The Design And Implementation Of General Purpose Stream Media Server Based On HADOOP

In recent years, the3G mobile network and mobile device performance is greatly improved and the demand for wireless video services is keep growing. The traditional streaming media server system usability can’t meet the challenges of the new application scene. Firstly, in heterogeneous networks, multiple protocols are used and the mobile device not only be a player, but also a camera. The traditional media streaming server is implemented for a particular environment and protocol can lack of capability for the general use purpose. Secondly, the current application usually contain massive video content, however the video transcoding operations are extremely complex and the traditional transcoding server lack of the scalabilty to support the computing needs.Combine with research project Nanjing radio and TV group3G censor system, the paper focus in the multimedia stream and video transcoding application, a general purpose multimedia streaming server with flexibility and scalability is designed and implemented, In the meanwhile, the cloud computing program pattern MapReduce is introduced into the video transcoding and give the fundamental solution for the shortage of the computing resource.The paper gave the analysis of the multimedia system, including H.264/AVC codec and typical transmission protocol. The MapReduce parallel computing framawork is introduced and is used in the video transcoding application. Meanwhile the paper discussed the limitation of the traditional tree network topology under the massive cluster environment and a new route algorithm is presented based on the TPL2-SW network. A last, a general purpose multimedia streaming server based on SPSC framework is designed and implemented.The paper has the following contributions:1) Design and implement the distributed multimedia streaming platform, with the flexibility and scalability to meet different workload by adjust the number of the cluster.2) A unified video source and sink interface is presented. The source, process and sink module could implemented by different instance so the unified service is provided for the different terminals.3) A hybrid network concurrency model is presented and the throughput of the system is up to700Mbps, higher than AISO and ACE.4) The cloud computing pattern is introduced into the video transcoding application and the Hadoop framework is used to solve the shortage of the computing resource.5) A new route algorithm for TPL2-S W network topology is presented which take the node work load into the route judge so the load will be more balanced. The paper’s work is deployed in Wuxi Taihu Water quality supervision system and could be extended to ITS and other environment.