Research On Content Caching Strategy Based On Substitution Theory

With the explosive growth of smart devices and mobile applications,the mobile core network faces the challenge of exponential growth in traffic and computing.Edge caching is one of the best solutions to this problem.The edge cache scheme mainly stores the content that users need to access at the edge of the network,and uses the edge to be closer to the user to reduce the delay of the user to obtain the requested content.At the same time,storing popular content on the edge also reduces the load on the backhaul link because it is frequently requested repeatedly.So,this thesis designed content replacement strategies based on popularity prediction and content placement strategies based on deep reinforcement learning to improve the content hit ratio of edge cache and reduce the content delay of user acquisition requests.The specific research work and results are summarized as follows:To solve the problem of content selection in edge cache,this thesis proposes a content popularity prediction algorithm based on minimal substitution theory and establishes a content substitution model based on minimal substitution theory.The growth mode of content is judged by Bayesian network,and the dynamic popularity of content is obtained according to the growth mode of content and content succession model.And design a content replacement strategy based on popularity prediction.Due to the limited edge cache space,this thesis constructs a content cache value function based on the popularity,size and the number of requests in a period of time,as a basis for evaluating whether the contents in the cache space need to be replaced.Periodic and conditional content replacement algorithms are designed to ensure timely elimination of outdated content,cache new popular content and improve content hit ratio.Simulation results show that the cache replacement algorithm proposed in this thesis has significantly improved the hit ratio compared with the traditional algorithm.For the overlapping coverage of multiple edge nodes,the base station fails to consider the location relationship between the user and the base station when delivering content.For the user at the edge of the base station,the transmission rate is low,and the transmission rate can be improved through mutual cooperation between the base stations.Therefore,this thesis proposes a transmission mode related to user location,and reduces transmission delay by jointly optimizing transmission mode and content placement.Specifically,considering the location relationship between the user and the base station,since the user of the overlapping part of the base station can be served by multiple base stations at the same time,the joint transmission and parallel transmission modes are designed to reduce the transmission delay of the content obtained by the user.An integer linear programming problem with user acquisition content delay as the optimization objective was established for the scenario with mixed transmission modes,and a content placement strategy based on deep reinforcement learning was designed.The experimental results show that compared with the traditional method,the proposed scheme has obvious delay reduction.