Design And Optimization Of Distributed Video Enhancement Transcoding System

With the rapid development of media services,requirements for high quality video have also gradually increased.Compared with high definition video,high dynamic,wide color gamut,and high definition 4K video is more popular for the audience.In order to satisfy the demand of massive high-quality video,content service providers intend to realize the video enhanced transcoding service by the popular cloud computing frameworks.However,traditional heavyweight distributed frameworks can't apply to the video enhanced transcoding cloud service flexibly.This paper gives the solution of distributed video enhanced transcoding system firstly.The system can be built on private clouds or public clouds to provide services to developers or users through the RESTful API.The system is designed containing three layer,including data storage layer,processing layer and presentation layer.When the system works,these three layers are connected by three flows,including control flow,state flow and data flow.This system supports chain processing,CPU-GPU collaboration and Windows-Linux collaboration.This paper then introduces the optimizations of efficiency,stability and scalability for video enhanced transcoding system.Firstly,a multi-tree-type chain control flow model is proposed,which realizes the functional requirements of the multiple transcoding scene after video enhanced processing.Then slice-merge processing pattern for video processing is proposed.Compared with the intact video processing,sliced video processing can significantly improve the video processing efficiency under the condition of sufficient resources.The data locality pattern for chain processing is implemented in the system.In the experimental system,the data locality pattern can increase the efficiency of 5.93%when the video is under single chain processing.Finally,we package the system components by Docker,which not only achieves the quick deployment,but also provides monitoring and fault tolerance pattern.This system has been successfully applied to the UGC online service platform.In this application,the system is deployed in the private cloud of China Telecom,and provides video enhanced transcoding service.