The Study On Using Distributed Constraint Optimization Problem To Model Data Caching In Edge Computing

Distributed Constraint Optimization Problem(DCOP)is a fundamental framework for Multi-Agent System(MAS),which can model multi-agent coordination optimization problems.At present,most of the work in the field focuses on algorithm theory and less work on practical applications.Edge computing is a new type of network computing paradigm which has the advantages of high real-time and robust privacy,and has shown an outstanding development prospect.Therefore,it has received extensive attention from scholars.However,there are still many unsolved problems in edge computing,one of which is data caching in the edge computing environment.Currently,the solution methods for data caching problems in edge computing are all in a centralized manner,which does not conform to the distributed nature of edge computing.For a type of data caching problems in edge computing,this thesis proposes a DCOP-based model and presents two algorithms for this model.The specific research work is as follows:(1)This thesis systematically reviews related work of data caching problems in edge computing,and analyzes and compares the shortcomings of current research work.Moreover,it gives the basic definition,representation method,and the background of DCOP,and elaborates two local search approximate algorithms for DCOP,DSA and MGM.In addition,the thesis proposes a DCOP–based model for a type of data caching problems in edge computing.The proposed model casts away the centralized solver in the traditional solution methods and effectively decomposes the problems that originally require global information to solve.In the model,the global hard constraint and optimization objectives of the original problems are factorized to local information for each edge server(i.e.,agent),which include local hard constraint and two-objective constraint costs.Accordingly,each edge server cooperates with each other to make its decision based on its local information to solve the data caching problems in edge computing.(2)Based on the DSA and MGM algorithms,two approximate solution algorithms for the DCOP-based model are proposed.Since the global hard constraint and optimization objectives are factorized to local hard constraint and two-objective constraint costs,two variants for DSA and MGM are presented,named EDCDSA and EDCMGM.The proposed algorithms transform two-objective constraint costs into single-objective constraint costs by summing weighed objectives,and then introduce a local decision-making strategy for each agent.Furthermore,different message processing modes are given for each agent according to whether it satisfies the local hard constraint.To verify the proposed model and its solution algorithms,we implement part of the core logic of the distributed constraint optimization platform(DCOPSolver)and add a data caching module of edge computing.The experimental results show that the proposed algorithms perform no worse than state-of-the-art centralized algorithms when solving larger-scale data caching problems in edge computing.