Design And Implementation Of Video Transcoding System Based On Hadoop Platform

With the rapid development of the current mobile Internet technology, traditional radio and TV industry gradually develops towards the field of new media, and they need provide videos to mobile TV, the Internet, network television, and other business platform at the same time. However, different business platforms have different requirements on video file formats, and the most common parameters which have different requirements include coding format, resolution and bitrate, etc. To meet this requirement, radio and TV industry must converse the original video file format. The current commonly-used traditional video transcoding system generally adopts one or several machines with high performance transcoding unified store video files offline, but the transcoding scheme has some disadvantages, including limited processing ability, one format each time and longer time consumption for large quantities of transcoding task. Therefore, a flexible and efficient transcoding scheme is urgently needed to meet the demand of current transcoding in radio and TV industry. At present, cloud technology has obtained unceasing development and has become mature. It has advantages of strong parallel processing ability, flexible scalability and low cost, so it will be widely used in the future.In this paper, the transcoding system of the whole network of Hangzhou TV Station and the problems are first described, and the method of using Hadoop to deal with video transcoding task is put forward. Then, the characteristics of core technology of Hadoop---HDFS and MapReduce are analyzed; combining with the video transcoding technology and the principles of file encapsulation, shard strategy based on GOP groups is put forward; the original video file is split into several video clips and stored on HDFS; MapReduce programming principle is used to do distributed transcoding of video subdivision; the transcoding system is flexibly configured and managed with the method of workflow. Thus, the cluster parallel computing ability is fully used to solve the bottlenecks problems of single machine performance, and at the same time, the need of transcoding and generation of various format video files in Hangzhou TV Station is well satisfied.Finally, based on the test of distributed transcoding system, the paper verifies that Hadoop-based video transcoding system can effectively reduce the time of processing video large file, and computing power and storage capacity of the system can freely expand with the growing demand for the transcoding. The system effectively improves the transcoding efficiency of video show of Hangzhou TV Station. It has very high practical significance at the time when HD video will be popularized.