Research On High Reliable Transmission Technology Of SDN Proprietary Trusted Network

While empowering enterprise management and business innovation,cloud computing in enterprises has also raised concerns in the industry about cloud data security.In order to prevent malicious users from accessing the network and using it for destruction,some institutions use a composite security mechanism of user authentication and frame verification in SDN scenarios to ensure security.This article refers to this network architecture as an SDN proprietary trusted network.Strict access control and communication control ensure security,but also significantly increase the network packet loss rate.This article optimizes the problem of high packet loss rate in SDN proprietary trusted networks,and experiments show that the optimization scheme can reduce the packet loss rate by half on average,effectively improving network reliability.This article first analyzes the verification mechanism of SDN proprietary trusted networks and provides a necessary and sufficient condition for switching devices not to lose packets in this scenario.Based on this condition,a partially backward compatible frame verification algorithm PBCFV is proposed to ensure that large streams are not discarded.Under the PBCFV algorithm framework,the controller will issue verification algorithm parameters for each switching device,which can ensure that the flow to a specific destination is not discarded on that device.The quality of the verification algorithm parameters determines the effectiveness of the PBCFV algorithm in reducing packet loss rate.In order to make the PBCFV algorithm parameters adapt to the future maximum flow as much as possible every time,this paper proposes a sorting prediction network RSTGCN based on spatiotemporal graph convolution.The RSTGCN model uses graph convolution to aggregate information from adjacent nodes in the topology and predict the traffic matrix of the entire network.In order to further improve the accuracy of RSTGCN in predicting the relative size of traffic,this paper introduces the sorting error PMML into the loss function of RSTGCN.The experiment shows that the accuracy of RSTGCN in predicting maximum flow is as high as 92.46%,which is an average improvement of8.81% compared to single node prediction methods such as ARIMA and GRU.In response to the situation where the maximum flow in different regions of the network may not be the same,the article adopts a region differentiation generation algorithm to further reduce the packet loss rate of the network.From the perspective of cooperative game theory,this paper proves that the optimal coalition structure of a graph is the optimal solution to the problem of generating regional differentiated random codes.This article proposes a parallel graph decomposition algorithm PGD to solve the optimal alliance structure problem.The PGD algorithm performs special processing on the cut points of the topology by decomposing the original topology into connected component pairs without repetition or missing,reducing the width of the search tree and further shortening the algorithm’s solving time through parallelism.In response to the problem of long computation time for PGD algorithm in large topologies,this paper adopts the communication based multi-agent reinforcement learning model DGN for solving,and designs the action space,observation space,and reward function of the agent based on the idea of infinite repeated games.Experiments have shown that multi-agent reinforcement learning algorithms can provide real-time results with a theoretical optimal value of around 92%.