QoS Sensitive Cloud Workflow Scheduling Optimization Method

Due to the high flexibility and scalability,and economy in cloud computing,many organizations migrate legacy their workflow applications to the cloud computing environments,which forms the cloud workflow.Cloud workflow scheduling refers to the allocation and execution of user-submitted workflows on the cloud platform.The scheduling process takes into account the quality-of-service(QoS)requirements,such as the execution time and cost.In order to solve the problem of QoS-aware cloud workflow scheduling,two kinds of cloud workflow scheduling optimization methods are proposed,which are respectively applied to the resource mapping stage and the task scheduling stage of cloud workflow scheduling.which are respectively applied to the resource mapping phase and the scheduling optimization in the task running phase of cloud workflow scheduling.The task-resource mapping phase pre-allocates the optimal resources for the tasks in the workflow while satisfying the user's QoS constraints.The existing cloud workflow scheduling algorithms are studied from both time and cost aspects,with little consideration of reliability.However,in the real case,the failure of resources and data transmission will have a negative impact on the successful operation of the workflow.The paper considers three important QoS factors: time,cost and reliability.Aiming at the problem of cloud workflow scheduling problem with minimization of time and reliability constraints,an optimal scheduling scheme search method based on firefly algorithm and dynamic priority is proposed.In particular,according to the characteristics of cloud workflow scheduling problem,the location,distance and position updating method of firefly algorithm are redefined.At the same time,dynamic priority algorithm is adopted to determine task order to reduce the workflow completion time for each scheduling scheme.According to the task-resource mapping relation,the task is scheduled to execute on the corresponding resource,and the scheduling overhead is generated in the scheduling process.Because there are many tasks with relatively short running time in cloud workflow,the overhead of scheduling such tasks in a distributed cloud environment is far beyond the running time of the task,which affects the execution time of the whole process.Task clustering combines fine-grained tasks intocoarse-grained tasks,dispatching them to the same resource,and reducing scheduling overheads to optimize process execution time.Unreasonable task clustering process will lead to problem of time and dependency imbalance,causing lower parallelism of task execution.Aiming at the problem of time imbalance,this paper proposes a time equalization clustering algorithm named RBCA,which employs the backtracking method for task clustering,so as to make the running time of clusters more balanced.Meanwhile,we propose DBCA,which defines the degree of similarity between tasks,and classifies tasks with high degrees of dependency into one class,so as to solve the problem of dependency imbalance.This paper employs workflowsim cloud workflow simulation platform for experimental simulation.The experiment results show that the proposed algorithms based on the firefly algorithm and dynamic priority is superior to the traditional cloud workflow scheduling algorithms for both convergence speed and optimal value.Meanwhile,compared with the traditional balanced clustering algorithm such as HRB and HIFB,the clustering algorithms proposed lead to the more balanced results and can optimize the execution time of the workflow.