Traffic Engineering On Hybird-SDN Architecture

SDN network is a new network architecture, that separates data platform and control platform. Control layer disposed SDN controller for centralized control of network traffic distribution. Management complexity of traditional routers is simplified and the device on the data layer is only responsible for forwarding data, so that the network architecture breaks the closure of traditional IP network. More diversified services provided by the network. The network architecture flexibly configures and uses network resources, load balancing, and improving network performance.In recent years, because of many advantages on SDN network, some companies deploy SDN network to its own data center and form hybrid SDN network based on the traditional networks. However, hybrid SDN development in its infancy, there is no uniform blend modes and routing mechanisms. The purpose of deployment of SDN nodes is to improve network performance, so this paper has an in-depth analysis and research on traffic engineering issues of hybrid SDN architecture. First, the paper describes the architecture and features of SDN network, and analyzes the current status of research on SDN / hybrid SDN traffic engineering. It also summarizes the existing hybrid SDN network model and algorithm for traffic engineering, and deeply analyzes its problems and shortcomings. Then hybrid SDN architecture based on node and routing mechanism is proposed. In order to meet that overall network delay and total delay of a single service are minimal, this paper proposes hybrid SDN traffic engineering algorithm based on Stackelberg Game. On this basis, in order to improve traffic engineering and find the best position to SDN nodes, the optimal deployment algorithm of SDN node is proposed based on bi-level programming. In summary, the main work is as follows:(1) This paper introduces the basic structure of hybrid SDN network, describes several possible hybrid network modes and application scenarios. With diagrams it analyzes the transition from the traditional network to SDN deployment scenarios, classifies existing hybrid SDN traffic engineering, and has a detailed description of the basic idea of existing algorithms.(2) Hybrid SDN architecture model exists two issues. One is that different routing mechanisms need coordination, the other is that granularity of control on SDN controller is difficult to grasp. The existing traffic engineering algorithms lack of global optimality, such as Min-Max link utilization, ECMP algorithm. To solve these problems, this paper use Stackelberg game model and Wardrop equilibrium theory to propose hybrid SDN network model based on hybrid nodes, which need to select the part nodes deployed as SDN nodes and design destination-based routing mechanism that will divide the process of generate the shortest path tree into non-SDN network routing process and SDN routing process. At the same time the interactive routing process of different mechanisms is described as Stackelberg game process. We set up routing policies for their game strategy. SDN controller is seen the total cost of the entire network system as the objective, non-SDN routing node is seen the cost of a single traffic flow as the objective. Mathematical model is built. Through interactive iteration and coordination between the two routings, the solution achieves global optimization when meets the standards of equilibrium. Simulation results show that LB-SG algorithm can significantly reduce the total cost of the network. In the larger network and more traffic demands, load balancing of the algorithm is more obvious.(3) In order to ensure optimal hybrid SDN traffic engineering, we analyze and design SDN nodes deployment algorithm. The algorithm takes into account the overall network costs of the minimum and the path of non-SDN services with minimal latency. Thus bi-level programming model is built. It overcomes the local optimal defect in single programming. This algorithm deploys SDN nodes from the perspective of bi-level decision-making. In the case of network operators to invest determined that the number of SDN node, the optimal algorithm guarantees the optimality of SDN node deployment location. Experimental results show that, compared with the traditional algorithm, this deployment scenario effectively improves network performance and reduces overall network costs.