Research On Service-flow Control Of Software Defined Network Towards QoE-driven

With the rapid expansion of the network scale and the emergence of the new services, the demand for service traffic is increasing. The complexity of network structure, low resources utilization and the diversity of user requirements lead to low user satisfaction. The network control is becoming difficulty, and the network is difficult to quickly deploy new capabilities. While the traditional network service-flow control used by admission control, traffic control and other control techniques have been difficult to play its role effectively, and also difficult to meet the E2 E users’ QoS requirements. However, software defined networking decoupled the data and control planes, which both can achieve global optimization and high-performance network forwarding capability, its open programmable interface provides a new solution for network service-flow control. Based on the analysis of traditional network service-flow control, using QoE as a services quality metrics, the paper does researches about service-flow control of software defined network towards QoE-driven. The main work is as follows:1. The model of SDN service-flow control towards QoE-driven is proposed. The model is based on the SDN environment, using self-service negotiation and optimization mechanism in SDN application layer in order to get the best service configuration. SDN control layer provides network service-flow transmission path according to the agreed service configuration, in order to meet the needs of individual users.2. The method of SDN multi-service-flow global optimization configuration based on selfnegotiation mechanism is proposed. This method adopts improved self-service negotiation and optimization mechanism for the limited network resource conditions, establishing a mathematical model based on MOS, computing the optimal utility value of network resources, and optimizing global utility function by using iterative MOS increase and adapted greedy algorithms, to obtain a set of overall user perceived quality of service which has near optimal service configuration. Finally, simulation results show the effectiveness of this method.3. The method of SDN multi-service-flow path optimization control based on the double reinforcement learning strategy is proposed. This method together with QoE-aware and reinforcement learning strategies, proposes a novel QoE-aware reinforcement learning strategies, calculating dynamically the most efficient path for each service level in order to transmit service-flow for each service type according to specific QoE requirements. Simulation results show that the proposed method can ensure the E2 E user QoE requirements and overall performance of networks.