Research On Traffic Engineering In SDN

In recent years, with the rapid development of the Internet, traffic in the network is increasing, network service needs more and more network resources, Due to the tight coupling between the control and forwarding platform in traditional network equipment, It is difficult to manage the network effectively, traditional Internet architecture becomes more and more difficult to achieve and adapt to the current business needs. Software Defined Network(SDN) arises at the moment, the control plane is separated from the traditional network equipment in SDN, network equipment only responsible for matching entries for data forwarding, control logic is concentrated into the software form of the controller. Through the open north interface provided by the controller, network administrators can dynamic control and manage the network, through a standardized south interface, the manufacturers of network equipment can be compatible. This centralized management architecture of SDN is very suitable for data center network, by SDN controller, traffic in data center network can gain more fine-grained scheduling, which can improve the utilization rate of data center resources and provide QoS guarantee for network service.In this thesis, based on SDN technology, SDN network traffic engineering technology is studied mainly about the load balance and QoS. In this thesis, introduce the research background and the present situation of current domestic and international traffic engineering at first time, after that, describe the SDN architecture and SDN related background, analyze data center network topology and routing status, and then describe the application prospects and advantages of SDN in the data center network. Then, the DMLB algorithm is proposed for the fat-tree topology in the data center network, and implement the algorithm on the SDN controller POX, using Mininet simulation platform to build the fat tree topology and two kinds of traffic model of data center network, through the simulation of these two kinds of models, the algorithm is analyzed and compared. Then, describe the multi-constrained QoS routing problem, solve the multi-constrained QoS routing problem using genetic algorithm,then build a topology with bandwidth, delay and packet loss rate constraint on the Mininet for test, realize genetic algorithm and the network model simulation test in the SDN controller POX, test results show that the genetic algorithm can complete multi-constrained QoS routing target, choose the right path for the traffic. Finally, summarize the work in this thesis,and explain the deficiency of this thesis, prospect the research work in the future.