Joint Scheduling Of Spectrum And Storage Resources For The Edge Computing Enabled Space-Ground Integrated Information Network

In the 5G era,mobile devices generate massive amounts of data,most of which are multimedia contents and occupy a lot of wireless spectrum resources.Transmitting such large-scale multimedia data over a wireless network will cause network congestions and long service delays,which can be addressed by caching content at the edge of the network(for example,base stations in Mobile Edge Computing(MEC)),however,edge caching may not be able to handle surge traffic during peak hours.Faced with the resource shortage challenge,this paper proposes to use satellite networks to enhance the performance of the terrestrial network,and design an edge computing enabled space-ground integrated information networks(SGIN)-oriented joint scheduling algorithm for network resources for content caching and distribution.The main contributions of this paper are as follows:1)Considering the SGIN scenario with content distribution as a typical service,a joint scheduling algorithm for storage and communication resources is proposed to design a content cache and distribution strategy.At the same time,the joint scheduling of communication resources and storage resources is considered to improve the system utility(improve cache hit rate and reduce bandwidth consumption);2)Decoupling communication resources from storage resources by means of phased scheduling.First,the Multicast Transmission(MT)stage based on a genetic algorithm(GA)is proposed as the user-node matching algorithm,and then a cache placement algorithm based on multi-agent deep reinforcement learning(MADRL)is proposed during the cache placement(CP)stage;3)The simulation and comparative analysis of the proposed algorithm are conducted from five aspects: the comparison of SGIN and the ground network system;the comparison among different user-node matching algorithms in the MT stage;the comparison among different caching algorithms in the CP stage;the comparison among multi-agent and singleagent algorithms,the comparison among different parameters.Compared with other benchmark algorithms,the proposed algorithm has achieved higher system utility.