Research On SDN Intelligent Routing Optimization Based On Deep Reinforcement Learning And Network Traffic State Prediction

With the development of new technologies such as 5G network,cloud computing and big data,the network scale has been continuously expanded,and various new network devices have been constantly emerging,which makes users put forward new requirements for network service quality.It is urgent to design an efficient intelligent routing optimization method.Due to the characteristics of tight coupling and decentralized management of control and forwarding,the traditional network architecture makes it difficult to obtain global network state information,which greatly limits the design and deployment of intelligent routing optimization methods and cannot meet the new needs of users.As a new network architecture,the Software-defined Network(SDN)obtains global network state information through the southbound interface,and the northbound interface provides application services for the upper layer.With the advantages of open and programmable network,decoupling control plane and data plane,and logically centralized control,it realizes centralized and unified control of the network,simplifies network management,and is very conducive to the design and deployment of intelligent routing optimization methods.In this paper,a network intelligent routing optimization method based on SDN single controller management is proposed to overcome the shortcomings of traditional routing methods,such as poor adaptability to dynamic and complex networks and inability to adaptively route and forward.At the same time,considering the problems of SDN single controller overload and network packet accumulation in large-scale networks,which can not make real-time intelligent routing decisions,a cross-domain intelligent routing optimization method based on SDN multi-controller management is proposed.The work and innovation of this paper are summarized as follows:(1)Network intelligent routing optimization under SDN single controller management:the traditional routing method makes use of limited information on the network links to make routing decisions,which makes it difficult to adapt to the dynamic and complex network and adjust the router’s forward strategy.To address these issues,this paper proposes an intelligent routing method based on the SDN,Dueling DQN(a Deep Reinforcement Learning algorithm),and network traffic state prediction.First,the global network awareness information is obtained with the SDN network measurement mechanism,which is converted into a traffic matrix consisting of multiple network link status information such as bandwidth and delay,etc.Then,the optimal forwarding route under the current network state is generated by predicting the network traffic matrix and the Dueling DQN.(2)Network intelligent routing optimization under SDN multi-controller management:message transmission and message synchronization for multi-controller interdomain routing in SDN have long adaptation times and slow convergence speeds,coupled with the shortcomings of traditional interdomain routing methods,such as cumbersome configuration and inflexible acquisition of network state information.These drawbacks make it difficult to obtain a global state information of the network,and the optimal routing decision cannot be made in real time,affecting network performance.This paper proposes an SDN crossdomain intelligent routing method based on multi-agent deep reinforcement learning and network traffic state prediction.First,the network is divided into multiple subdomains managed by multiple local controllers,and the state information of each subdomain is flexibly obtained by the designed SDN multithreaded network measurement mechanism.Then,a cooperative communication module is designed to realize message transmission and message synchronization between root and local controllers,and socket technology is used to ensure the reliability and stability of message transmission between multiple controllers to realize the real-time acquisition of global network state information.Finally,after the optimal intradomain and interdomain routing paths are adaptively generated by the agents in the root and local controllers,a prediction mechanism for the network traffic state is designed to improve the awareness of the cross-domain intelligent routing method and enable the generation of the optimal routing paths in the global network in real time.