Research On Key Technologies Of Resource Allocation And Optimization In Edge Computing Environment

With the advent of the era of Internet of things,the task and data volume of mobile terminal devices are growing rapidly,which brings the demand for continuity of service,high reliability and real-time task processing.Edge com-puting paradigm provides low latency and efficient services for delay sensitive applications.Edge computing can provide more resources at the edge of the network than terminal devices.However,compared with traditional cloud com-puting,the resources provided by edge computing servers to support task trans-mission and processing such as computing and storage are limited.Reasonable resource allocation can reduce the processing time of terminal tasks,improve efficiency and accuracy,and improve user service experience,service provider benefits,and system stability.This paper is oriented to the edge computing en-vironment and launches the research on the key technologies of resource alloca-tion and optimization.It mainly includes the following four aspects of research content:(1)Aiming at the problem of workload difference in time and space caused by the mobility of users,this paper proposes an edge computing server deploy-ment method under the influence of two-dimensional space-time.First of all,from the perspective of service provider's benefit,on the basis of meeting the delay requirements of task processing,a server resource requirement estimation algorithm considering user preferences is proposed.Secondly,considering the mobility of users,the real data is abstracted from the two dimensions of time and space,and a small full functional area is established to estimate the mobility and density of users.The location of servers is arranged based on the centroid of grid.On this basis,the competition and cooperation between edge computing service providers are studied,and a Bayesian game-based model without com-plete information of other competitors is proposed to obtain the layout scheme of edge computing servers.Simulation results verify the advantages of the pro-posed method in resource utilization and task processing delay.(2)Aiming at the problem that the remaining available resources of the edge computing server may not be enough to support the processing of large and indivisible computing tasks,this paper proposes an edge computing resource pre-allocation algorithm based on usage prediction,and explores the optimal auction scheme for both sides according to the different personality charac-teristics of users and edge computing service providers.Considering that the mobility of terminal equipment may change the location and area of receiving service,a state search algorithm is proposed to predict the staged destination of terminal equipment to ensure timely processing and completion of terminal equipment requests and tasks.Simulation results verify the performance of the proposed algorithm in terms of satisfaction and resource utilization.(3)Aiming at the problem that the resource of some servers is idle for a long time due to the difference of workload between edge computing servers,a fairness-oriented strategy for task offloading and load balancing of edge com-puting server is proposed.Firstly,the task is divided into several subtasks based on application workflow,and the resource of edge computing server is virtual-ized.The task offloading problem of terminal device is simplified as the map-ping problem of subtask-virtual machine resource.Secondly,a task scheduling supplement mechanism is proposed to improve the resource allocation,system utility and fairness between edge computing servers.Simulation results verify the performance of the proposed strategy in task delay,system utility and user satisfaction.(4)Aiming at the problem of service stagnation or interruption caused by the high dynamic of users,and the problem of sensitive information leakage caused by the instability of users' mobile state,an edge computing service mi-gration mechanism that guarantees user privacy is proposed.First of all,consid-ering that the user may be in a high-speed mobile state,and the user's location information is sensitive and not leaked,the user's mobile location and path are predicted based on historical data and short-term real-time behavior detection.Secondly,considering that the server clustering based on service similarity can be mapped to user interest clustering,the service type and resource requirement of the target server are determined by user interest clustering and user mobility.Finally,the target server of service migration is determined by evaluating the reputation of the edge computing server.Simulation results verify the commu-nication overhead and satisfaction of the proposed mechanism.