Intelligent Scheduling Optimization Method For Scientific Workflow Based On WorkflowSim Toolkit

As a systematic and integrated form of data with diverse topologies,workflow is often used to build tasks and services with complex dependencies.Faced with the dilemma of low execution efficiency of large-scale complex workflow,more and more industrial and academic workflow-based applications are deployed in the cloud computing environment,which enables workflow datasets generated by complex applications to be quickly processed based on the new paradigm of highperformance computing provided by the cloud center.In the real-world situation,unreasonable scheduling order for the given workflow datasets on the cloud resources may increase computational consumption significantly and unnecessary service time delays.Generally,scheduling workflow data in the cloud computing environment mainly relies on heuristic algorithms and intelligent optimization algorithms,which can generate proper scheduling strategies.Heuristic algorithms can find solutions for the given scheduling problem within few minutes,but the results of obtained solutions may have poor qualities.While intelligent algorithms can explore good results,they often consume a lot of computational time.This dissertation combines the characteristics of these two kinds of algorithms,and proposed a hierarchy-based particle swarm optimization(PSO)algorithm which contains hybrid heuristic strategies.Firstly,this paper formulates a single-objective constraint optimization model.This model is designed for the scientific workflow scheduling problem in the heterogeneous cloud environment.The goal of the model is to minimize economic expenses of executing all tasks of the workflow within a given deadline.To simplify complex topologies of scientific workflows,a preprocessing strategy is proposed for the model.Then in the initial phase of the proposed algorithm,an initial solution generation method and a dynamic expansion search space strategy are proposed,which help generate the random population.Based on the initial particles,the population division and hierarchical evolution method is used to update all particles.In the update process,this paper adopts the "step by step" search method to explore potential solutions,trying to find better solutions.In the experimental part,this paper simulates the heterogeneous cloud environment based on the Workflow Sim framework,and calibrates the proposed algorithm’s components and parameters accordingly.To verify the effectiveness,the performance of the proposed algorithm after calibration will be compared with several state-of-the-art algorithms on scheduling different types of scientific workflow with different deadlines.The experimental comparison results show that the proposed algorithm can schedule workflow tasks in a heterogeneous cloud environment with a higher completion rate and a lower economic cost.