基金
安徽省优秀青年科技基金项目(04048046)资助
新世纪优秀人才支持计划项目(NCET-04-0564)资助
国家"八六三"计划项目(2006AA01Z114)资助
关键词
分段缓存
缓存决策
区分优先级缓存
冗余释放
segment-based caching
caching decision
priority caching
redundant release
