Research On Workflow Scheduling,Reliability Optimization And Energy Optimization In Cloud Computing Environment

Scientific Computing,which consists of abundant and heterogeneous calculation tasks and data,has been recognized as an important part of today's scientific research and engineering practice.Scientific workflow is one of the significant solutions for large scale scientific computing.The growing complexity of scientific computing complicates the workflows and enlarges the scale of workflows.This increases the demand of computational resource.Cloud computing is a rising service model of computational resource.Cloud offers computational resource to users according to their demands.And the payment follows a pay-as-you-go manner.Among different cloud service models,IaaS offers users virtualized resources in the form of virtual machine(VM)instances,which satisfies the requirements of workflow deployment and draws attentions form the workflow builders.However,due to the difference between Cloud computing and traditional computing pattern in resource provision manners and billing,to schedule the workflow in cloud,optimize the workflow reliability and other tasks have become more challenging.In this dissertation,several key problems related to the workflow scheduling,workflow reliability and energy management in cloud environments have been studied.The main contributions of our work include:(1)Considering the limitations of regarding the QoS parameters as static constants,which is implemented by most current workflow scheduling methods,this study presents a QoS fluctuation-aware workflow scheduling algorithm.This algorithm based on the QoS prediction data of cloud resources acquired by ARIMA model and intelligent optimization method.This algorithm aims to optimize the total operational cost of the workflow scheduling while satisfying the deadline-constraint.Experiments show that proposed algorithm can achieve higher resource utilization and lower SLA-violation than the state-of-the-art algorithms.(2)Considering the structure features of the workflow and the heterogeneous characters of IaaS resources,this study investigates three active fault-tolerance schemes to fully explore the possibilities offered by the ability of cloud to scale up and out computing resources.(3)This study formally model the problem of optimizing the reliability of a cloud workflow system under budget constraints with three active fault-tolerance schemes belongs to Mixed-Integer NonLinear Programming(MINLP)and prove that they are NP-hard.Finally,we solve the three problems with Constraint Programming(CP).(4)Due to the hardness of CP,we adjust the reliability optimization models based on Max-Min technology.Experiments show that Max-Min-based solutions approach CP-based ones.Moreover,this thesis proposes a meta-heuristic optimization based algorithm,GA-WFT.Experiments show that solutions obtained by GA-WFT approach CP-based solutions and GA-WFT is the most efficient.(5)The energy consumption of the private cloud owned by workflow builders costs a part of total workflow budget in hybrid cloud environment.To reduce the total cost of processing a workflow in hybrid cloud environment,this thesis proposes a method to improve the energy efficiency of cloud data center.This method includes a VM selection method and a VM allocation method.Experiments show that POM-Swap is able to reduce more energy comsuption than the state-of-the-art methods while keeping a low SLA-violation.