Research On QoS-Based Traffic Classification And Multi-Objective Optimization Routing In SDN

As a new network architecture pattern,Software Defined Networking(SDN)can effectively solve the limitations of traditional network in meeting the requirements of today's complex network by decoupling the control plane and the data forwarding plane and centrally controlling the resource allocation in a programmable way.It is the centralized control ability and programmable ability of the controller that lays the foundation for realizing traffic management engineering and routing planning engineering in the network.Up to now,the identification of traffic type in SDN controller is based on To S field in Open Flow protocol.Such traffic identification accuracy is not high and it is difficult to distinguish more and more traffic types.The shortest path algorithm based on hops is used for routing planning.This routing strategy is easy to generate network congestion when the traffic data is large.In order to improve the quality of service(Qo S)in SDN,the traffic classification algorithm and routing planning algorithm in SDN are studied in this paper.The main work contents and innovations are as follows:(1)The research background and significance of SDN are introduced,the research status of traffic classification and multi-path routing in SDN are analyzed emphatically,and content and innovation of this paper are briefly pointed out.Through the detailed introduction of SDN structure framework,convolutional neural network model and multi-objective optimization algorithm,this paper lays a theoretical foundation for the study of traffic classification and routing planning in SDN.(2)Aiming at the traffic classification problem in SDN,an online traffic classification algorithm based on one-dimensional convolutional neural network is proposed.The classification algorithm uses convolution neural network model for traffic classifier.First of all,based on the traditional convolution(Le Net-5)a new network model structure was designed to guarantee the high classification accuracy and reduce the time needed for classification.To more suitable for online traffic classification implementation.Then the open data set is used for offline training of the designed convolutional neural network model,and the optimal model is embedded into the SDN controller.Finally,the data packets are sent into the classifier for traffic service category prediction after online preprocessing by SDN controller,and the total accuracy rate of traffic classification reached 92.13%.(3)Aiming at routing planning problem in SDN network,an adaptive multi-objective optimization routing planning algorithm is proposed.The rapid non dominated sorting,elite strategy and crowded degree of multi-objective optimization genetic algorithm(NSGA?)are introduced to solve the problem of route planning in SDN.Firstly,the controller is used to obtain the residual link bandwidth,link loss rate,link delay and other information to determine the multi-objective optimization function.Then the first generation population is generated by greedy strategy,the first generation population is ordered by non-dominance,the crowding degree operator is used for selection,crossover and mutation,and the elite strategy is introduced to ensure the diversity of the population.Finally,the network state is monitored in real time by the monitoring module to adaptively schedule the multi-objective optimization routing algorithm to prevent network congestion.The simulation results show that the routing planning algorithm is better than the traditional routing algorithm in link utilization,throughput,packet loss rate and delay.