Multi-workflow Scheduling With Time Constraints Onto Cloud Computing Resources

Recently,the problem of multi-workflow scheduling based on cloud environment has gradually attracted people’s attention.Cloud computing satisfy the quality of service(QoS)requirements,such as deadline constraints through reasonable scheduling schemes.Scheduling strategy will directly affect resource utilization rate(UR)and job success rate(SR),which are closely related to the economic interests of service providers and users.Therefore,the research topic of this paper is multi-workflow scheduling with time constraints onto cloud computing resources.The scheduling goal is to improve UR and SR.In instance-intensive workflow applications,it is common that multiple concurrent workflows are submitted to the scheduler for processing.This paper proposes CSJM algorithm based on job merging to ensure the fairness of scheduling.The algorithm first introduces the merging parameters,limits the number of single merging operations,then merges jobs of the same group into a new job according to the merging method.Meanwhile,this paper designs a preliminary experiment to explore how to set the value of the merge parameter reasonably.In view of the large amount of communication between data-intensive workflow tasks,this paper proposes the KSTA algorithm and the STLA algorithm.In the task scheduling sequence,these two algorithms both adopt a task sequencing strategy based on horizontal weight.In terms of processor selection,this article introduces the concept of reference value(RV).When calculating the RV,for subtasks that meet the conditions,the communication volume compensation factor is subtracted to strengthen the connection between the parent and child tasks.The difference is that the KSTA algorithm only considers the relationship between the key parent-child tasks,while the STLA algorithm considers each subtask in turn according to the amount of communication.In this paper,the RTWSim simulator is used to perform random workflow generation and simulation experiments.The newly proposed algorithm is compared with the classic scheduling algorithms.The experimental results show that the algorithms proposed in this paper can significantly improve the UR and SR.