Fuzzy Workflow Scheduling Method Based On Pareto Front In Cloud Computing Environment

This paper studies the multi-objective optimization problem of fuzzy workflow scheduling based on Pareto front in the cloud environment.The workflow of this problem has the characteristics of large scale,computing intensive,fuzzy and dependency,and is executed on elastic cloud resources with various price structures.In this paper,a fuzzy problem model is established based on triangular fuzzy number,and an adaptive iterative heuristic workflow scheduling method is proposed to minimize the rental cost of virtual machine and maximize the user satisfaction.This method includes three different scheduling stages.First of all,according to the priority arrangement stage of the task in the scheduling process,the priority is sorted according to the earliest start time of the task.In this paper,the earliest possible start time calculation method PEST?Possible Early Start Time?and the earliest feasible start time FEST?Feasible Early Start Time?of the task are proposed.Secondly,in order to further improve the scheduling ability,this paper use a variety of heuristic optimization algorithms,such as Variable Neighborhood Descent?VND?and Simulated Annealing?SA?.Then in order to be more practical,this paper proposes to use the Relaxed Task Deadline?RTD?and the Tight Task Deadline?TTD?,and allow the task to expire.In addition,in order to find a better balance between low rental cost and high user satisfaction,this paper introduces Pareto multi-objective optimization theory,and uses Pareto Front evaluation indexes such as Distance Average metric(d_?6?,Distance Max metric(d₈?6?),Maximum Spread metric?MS?and Average Quality metric?AQ?to evaluate the solution quality.Finally,the method of each scheduling stage is calibrated by detailed experimental design and reasonable statistical technology.The experimental results show that the workflow scheduling method proposed in this paper is robust and effective to the problems existing in the research.It can find a good balance between the number of virtual machines and user satisfaction,effectively reduce the cost of renting virtual machines and improve user satisfaction as much as possible.