为了实现能够快速高效的使用计算集群解决视频计算问题,提出一种基于Celery的分布式视频处理框架,该框架借鉴了Hadoop的架构设计,提出了API服务器层,JobTracker和TaskTracker的三层架构,并对其进行了优化使之能够兼容计算资源与存储资源的横向扩展和新的计算应用的纵向扩展。详细描述了框架的架构设计和优化设计。实验数据证明分布式计算框架能够高效地利用多节点实现视频计算。
In view of using computer cluster to solve video processing problem, a distributed video processing system based on celery is proposed in this paper. Using Hadoop' s structure as a reference, a three-tier structure is presented, including API server layer, JobTraeker layer and TaskTracker layer. This structure enables the system to be compatible with both computing resource and functional extension. In this paper, structure of the system is presented in detail together with optimization. The experimental data indicate that the system is proved to be efficient with muhiple calculating nodes.