High-reliability And Low-cost Task Scheduling In Edge Computing

With the rapid development of mobile communication technology and the Internet of Things(Io T),emerging technologies such as collaborative driving,behavior recognition,and virtual reality are applied in large numbers,and the resources of Smart Mobile Device(SMD)are limited to meet the huge demand of such applications.Mobile Edge Computing(MEC)has been proposed to solve the problem of scarcity of various resources due to the portability of smart mobile devices.However,current scheduling algorithms still have many shortcomings: on the one hand,the workflow scheduling process is inevitably characterized by edge network fluctuations and server failures,which may lead to privacy leaks and task offload failures,so the workflow scheduling must consider reliability constraints;on the other hand,users not only put forward higher requirements for the response speed of smart mobile devices,but also require minimizing the computational cost of task scheduling,and the existing The existing scheduling algorithms still have room for improvement.To address the abovementioned problems of workflow scheduling in MEC environment,this thesis conducts an in-depth study,and main innovations and work are as follows:1.A reliable task scheduling algorithm based on push-pull search(RT-PPS)is proposed for the reliability problem of task scheduling in MEC environment.The RTPPS algorithm firstly calculates the reliability and filters nodes by grouping edge nodes,and generates the ready queue by processing workflow using HEFT algorithm.Next the initial population is generated using random numbers based on a genetic coding scheme.Then according to the push search strategy,the Pareto front is found under unconstrained conditions and the change rate is updated.Finally,the optimal new generation is obtained by selecting a non-dominated solution set.When the change rate exceeds a threshold,it is transformed from push search to pull search,and the optimal scheduling solution is found from the Pareto front using the non-dominated sorting genetic algorithm Ⅲ under the constraint,which improves the success rate of task offloading execution under the condition of protecting user privacy.2.To address the problem of pricing edge services when SMD performs task scheduling,this thesis constructs a server pricing model in MEC environment and proposes a cost and latency-aware task scheduling algorithm based on SPEA2(CLASPEA2).The algorithm first processes the workflow to generate the ready queue.Next,the initialization generates the population of energy consumption constraints.Then,diverse population individuals are generated during the iterative process by operations such as non-dominated sorting,adaptive cluster selection and cross-variance,while domination-resistant solutions are eliminated by continuously maintaining individual profile records.Finally,the optimal individuals satisfying the conditions are selected to generate a scheduling scheme to achieve low computational cost task scheduling.3.In this thesis,RT-PPS algorithm and CLA-SPEA2 algorithm are implemented separately using Edge Cloud Sim,a mobile edge simulation platform.Using the platform to simulate the MEC environment for multiple experimental comparisons,the results show that under the reliability constraint,the RT-PPS algorithm proposed in this thesis is 2.8%,10.3% and 29.0% lower than the NSGA-III,NSGA-II and GA algorithms,respectively,in terms of combined optimization of device energy consumption and workflow execution time;the CLA-SPEA2 algorithm reduces the joint optimization of execution time and computational cost by 21.3%,13.4%,and 5.6%,respectively,compared to RR,PSO,and SPEA2.