Research And Implementation Of 5G Slice Management System Based On Network Traffic Analysis

With the rapid development of mobile network technology,5G has been widely used in many vertical industries due to the capability of providing high-quality network services with large bandwidth,low latency,wide connectivity and high reliability.Particularly,power Internet of Things(PIo T)is one of the typical applications.The adoption of 5G to PIo T can match with the internal services and meet diverse power needs.Besides,network slicing is a key enabling technology for 5G-enabled PIo T.It separates a physical network into multiple logically independent,mutually isolated and functionally different virtual networks to provide different network services.In 5G-enabled PIo T,the network slice management involves traffic analysis and resource allocation.The 5G peak bandwidth continuous and the characteristic of rapid decline are different from the traditional network traffic.At the same time,the creation of network slices often fails to obtain the most underlying physical resources,resulting in critical problems such as high energy consumption and low resource utilization.To overcome the above difficulties,this research focuses on the 5G-enabled PIo T application scenario and takes it as an example to realize the intelligent slice management,and first presents a network slice management system,afterwards tackles two key optimization problems of the system.The details are shown as follows:(1)A 5G-enabled end-to-end network slice management system is designed for the application scenario of PIo T.The system maps the resources required by power users into corresponding slice subnets through virtualization technology,and uses software-defined network controllers to realize customized allocation of network resources.The PIo T slice management system realizes the allocation and management of network resources of each slice,and completes the experimental verification of system service reliability guarantee by improving the supply of slice resources or changing the method of network slices.(2)A prediction method based on CNN-GRU hybrid neural network is proposed for solving the traffic analysis problem for network slicing.This method can fully mine the latent features of the data,and fit the time series and complex nonlinear relationship of 5G bandwidth data.The parameters of 5G network bandwidth,signal-to-noise ratio and reference signal received power are predicted by the method.The experiments show that this method performs better in error control and training speed than traditional algorithms.(3)A promising approach based on deep reinforcement learning is proposed for addressing the challenging resource allocation issue for network slicing.For the core network and wireless access network,an end-to-end computing and communication resource allocation model(including energy consumption model,delay model and reliability model)is formulated.To meet the general communication requirements of PIo T,the optimization goal is to minimize the energy consumption of end-to-end network slicing.Through comparative experiments,the feasibility and performance of the resource allocation approach based on deep reinforcement learning are verified.The simulation results show that,compared with the traditional scheme,the proposed scheme achieves higher system reliability and lower energy consumption,thus enabling most of the network services to have the better quality of service.