Research And Implementation Of Resource Deployment And Scheduling Mechanism In Edge Computing Server

With the rapid development of Internet of Things,data traffic from the edge are explosive growing.Traditional cloud computing gradually shows its deficiencies in dealing with delay-sensitive and intensive applications.Mobile Edge Computing(MEC)has emerged as a new solution,which reduces the response delay and alleviates the network congestion by sinking computing ability to the edge of the network to achieve nearby response.However,the lack of computing resources makes MEC servers face challenges when dealing with massive demands.The mobility of equipments always leads to inefficiency of static resource deployment.Software Defined Network and Network Function Virtualization make it possible to do service deployment dynamicly,and it can significantly improve the resource utilization.But the frequent deployment or replacement alse bring extra delay.In order to better match the changing requests,the computing ability should be elastic to improve service quality.Therefore,this paper studies the MEC dynamic service deployment and resource scheduling mechanism,which mainly includes the following three aspects:(1)In order to maximize the use of MEC's limited resources and improve the quality of service(QoS),this paper proposes a computing unit deployment mechanism based on service popularity and equipment reliability.First,it uses the gated recurrent unit to do data prediction for request frequency and service reliability,so as to obtain a higher request hit rate.Then,based on the request frequency,capacity of MEC,and reliability of MEC,the MEC calculation unit deployment decision is made with the optimization goal of minimizing the total response delay and minimizing the number of calculation units.This problem is multi-objective and the heuristic ant colony optimization algorithm is used to solve the problem.The effectiveness of the proposed algorithm in terms of delay and service stability is verified through simulation experiments.(2)To adapt to the fast changing requests,this paper proposes a dynamic resource scheduling mechanism based on service intention and load balance.It first establishes service satisfaction model and load balance evaluation model to ensure the QoS of requests and the efficiency of MEC.Then,these two parameters are used as constraints,and the response delay is minimized as the optimization goal to solve the MEC resource scheduling mechanism.This problem can be modeled as a Markov decision process and solved by deep reinforcement learning algorithms.Simulation experiments prove that the proposed mechanism has better performance in terms of response delay,service quality and load balance.(3)Based on the above researches,this paper implements an MEC service deployment and resource scheduling system,which visually presents the simulation environment and the algorithm results.This system is divided into four parts,including the topology display,MEC node display,service deployment,and resource scheduling.Functional and performance tests verify the correctness and efficiency,and the operation results show that the system can provide researchers with a visual page for MEC service deployment and resource scheduling operations.