Traffic Analysis And Resource Allocation Of 5G Optical Fronthaul Network Based On Deep Learning

During the rapid development of mobile communication for decades,billions of intelligent devices connect information objects.Lots of data-intensive,personalized and diverse applications bring great pressure to the existing access and fronthaul network.Base station as the carrier of information interaction and network service supply is vital in people's communication requirements,hence,the maximum excavation of effective information of base station traffic flow is conductive to achieving the flexible allocation and arrangement of resources.However,spatio-temporal dynamic condition and uncertainty exist in the business request of base station so that the timing problem is analyzed and predicted difficultly.Besides,it has become the key problem to be solved for 5G as the next generation of cellular system to meet the fronthaul network's demands in high data rate and high service quality,to adapt the explosive growth of flow.C-RAN is one of key architectures for 5G.It is also crucial to achieve the intelligent and efficient configuration of resources by fully utilizing collective and collaborative features.To solve these problems,this paper focuses on the C-RAN and takes the fronthaul network composed of low-latency and high-bandwidth fiber as the research object.The main achievements and innovations of this paper are as follows:1)Aiming at the problem that it is difficult to analyze and predict non-stationary traffic,four traffic prediction schemes based on deep learning algorithms are proposed.In the process of data analysis,it introduced the related work of feature engineering such as data cleaning and feature coding,aiming to maximize the effective information in the traffic and remove the redundancy for the model to play an excellent performance.In order to ensure the prediction effect,deep learning model is continuously optimized through strategies such as hyperparameter tuning,selection of appropriate optimization algorithms,and fitting correction.Finally,PA of Prophet model can reach to 92.16%at most.2)Aiming at the problems of low resource utilization and rigid resource management of 5G optical fronthaul network,an energy-saving strategy of dynamic load transfer based on traffic perception is proposed.On the basis of traffic prediction,the improved bin packing algorithm is used to migrate the traffic load processing task to a few BBUs and close the idle BBUs.Then achieve the purpose of energy saving.Through the load migration solution,the number of active BBUs has been reduced by 66.67%,and the average utilization rate of BBUs has increased from 15.60%to 85.83%.3)Aiming at 5G high-quality,low-latency business requirements,an optical forwarding load balancing scheme based on traffic sensing is proposed.Based on traffic awareness,the mapping relationship between RRH-BBU is dynamically configured through clustering and particle swarm optimization algorithms to solve the problem of non-uniformly distributed load,and it is convenient to achieve resource aggregation and dynamic allocation.Simulation results show that the load balancing factor of this scheme can reach to 90.32%,and the blocking rate is reduced by 14.01%.