Research On OpenFlow Based Session Migration Technique For Media Streaming Cloud

In recent years, as the scale of the Internet users expands constantly and mobile terminal devices become more and more popular, large-scale concurrent user access, multiple terminal access methods and high quality of service requirements have became new challenges that streaming media service is facing. But traditional CDN and P2P based service mode is difficult to support terminal switching in service (multiple screen switching) and service switching on the move, as well as its poor scalability.The rise of cloud computing lets people think about shifting streaming media service to clouds form. Convenience, reliability and other advantages of cloud computing provide a good platform to traditional services, but existing cloud computing platforms are lack of corresponding design for streaming media service. And the long sessions and interactivity features of streaming media bring a great challenge to the elastic scheduling of media streaming cloud. Therefore session migration technique which can solve this problem becomes a key technique of media streaming cloud. But traditional techniques based on a middleware or a proxy at the application layer are likely to cause a single point of failure and have poor scalability, and it’s difficult for them to be transparent to users which is the essential requirement of cloud computing. In addition, since the cost of service migration is high, how to optimize the quantities of session migration is a key problem of migration strategy. Aiming at the above three problems, the main content and achievements of this dissertation can be described as follows:1) A basic architecture of streaming media cloud computing platform has been proposed, which aims at the features of streaming media service and the essential characters of cloud computing and combines the ideas of CDN, DHT, MapReduce and distributed monitor.2) A migration method for media streaming cloud has been proposed, which aims at the problems of traditional streaming media service. The method separates the migration function from application layer service with the help of OpenFlow technique in software defined network. For verifying the effect of the method, a simulation system of streaming media cloud has been designed and build. The results of experiments under the system verified the effectiveness of the migration method. 3) A request prediction based migration strategy has been proposed, which aims at the cost of service migration at network layer. The strategy can improve the system capacity while optimize the quantities of session migration. Simulation results show that the strategy can effectively increased the acceptance rate of requests, thus improving the service capability of the system.