Research On Traffic Analysis And Routing Optimization For 5G Optical Fronthaul Network And Metropolitan Area Network

The rapid development of 5G technology has brought many positive changes to human society,but at the same time,the number of end users continues to grow explosively,followed by an increasing amount of data,which has brought great pressure to the optical network that transmits data.In order to meet the transmission requirements of various scenarios,especially delay sensitive services,it is bound to put forward higher requirements for network performance.Quickly mining the effective characteristics of network traffic is conducive to understanding the network status and making subsequent decisions,such as energy conservation,resource arrangement and routing planning.Therefore,timely and accurate network traffic prediction has great significance and good application prospects for network operation and maintenance.Based on the results of traffic prediction,the optimal resource allocation and routing strategy can be calculated in advance,and then the specific routing optimization algorithm can be used to achieve online routing decisions in real optical networks.Aiming at the above problems,this paper focuses on the traffic planning and routing optimization in 5G optical fronthaul and metropolitan area network.The points of elemental work and innovations in this paper are shown below:Due to network fluctuations,burst traffic and other reasons,time series traffic will produce uncertainty and periodicity is not strict,which will cause great interference to traffic prediction and analysis,and traffic in different modes has different trends.The traditional traffic prediction model based on convolution or recursive operation has defects in predicting speed or solving long-term dependence.In response to these problems,this paper proposes a Transformer traffic prediction model based on the self-attention mechanism,and constructs two common neural networks as a baseline to compare with the proposed model,reflecting the advantages and characteristics of the proposed model in terms of accuracy and speed in dealing with such traffic prediction problems.In order to ensure the optimal prediction effect,a series of data processing work will be carried out,including eliminating redundant information,supplementing default information and data normalization.Experiment with a real set of optical fronthaul network traffic data,the results show that the prediction accuracy of Transformer model based on the selfattention mechanism is 94.6%,and its prediction time is 1.28s,which has obvious advantages over other models.In order to make full use of traffic prediction results,in an optical metropolitan area network disaster scenario,in order to quickly recover routing,solve data forwarding,and maintain network utilization.In this paper,we propose a traffic-driven multi-constraint multi-path routing algorithm in optical transport networks,which combines traffic with routing algorithm,uses dynamic traffic pruning to reduce network complexity and solve routing optimization under multiple constraints.Based on the real network topology and its connection requests,simulation analysis is conducted.The results show that the average time for traffic routing using the traffic driven multi constraint multi path routing algorithm proposed in this paper is 20.4 ms,and the standard deviation of load occupancy is 0.0122.The time performance and load balancing performance of the algorithm are verified.