Research Of Video User QoE Optimization Over OpenFlow Network

As streaming media business is one of the main Internet service, how to ensure the quality of experience (QoE) is an important research topic in the academic field, and also a key evaluation index in the industry. It is gradually formed consensus that software defined network (SDN) is an efficient way to improve user QoE of streaming media services, as SDN is intelligence and flexibility in network resource scheduling and traffic management in recent years. Many existing optimization strategies based on the network quality of service (QoS) have been proposed, thus to improve the user video quality; however, the network QoS and user QoE is not equivalent, and the network Qos did not indicate the user intuitive feeling. Sometimes the user QoE is poor while the network QoS is good. Some existing work proposed optimization strategies triggered by the client business requirements, through the client requests to network, and the way to ensure that the user perceived video quality; this approach requires the feedback from the client, both network bandwidth is occupied and the client is generally modified, will bring additional overhead and cost. Dynamic Adaptive Streaming over HTTPUnlike existing work, this paper proposes a new optimization strategy to improve the streaming media user QoE, using OpenFlow network traffic information collection to detect the video player buffer state of client, and to adjust the network resource allocation strategy. As DASH (Dynamic Adaptive Streaming over HTTP) is the research object in this paper. The main work and innovation includes two parts.The first part of the work proposes a client buffer state detection model, the target of which is to infer the buffer filling level of the client player by using the network traffic information simply. The detailed works are:selects the basic feature set by observing and analyzing the relationship of the network flow parameters and the client player buffer length; based on the principle of maximum mutual information and the characteristics of DASH, offers some new feature sets generated from the basic feature set; evaluates and analysis the decision tree model trained by different feature set; tests the decision tree model trained by feature set generated using exponential smoothing.The second part of the work proposes a QoE optimization framework and QoE optimization model. Based on the result of the first part, the framework starts the QoE optimization model to achieve the optimization of all streaming media users. On the basis of maximum minimum fair principle, clarity and continuity are used to evaluate the user QoE in the QoE optimization model of which the optimization goal is maximizing bandwidth of the streaming media user whose experience quality is the worst.