| In recent years,with the continuous progress of network technology and high-definition video technology,the communication network is also evolving,and various video sharing platforms are becoming more and more perfect.Online video technology has made rapid progress.The amount of online video VOD platform corresponding also becomes more and more,and the different requirements for the format of the video on demand platform is often different,which leads to the change of some parameters such as resolution,frame rate and encoding standard,and so on in the massive video transcoding has become increasingly difficult.The traditional transcoding software often only for transcoding some specific format,this way can not handle the volume of data,the number of transcoding tasks,and can not be personalized video,video transcoding efficiency is relatively low.When faced with massive transcoding needs,the traditional data transcoding system can not be transcoding in a timely manner,which requires upgrading the existing system to meet the current needs.This paper introduces the design and implementation of a new distributed transcoding system instead of the existing centralized transcoding system based on the distributed platform.Task scheduling for equilibrium problems and possible data blocks,this paper presents the design of a load balancing strategy based on Gearman,through the management station for task scheduling,tasks are assigned to different workstations for specific treatment.The station contains a PreTreater preprocessing module and JobTracker task scheduling module,the PreTreater module is configured by pipeline way to allocate resources,according to the number of each resource obtained by pipeline pretreatment on the task,the task allocation for JobTracker,and the JobTracker module is received on the task information,do a balanced allocation of queue length and server performance,which is divided and distributed to each workstation.Then the workstation is responsible for segmentation,merging and transcoding of the received video,this structure give full play to the characteristics of parallel processing workstation,and then realize the transcoding operation of massive video,greatly improving the efficiency of transcoding.Based on the analysis of the principle of distributed file storage system design and distributed computing,knowledge and video codec,how to solve the problem in a distributed platform for video coding and decoding operation and data storage.According to the principle of transcoding,a unified and standard communication interface is designed in this paper.At the same time this paper also established a fault tolerance mechanism of data transmission failure and runtime error based on the heartbeat detection,query operation,Job_Id,to make different treatment of communication errors and possible errors in the transmission station respectively,greatly improve the stability and the rate of fault tolerant system.Finally,this paper does some test and analysis of the distributed transcoding system,and the test data show that the transcoding system built on the distributed platform has achieved good results.In small batch processing of video,performance advantage of this system is obviously,when dealing with large quantities of video,the performance of the system compared with other systems,greatly reducing the time required for video transcoding;at the same time,this system has very high reliability and stability,the work load is spread to many machines,which greatly improves the efficiency and fault tolerance mechanism of transcoding;there,greatly reduce the error rate,while improving the efficiency of transcoding provides reliable technical support for users. |