Research And Implementation Of QoS Routing Technology Based On Traffic Classification And Reinforcement Learning In SDN

With the continuous development of network technology,the proliferation of diversified new network applications has put forward differentiated requirements for indicators such as latency,packet loss rate and link utilization in network data transmission,and how to formulate reasonable routes for different scenarios and different application types based on quality of service(Quality of Service,QoS)requirements has become an urgent problem to be solved.Traditional network architectures have problems such as poor flexibility and limited global network views,and it is difficult to guarantee QoS for different applications in a timely and flexible manner.Therefore,based on the programmable and easy-scalability characteristics of Software Defined Networking(SDN),this thesis uses traffic classification technology to classify different data flows,and then constructs multiple constraints according to the QoS requirements of different data flow alienation,and finally formulates routing strategies based on reinforcement learning technology to ensure different application qo S.The main research content of this thesis are:Aiming at the problems of partial overload and network congestion when using the traditional Equal cost multi-path routing algorithm(ECMP)to formulate routing policies for elephant and mouse flows in SDN network data center scenarios,an Intelligent Routing Algorithm(IRA)based on threshold traffic classification and Q learning is proposed.Firstly,the threshold traffic classification method is used to divide the service flow into elephant flow and mouse flow,and then the elephant flow is disassembled into a subflow through the disassembly module,and then a multi-constraint condition is constructed according to the QoS requirements of the two traffic types;finally,Q learning is used to dynamically and intelligently plan the link forwarding weight according to the link state and generate the routing strategy.Simulation results show that compared with dijkstra algorithm and ant colony optimization algorithm,the proposed algorithm reduces the end-to-end latency and packet loss rate of network nodes,improves the throughput of network nodes,effectively avoids network congestion,and improves network performance.Aiming at the QoS requirements of different application data streams in multiple application scenarios of SDN,this thesis proposes a QoS routing optimization scheme based on traffic classification and DQN.The traffic classification module consists of three parts: the traffic characteristic optimization submodule based on Relif F algorithm,the data set balance optimization based on K mean and SMOTE algorithm,and the network traffic classifier based on XGBoost algorithm,which effectively alleviates the problem of traffic classification accuracy due to the imbalance in the proportion of multi-application network traffic.The route optimization module consists of a multi-constraint routing module based on QoS and a routing module based on DQN algorithm,the former constructing multi-constraint conditions for different applications of QoS,and the latter mapping the relevant information of constraints and network traffic data to the state space,action space and reward function of the DQN algorithm,and generating the routing path scheme through policies.Simulation results show that the proposed algorithm has high classification accuracy,lower end-to-end average latency and packet loss rate.In this thesis,two different routing algorithms are set up for two different scenarios of data center and multiple applications,and experiments are carried out in the simulation environment Mininet.Experimental results show that compared with the comparison algorithm,the proposed algorithm has great advantages in converging speed,improving network link utilization and stability.In summary,the routing algorithm proposed in this thesis can effectively improve the network congestion problem and effectively improve network performance.