The Balancing Optimization Of Workflow Scheduling In Cloud Computing Environment

With the development of cloud computing,workflow systems can gain powerful computing and expansion capabilities in the cloud environment,and the “pay-as-you-go” usage model of cloud resources greatly reduces payment costs.In addition,workflow support for abstract definition,flexible configuration,and automated operation of cumbersome applications can effectively improve cloud resource utilization.However,the demand of users is increasing day by day,especially the application of workflow technology in the cloud environment,which further makes process more complicated.In particular,workflows in some business areas have many tasks,large scale,and intensive instances,which brings great challenges to cloud workflow scheduling research.As we all know,establishing the best mapping between tasks and virtual machines is an NP-hard problem.Furthermore,compared with traditional scheduling,the tasks dependence increases the complexity of scheduling allocation.Especially when a large number of instance-intensive tasks arriving at the same time,low-cost and high-quality virtual resources are frequently called,resulting in low scheduling efficiency and poor resource utilization.Aiming at the above situation,we propose a Two-Phase Workflow Scheduling Optimization(2PWSO)strategy including pre-scheduling optimization and dynamic scheduling optimization.Based on the strategy,a cloud workflow scheduling and monitoring platform is designed.The platform can intelligently schedule workflow and monitor the resource consumption to visually demonstrate the feasibility and effectiveness of the proposed scheduling strategy.The main innovations of this paper include:(1)A pre-scheduling optimization method is proposed.Different from the traditional static scheduling method,the Improved Shuffled Frog Leaping Algorithm(ISFLA)apply time greedy strategy to optimize the initial population quality,thus improving the search efficiency of the optimal solution.In addition,a reconstruction strategy is put forward to go out of the dilemma of local optimum.The experimental results show that ISFLA is superior to the traditional shuffled frog leaping algorithm and particle swarm optimization algorithm.(2)A dynamic scheduling optimization method based on load sensing is proposed.Different from the existing dynamic scheduling methods,we apply load sensing mechanism to the scheduling process.By establishing load and resource selection model,the proposed Candidate Queue Generation Algorithm(CQGA)and the Dynamic Selection Algorithm(DSA)is used to search candidate resources,thus balancing the vm load in task execution.(3)Based on the above optimization method,a two-phase scheduling optimization of cloud workflow is proposed.On the basis of pre-scheduling optimization,dynamic scheduling optimization based on load sensing is implemented.On the one hand,a dynamic monitoring mechanism is applied to the 2PWSO during task execution,which improves the shortcoming of static scheduling when multiple workflows arrive in parallel,which is prone to excessive single-load overload and slow down overall execution efficiency.On the other hand,based on the one-stage pre-scheduling result,2PWSO can avoid the limitation that conventional dynamic scheduling can not take into account global constraints.(4)Based on the above research,a cloud workflow scheduling and monitoring platform is designed and implemented.The platform can schedule the workflow and dynamically monitor the resources during task execution to evaluate the effectiveness of the scheduling strategy.