Research On Virtual Network Construction And Optimization Of Flow Distribution Under The Environment Of Hybrid SDN

Along with the continuous expanding of network size and the complexity of the business, the demand to the routers is increasingly high, which results in the complication of router and the increase of network loads. In order to reduce the workload of routers, Software Defined Networks(shorts for SDN) rises in response to the situation. SDN separates control function from routers which realize the decoupling of controlling and forwarding. But centralized control of SDN brings advantages as well as problems such as the weakness of expandability; at the same time it is hard to realize the unified standard and the fusion with traditional networks. Specific to these problems, researchers come up with the concept of Hybrid SDN. As the transition from traditional networks to SDN, it relieves the load of routers and provides expandability to SDN.Hybrid SDN can be classified in different ways, in this paper we put forward flow-based hybrid SDN according to the variety classes of flow in the networks. There are traditional networks flows and SDN flows co-existence in this flow-based hybrid SDN which create problems of using network resources between different flows. As a result, we need to find out how to control and allocate the flows reasonably.In this paper(1) we first introduce the research background of SDN, analysis the pros and cons it brings to us. We introduce the studies the researcher did about SDN and how to solve problems that SDN faces.(2)We propose the flow-based hybrid SDN structure, all the nodes in the network are unified, but two different kinds of flows(SDN flows and traditional networks flows) exist in the network. The controllers formulate forwarding strategies for SDN flows and traditional networks flows forwarding freely. We analyze the features of flow-based hybrid SDN structure, come up with the problems we need to solve in the structure.(3)In this paper, we allocate flows in two conditions: static network and dynamic network. In static network, we build model based on the thoughts of wardrop balance and bi-level programming according to the interaction of different flows. To be specific, the upper-level of the model is about controllers make choice for SDN in how to choose lane, the result of this influences the decision of lower-level. The lower-level is that traditional networks flows choose lane according to principle of stochastic user equilibrium assignment. After the interaction of flows, we reach the goal of rational utilization of network resources. We use tabu search algorithm to solve bi-level programming problems, and have a good result in the simulation experiment.(4)In dynamic network, there are n OD pairs in where SDN flows and traditional network flows game for the minimum of cost in the network. We use Stackelberg game model to describe the problem, where SDN flows are the leader and traditional flows are the follower. SDN flows make the decision based on the possible decisions that traditional network flows made. After that, traditional network flows observe the decision SDN made and make their own decision to realize their own interests. At last, we reach the equilibrium after the game between two flows. We use genetic algorithm to solve the problem and have a good result in the simulation experiment.