Research On System Delay Optimization For Mobile Edge Computing

With the rapid development of Internet,mobile devices such as mobile phones have been able to implement complex applications such as multimedia processing,face recognition,and augmented reality.Due to their small size,mobile terminals have very limited computing power and battery reserves,so they do not have the ability to handle complex computing tasks.Offloading computation-intensive tasks on mobile devices to the central cloud for execution is an effective solution.However,the transmission of large amounts of data on the core network will not only increase the network load,but also make the transmission time of task be too long.Mobile edge computing is an effective paradigm that delivers cloud services and functions to the edge of the network.This architecture improves the quality of service(Qo S)of mobile users by offloading their computation-intensive tasks to the edge server for execution.In mobile edge computing architecture,tasks will not be transmitted through the central network,but are processed at the edge of the network.For a mobile edge computing system,task offloading strategy and resource allocation method will greatly affect system efficiency.Therefore,this paper mainly studies the task offload scheduling and power allocation algorithms in mobile edge computing systems.This article considers a mobile edge computing system with multiple users and multiple servers.Each user can select an edge server in the system to perform its computing tasks.The ultimate goal is to minimize system latency(ie,maximum server latency).The main research contents,contributions and innovations of this article are as follows:1)A system delay optimization problem based on task scheduling and power allocation is defined.To solve this problem,this paper decomposes the original problem into two parts:task offload scheduling problem and transmit power allocation problem.Finally,two subproblems are used to solve the original problem in a loop and iterative solution.2)Decompose the task offload scheduling problem into two sub-problems,user-server matching and user-subchannel matching.Using matching theory,user-server and usersubchannel performs one-to-one matchings respectively.Secondly,by discretizing the power value,this paper proposes a heuristic algorithm to allocate transmit power to the terminal equipment.3)The algorithm JTOTPA proposed in this paper is compared with the three existing algorithms(DCORA,SDS,RS),in terms of system delay,energy consumption and comprehensive performance.The experimental results show that JTOTPA not only obtains a small system delay,but also generates the smallest user energy consumption,and its overall performance is superior to other methods.4)Existing research mainly aims at shortening user waiting time / reducing user energy consumption.From the perspective of the operator,this paper aims to shorten the system delay as the optimization goal.This method helps to improve system throughput.Most of the existing research only consider the data size of the computation tasks.This paper adds the consideration of unit data workload,which makes the algorithm model of this paper more practical.