Research On The Traffic Matrix Estimation And Routing Optimization Of Hybrid SDN

Software defined network(SDN) is an emerging network architecture that has drawn the most attention of academics and industry in recent years. The key idea of SDN is separating control plane and forwarding plane to simplify the management of network, the deployment of application and so on. Affected by investment protection,risk control and other factors, the full deployment of SDN will not be finished in the short term,and result into a coexistence state of traditional IP network and SDN which is named hybrid SDN.As the-intermediate state of the fully—deployment-of SDN,. hybird SDN introduces part advantage of SDN into traditional IP network, and the same time, was restricted by traditional IP network’s inherent drawbacks.So,how to make fully use of SDN’s advantages to optimize traditional IP network becomes the hot point on current research. In the current traffic engineering of hybrid SDN, traffic matrix is usually assumed to be known. But in fact,it is hard to measure it directly. Existing estimation methods are mainly for the traditional IP networks, not fully applicable to hybrid SDN.Route optimization in current hybrid SDN traffic engineering does not concern the limitation of SDN nodes.This might exceed the service limit when blindly seeking optimal network performance. Aiming at the above problems in hybrid SDN traffic engineering, the main content and achievements of this dissertation can be described as follows:(1) An algorithm named adaptive multi-Elman neural network(AMElman) algorithm which based on Origin-Destination(OD) flow clustering has been proposed.This algorithm can accurately estimate hybrid SDN’s traffic matrix. By using the Elman neural network model,this algorithm can avoid the influence of prior information on estimation process. Meanwhile,because of the precise measurement of the OD flow by SDN node, the algorithm’s adaptive can be improved without introducing additional measurement cost.(2) A route optimization model based on the sum of link cost and a route optimization algorithm named minimal cost sum(MCS) have been proposed. From an overall viewpoint of network,this model is used to minimize the sum of link cost.The actual processing capacity of SDN nodes were given as the restriction conditions. This model can realize the entire network link load balancing optimization and overcome many limitations of existing models.(3) A series of simulation experiments are performed using the real data of Abilene. Experimental results show that AMElman has higher estimation accuracy and better adaptability than existing estimation algorithms and MCS makes the link load more balanced. Also, an experimental analysis is conducted to verify the application result of traffic matrix estimation in the process of route optimization. The result shows that using the traffic matrix estimation based on AMElman to optimize the route is practical.