基于软件方法提升流媒体服务器性能  

Improving the Performance of Streaming Server based on Software Methods

作　　者：伍时扬 王功乾[1] WU Shi-yang;WANG Gong-qian(FiberHome Telecommunication Technologies Co.,Ltd.,Wuhan 430074,China)

机构地区：[1]烽火通信科技股份有限公司,武汉430074

出　　处：《光通信研究》2020年第6期5-7,24,共4页Study on Optical Communications

基　　金：国家重点研发计划资助项目(2017YFB1401300)。

摘　　要：针对流媒体服务节点硬件成本高、流媒体服务器硬件能力未被充分发掘利用的问题,文章提出了优化流媒体服务器性能的软件方法。首先,基于性能测试分析的结果,流媒体服务器性能瓶颈在于硬盘组负载不均衡,硬盘组的整体吞吐能力未被充分利用;其次,分析了Linux文件系统输入/输出(I/O)访问模式,提出了直接I/O与异步I/O相结合,加速文件读取的方法;然后,针对流媒体文件的特点,提出了分片存储技术,以支持并发读取,均衡硬盘组的负荷;最后,提出了基于内容热度的动态副本策略,以支持热点内容巨大的并发访问。经测试验证,采用这些技术优化流媒体应用软件后可大幅提升流媒体服务器的性能。Aiming at the problems of high hardware cost of streaming service node and insufficient development and utilization of streaming server hardware capability,we propose a software method to optimize the performance of streaming servers.Firstly,based on the results of performance testing and analysis,the performance bottleneck of streaming server is that the load of disk groups is unbalanced,and the overall throughput of disk groups is not fully utilized.Secondly,the Input/Output(I/O)access method of the Linux file system is analyzed,and a method combining direct I/O and asynchronous I/O is proposed to improve the file reading speed.Thirdly,according to the characteristics of streaming media files,shard storage technology is proposed to support concurrent reading and balance the load of hard disk groups.Finally,a dynamic replication strategy based on content popularity is proposed to support massive concurrent access to hot contents.After testing and verification,using these technologies to optimize streaming application software can greatly improve the performance of streaming servers.

关 键 词：流媒体服务器 直接I/O 异步I/O 分片存储 动态多副本 

分 类 号：TP37[自动化与计算机技术—计算机系统结构]