Research On Computation Offloading Strategies For Multi-service Edge Networks

In recent years,with the development of mobile communication technology,different terminal services have higher and higher requirements for communication system.Mobile edge computing technology is introduced to achieve the goal of low delay and high speed in order to meet the business needs of different scenarios.But the requirements of different business types and the characteristics of computing tasks are different.The state of the network is constantly fluctuating over time.It is one of the problems that need to be solved in practical engineering application how to make lowdelay unloading decision in edge network and ensure that the edge server cluster is in load balancing state at different times.This paper aims to construct different communication parameters and service scales of various kinds of services on the edge network,so as to achieve a model closer to the real scene and make the solution of decision vector more engineering realizable.At the same time,an efficient computing and unloading strategy is proposed to reduce the delay of the terminal to obtain the calculation results of the unloading task,and optimize the performance of load balancing of the edge server cluster.Firstly,this paper expounds and summarizes the development of mobile edge computing support technology,modeling of computing tasks and research status of computing offloading strategy in edge network.Then analyzes the defects existing in the existing research and the characteristics of various services,and builds simulation models of various services on this basis.It includes the size and scale of node computing tasks as well as the performance parameters of terminal devices.Finally establishes the flow network model under the scenario of terminal server,which provides the support of the scene and model for the research in the following paper.Then,in order to realize the initial matching selection of terminal and server,this paper uses the minimum cost and maximum flow algorithm to solve the optimal matching of dichotomous graph of terminal and server on the basis of the previously constructed stream network model.This builds the initial terminal server match list.With the fluctuating of network state,the use of search algorithms of each dynamic moment mean network are applied to solve the minimum delay cost decision vector.The simulation results show that the time delay obtained under three single service type scenarios can meet the requirements of maximum delay tolerance under various service types,and the delay performance is significantly improved compared with the requirements of maximum delay tolerance.Finally,in view of the unbalanced load state of the server cluster after computing unloading,an index to measure the performance of load balancing is proposed.The load balancing state is introduced into the flow network to rebuild the edge weight,and the minimum delay computing unloading strategy considering the load balancing state is proposed.Finally,the central cloud server is introduced as one of the decision objects of computing offloading.Simulated annealing algorithm is introduced to optimize the search process of computational unloading decision.The simulation results show that the calculation of unloading delay meet the needs of different services and the performance of load balancing is improved in the hybrid business scenario.