基于MapReduce的H.264/AVC并行视频编码

提出一个基于MapReduce的并行视频编码架构, 将源视频切分后以任务的形式分发到不同的处理器上并行地进行编码处理, 以达到提高编码速度的目的。为使整个任务的完成时间最短并平衡负载, 系统综合考虑视频编码特点及处理器的处理能力, 给出LBMM(load balance maximal-minimal complete time)算法。仿真结果显示提出的并行视频编码架构极大地改善了大数据量视频序列的编码效率, 减少了作业的平均响应时间。LBMM与Min-Min算法和CloudSim现有的轮循调度算法相比视频编码速度更快。

H.264/AVC parallel video coding based on MapReduce

This paper proposed a parallel video coding scheme based on MapReduce. It divided the input video sequence into segments and mapped to multiple computers in the form of tasks. After that, it encoded all these tasks in parallel at different computer nodes to achieve better encoding speed. In order to minimize the makespan of a given tasks set and balance the system load, this paper proposed a LBMM algorithm according to the computing capacity and the coding charactertics. The simulated results show the proposed parallel video coding system greatly improves the big volumn video sequence coding efficiency, reduces the average job response time. The LBMM algorithm outperforms Min-Min(minimal-minimal complete time) algorithm and RR(round robin) algorithm using in CloudSim.