Research On 5G End-to-end Network Architecture And Resource Allocation Strategy Based On Network Slicing

With the development of 5G,user equipment is expected to exchange information anytime and anywhere in a broader geographical area.The management of high-speed mass mobile data will bring great challenges to the existing network architecture and resource management methods.In order to collect data faster and transfer it to the data center for effective management,this paper studies the network architecture and resource allocation methods based on the concept and characteristics of network slicing,and achieves the following contributions:(1)For the problem to transmit and process massive data in the future,a distributed application plane of end-to-end network slice architecture is designed.In this paper,the component of slice life cycle management in architecture and application plane is proposed,and the layered arrangement and management architecture of network slice based on business is realized.(2)The quality of service can no longer reflect the future users' perception of network performance and service quality,this paper proposes a demand characterization function based on the slice,and then combines the user feedback to get the actual demand characterization function value,Through the ratio to quantitatively express the quality of experience,and introduces the concept of virtual resource utilization ratio to represent the performance of the whole slice system.(3)The resource allocation model is established.Taking the weighted normalization parameters of the quality of experience and resource utilization as the optimization objective,based on this,a joint optimization algorithm is proposed to allocate radio resources to provide users with better the quality of experience.The deep reinforcement learning algorithm is used to formulate better virtual access unit allocation strategy in the edge cloud.Online auction is used and compete for a better virtual resource solution in the market.The simulation results show that the model has better performance than the traditional way.Finally,the related work of this paper is summarized and prospected.