Research On Caching Strategy In Edge Cell Cellular Network

With the widespread deployment of 5G networks and the explosion of mobile data traffic worldwide,a large number of edge nodes are deployed in cellular networks to accommodate user demand,but the widespread deployment of edge nodes increases the burden on backhaul links.To solve this problem,edge caching is considered to be an effective solution.Edge caching shortens the distance between content servers and users,thus the users can get the required file data from the edge nodes instead of served by the data center in the core network.However,as the number of small base stations(SBS)and users in cellular networks increase,the energy consumption in content caching and file delivery is considerable,so energy efficiency remains an important aspect to be considered in caching strategy design.Therefore,an efficient edge caching strategy not only reduces user latency,but also improves system energy efficiency.While the storage capacity of SBS used for caching in cellular networks is limited,the contents in the network are huge.The research and analysis have shown that the files frequently requested by users tend to be a small portion of the tatal content,and this popular content accounts for the majority of network traffic.Therefore,effective caching can be achieved by determining part of the popular content through file popularity prediction.Since most of the existing file popularity prediction algorithms are based on user requests only and do not consider the file characteristics,the prediction accuracy is not ideal.In order to improve the caching performance,this thesis proposes a deep learning-based file popularity prediction algorithm,which first extracts file features by the training network,then associates file features with user requests to construct an association matrix,and finally calculates the file popularity value based on the association matrix.Hence,we can achieve the prediction of file popularity in the network.In this thesis,an energy efficiency optimization caching strategy is proposed under the premise of known file popularity.The content is cached by mixed partition caching in SBS of heterogeneous cell network,and the requested content is delivered by joint transmission,parallel transmission or direct transmission according to the user's location and file caching mode.The energy-efficient optimal caching model is established to analyze the probability of successful file transmission under different transmission methods,and then the energy-efficient problem is divided into two sub-problems by stepwise optimization algorithms.Firstly,the location of the file cache is determined under the premise of maximizing the cache hit,and the corresponding cache placement strategy is obtained by solving the non-convex problem through heuristic simulated annealing algorithm.Moreover,part of the cache files are transformed into the most popular cache mode under the premise of maximizing the energy efficiency,and finally the optimal cache strategy design is realized.For the high request density scenario,this thesis investigates the caching policy under the multicast and proposes an energy-efficient optimization method.Simulation results show that the proposed caching policy achieves a higher cache hit rate and outperforms the related comparison algorithms in terms of energy efficiency.