基于Hadoop视频转码的任务调度算法

现有的分布式转码系统大多没有考虑到根据计算节点的处理能力分配转码任务量,对转码过程造成影响,导致转码效率降低。针对这一问题,在对现有Max-Min算法的研究基础上,提出一种新的任务调度算法。利用佩奇排名算法评估机器转码性能,输入视频被分成大小不同的段,依据其复杂度和机器的转码能力进行映射。处理能力强的节点将分配到更多的视频块,最终所有计算节点的完成时间得到均衡化。实验结果表明,该算法能够显著提升视频转码效率,在很大程度上改善了云转码系统。

Scheduling Algorithm Based on Video Transcoding Using Hadoop

Allocation videos in accordance with the processing capacity of computing nodes has a significant impact on transcoding performance,which is currently not taken by most existing distributed transeoding systems into account, resulting in degradation of transcoding efficiency.To address this problem,based on the research of existing scheduling algorithm Max-Min,a modified algorithm is proposed.This paper used a PageRank to evaluate the computer performance. Input video sequences are divided into segments, and mapped to multiple computers.Computing nodes with higher handling abilities will get more video segments.Obviously,it would average the finish time of the computers. Experimental results show that the proposed algorithm can achieve performance improvement of transcoding efficiency,and enhance distributed transcoding system to a large extent.