Research On Dynamic Cloud Workflow Scheduling Based On Genetic Algorithm

Workflow scheduling algorithm in cloud environment provides an efficient scheduling solution for cloud computing platform.Workflow scheduling problem has always been considered as a NP hard problem,many scholars have done in-depth research on this hot issue.However,some algorithm ignores the dynamic environment when designing the algorithm to solve the problem of workflow scheduling,and the designed algorithm can not continuously schedule the workflows.Or in determining the workflow scheduling solution,the dynamic changes in the cloud environment in the actual scheduling process are ignored.Aiming at the above defects of workflow scheduling algorithm at present,this paper does the following work.Firstly,the paper summarizes a series of heuristic rules based on the critical path method,encodes the heuristic rules into genes on chromosomes,and proposes a heuristic rules based genetic algorithm(HGA).In order to apply HGA to multi workflow scheduling problem,this paper adopts the method of combining genetic algorithm with batch processing,by dividing the workflow scheduling process of cloud platform into different periods,and uses genetic algorithm to generate scheduling solution for each period.In this way,genetic algorithm can schedule multi workflow continuously.In order to verify the superiority of genetic algorithm based on heuristic rules,this paper compares HGA algorithm with genetic algorithm based on location relation and genetic algorithm based on mapping relation,and conducts comparative experiments with meta heuristic algorithms HEFS,Max-min and Min-min.The results show that HGA's ability to find the optimal scheduling solution is better than other algorithms,and it can reduce the total execution consumption of the cluster.Secondly,in order to adapt to the dynamic environment of cloud platform,this paper improves the original heuristic rules.In order to make the generated scheduling solution adapt to the dynamic workflow environment in practical application,this paper designs a series of rules to make the generated scheduling solution adjust according to the specific workflows in the actual scheduling,so as to improve the effectiveness of the scheduling solution in the dynamic environment.The experimental results show that the genetic algorithm based on improved heuristic rules is more robust than the genetic algorithm based on location relation and the genetic algorithm based on mapping relation.