基于统计结果的文件I/O优化

File I/O optimization based on statistics

对目前主流操作系统的文件大小分布和空间占用、网络文件服务环境下的文件读取请求和视频下载点播等典型应用进行了统计分析,总结了对交换文件、小尺寸文件和访问集中文件进行优化的必要性和可行性。基于目前已有的存储器件,提出了基于统计结果的混合加速外存储体系结构及相应的调度算法。实验结果表明,可以缩短I/O响应时间,提高传输率,改善系统性能。

This paper analyzed the following items：the files size distribution and space occupation in the popular operating systems such as Windows and Linux;the block sizes distribution of accesses requests in a network file services environment;the layout of downloading or online playback activities of video clip files on the Internet. It verified the necessity and feasibility of optimizing I/O performance of swap file, small size files and frequently visited files based on the statistics. This paper proposed a new architecture and scheduling algorithm of hybrid external storage system composed of iRAM, solid state disk and zoneddisk. Experimental results show the new architecture can reduce the I/O response time, increase the transmission rate and improve system performance.