基于Hadoop的流媒体转码系统设计  

Design of Streaming Media Transcoding System Based on Hadoop

作　　者：李增禄[1] 张武威[1] 刘孙发[1] LI Zenglu;ZHANG Wuwei;LIU Sunfa(Modem Education Technology Center Sanming University,Sanming Fujian 36500)

机构地区：[1]三明学院现代教育技术中心,福建三明365004

出　　处：《山东农业工程学院学报》2018年第9期31-35,共5页The Journal of Shandong Agriculture and Engineering University

基　　金：福建省中青年教师教育科研项目(JAT160471);福建省中青年教师科研项目(JAT170553)

摘　　要：为了满足移动设备视频流播放需求,设计一套针对HTTP非实时视频流(HLS)与HTTP动态自适性视频流(DASH)的云计算流媒体转码系统。为适配HLS与DASH的标准,每部影片都会切割成较小的视频片段,而且每个视频片段都可能要求支持数种编码格式,以满足视频传输与显示的需求。为了加速视频转码的速度,在本系统中通过Hadoop将所有的视频片段分散储存至所有的节点中,并在每个节点上执行视频转码程序来对储存在节点上的视频流片段进行视频转码。实验测试的结果显示,相较于单一主机的情况,在一个拥有3个节点的Hadoop系统上执行视频转码,平均节省38.64%转码时间。In order to meet the requirements of video streaming for mobile devices, a cloud computing streaming transcoding system for HTTP non-real-time video streaming （HLS） and HTTP dynamic adaptive video streaming （DASH） was designed. To adapt the HLS and DASH standards, each video was cut into smaller video clips, and each video clip may require support for several encoding formats to meet video transmission and display requirements. In order to speed up the video transcoding, all video clips were distributed to all nodes through Hadoop in this system, and a video transcoding program was executed on each node to Transcoding video clips. According to the results of the experimental test, video transcoding was performed on a three-node Hadoop system, which saved an average of 38.64% transcoding time compared to the single-host scenario.

关 键 词：云计算 HADOOP 流媒体转码 

分 类 号：TN948.64[电子电信—信号与信息处理]