基于云计算框架的视频转码系统

VIDEO TRANSCODING SYSTEM BASED ON CLOUD COMPUTING FRAMEWORK

随着视频服务和高清视频的普及,视频服务提供商正面临着越来越多的高清视频转码需求。但是由于高清视频分辨率高、码率大,其转码复杂度是标清视频的数倍,而常规的转码系统虽然利用了多核并发来提高转码速度,但终究受限于单个计算节点的物理条件和转码算法本身的并发能力,使得转码速度提升空间有限。提出一个利用云计算的"Map-Reduce"计算框架的转码系统,将单个转码任务并发至多个计算节点,每个计算节点对同一个视频文件的不同部分分别进行转码,再将转码后的视频文件合并,从而显著提升了转码速度。

With the popularisation of video services and high definition（HD） videos,the video service providers are encountering growing needs of HD videos transcoding.Because of the high resolution and bit rate of HD video,the complexity of HD video transcoding is several times of that of the standard definition（SD） video＇s.Although the general transcoding systems have utilised the multiple core concurrency to increase the transcoding speed,it is still limited by the physical condition of single computing node and the concurrency ability of transcoding algorithm itself,which results in the limited increasing room of transcoding speed.In this paper,a transcoding system using the ＂Map-Reduce＂ computation framework of cloud computing is proposed.It assigns the single transcoding task to several computing nodes concurrently.Each computing node transcodes different parts of the same video file separately,and then each piece of the transcoded video will be merged afterwards,which significantly increases the transcoding speed.