Design And Implementation Of Workflow System EdgeWorkflow Based On Edge Computing

With the advent of the Internet of Things era,edge computing has become the ideal computing paradigm of intelligent logistics,intelligent health,intelligent transportation and other intelligent systems.Compared with the traditional cloud computing platform,the edge computing environment has the characteristics of low latency,distribution and heterogeneity.However,the resource heterogeneity and network complexity of edge computing environment make the resource management algorithm of workflow system in traditional cloud computing environment no longer applicable.Therefore,how to run complex scientific workflow more efficiently in edge computing environment is an urgent problem to be solved.Current research focuses on various resource management optimization algorithms in edge computing environment,and the effectiveness of the proposed scheme is verified by simulation experiments.At present,the research on workflow system mainly focuses on scientific workflow modeling,simulation of workflow task execution,simulation of cloud/fog/edge computing environment resource types and network topology.However,the resource heterogeneity,network dynamics and workflow task attribute diversity of edge computing environment cannot be completely reproduced in the simulation experiment environment.This leads to a big gap between the experimental results of various existing resource management methods in simulation and real environment.Therefore,an easy-to-use and efficient edge computing workflow system is urgently needed to promote the research of resource management.In view of the above problems,this thesis has done the following three aspects of research:Firstly,this thesis designs Edge Workflow system.Its framework is divided into four layers: infrastructure layer,middleware layer,service layer and user interface layer.The infrastructure layer is mainly responsible for establishing real edge computing resources and network environment.The middleware layer is mainly responsible for receiving the instructions of edge computing environment setting and workflow tasks of the service layer,and sending the instructions of constructing edge computing environment and executing workflow tasks to the infrastructure layer.The service layer generates the optimal execution scheme of workflow tasks through workflow management service,and deploys workflow tasks to the real edge computing environment through workflow engine.The user interface layer is responsible for interacting with users.Then,due to the premature and local convergence of workflow task unloading and scheduling algorithm in edge computing environment,the optimization effect and efficiency of workflow task execution time are not high.In this thesis,Genetic Algorithm(GA)with strong searching ability is used in the first half of iterative process of workflow task unloading and scheduling algorithms respectively.In the last half of the iterative process of the algorithm,Particle Swarm Optimization(PSO)algorithm with fast convergence speed is used,and GA-PSO Hybrid algorithm based on GA and PSO is proposed.This algorithm combines the strong searching ability of GA algorithm and the fast convergence of PSO algorithm,which can reduce the execution time of the algorithm while generating the optimal task unloading and scheduling scheme.Experiments show that GA-PSO Hybrid algorithm has fewer workflow tasks and algorithm execution time compared with GA and PSO algorithm.Finally,this thesis implements the Edge Workflow system Edge Workflow based on the above system framework.The Edge Workflow system can select workflow files from workflow libraries,or users can use visual modeling tools to generate executable workflow XML files.With the workflow system engine of Edge Workflow,users can automatically generate a customized real edge computing environment.Deploy the user-selected workflow instance to a real edge computing environment for execution.Users can also select various resource management algorithms contained in the system or apply their own resource management and task scheduling algorithms.The graphical interface can monitor the execution status of workflow tasks in real time and obtain multi-angle data about the execution results of tasks.We use UAV(Unmanned Aerial Vehicle)last-kilometer delivery system based on edge computing as a real case study,and use some representative scientific workflows(such as Montage,Cyber Shake,etc.)for our experiments.Experimental results show that Edge Workflow can effectively evaluate the performance of different resource management and workflow task scheduling algorithms in real edge computing environment,and can automatically deploy and execute scientific workflow instances in user set edge computing environment.In this thesis,Edge Workflow is designed and implemented for the first time in edge computing environment.After summarizing the defects of other workflow systems,this system designs and implements a workflow system that can deploy edge computing environment with one click,and verifies the efficiency and effect of the proposed Edge Workflow system in optimizing workflow task execution in edge computing environment from simulation and real experiments.Second,edge computing environment workflow task for and unloading scheduling algorithm may be premature and local convergence phenomenon,design and put forward the workflow task scheduling method based on time optimization(GA-PSO Hybrid),combined with the advantages of GA and PSO algorithm,improves the edge computing environment workflow task to unload and scheduling time efficiency.