Research On Service Resource Allocation Model In Edge Computing Environment

Edge computing greatly reduces the delay of data transmission because the edge servers are closer to terminal devices.However,because the capacity,bandwidth and computing power of an edge server are weaker than a cloud server,it is impossible to deploy all resources on one edge server to meet user needs,allocated the limited service resources to edge servers is a research hotspot.At present,the allocation of service resources under the edge computing environment mostly studied how to minimize the delay to meet the needs of users.However,it ignores the influence of service popularity on the allocation of service resources and the benefit demand of service operators,it caused low resource utilization rate and less revenue of service operators.Therefore,the allocation of service resources based on comprehensive consideration of service heat,time delay and service operation cost can improve the service experience of users and meet the benefit needs of service operators effectively.In this thesis,we will study service allocation in the scenario of edge computing service resource allocation,a service resource allocation framework is proposed,and the problem of edge computing service resource allocation is transformed into a nonlinear 0-1 programming problem.Firstly,Hacker News algorithm is used to calculate the service heat and construct the decline sequence of service heat.Then,considering the influence of service heat,edge server capacity,bandwidth,the request pf end users and the other factors on service resource allocation in edge computing,a service resource allocation model based on service heat is designed aiming to jointly reduce the delay of end users and the operating cost of service providers.In order to solve the service resource allocation model,the original harmony search algorithm is improved in this paper.On one hand,the two main parameters and variation of the algorithm are adaptive processed,so that the algorithm can avoid local optimization by expanding the search scope and increasing the probability and the degree of variation in the early stage.In the later stage,the convergence speed and search accuracy can be accelerated by improving the local search ability,transmitting the information of the current optimal solution to the next generation,and reducing the mutation rate and the mutation degree.On the other hand,by introducing penalty functions and adding penalty constraint terms to the objective function to eliminate service resource allocation solutions that do not meet the constraint conditions in the iterative process,we can accelerate the rate of convergence and improve the accuracy of finding the optimal solution.The experimental results show that the heat based service resource allocation model proposed in this thesis has obvious advantages in reducing the average request delay of end users and the operation cost of service providers.