Research On Task Scheduling And Virtual Machine Consolidation In Cloud Computing Environment

Cloud computing can enable users to make access to all kinds of shared computing resources,such as servers,networks,storage and applications,in a ubiquitous and convenient way through the Internet.As one kind of business model,cloud computing provides the users with customizable and on-demand model of obtaining the required resources.More and more users trend to use cloud resources to resolve their own business,which makes cloud computing becoming one of the hot topics in the field of information and communication technology.Some factors of cloud data center bring challenges to researches on task scheduling and virtual machine consolidation in cloud environment,which attracts much attention of industry and academia.These factors include resource heterogeneity of computing nodes,dynamic and massive natures of users' tasks,fast growing scale of the cloud data center,and so on.Although some efforts on the researches of task scheduling and virtual machine consolidation,there are still a series of critical problems remaining to be further studied,such as the following four points.First,most of the existing studies on job scheduling in cloud environment only address the incentive for one party(e.g.,either the users or the cloud provider),which can't guarantee the incentives for both parties.Second,most of the existing studies on workflow scheduling don't consider the fact that failures may happen on computing nodes.The scheduling models and algorithms obtained by these studies can't reflect the real-life scenarios of modern cloud data centers.Third,most of the existing studies on virtual machine consolidation don't consider load balance of active servers,which is an important factor that should not be ignored.Fourth,most of the existing studies on virtual machine consolidation adopt resource-based factor or the number of migrations to choose the virtual machines needed to be migrated,without considering the migration cost of virtual machines.These algorithms can reduce energy consumption to some extent,but may also lead to significant costs of virtual machine migrations.According to the above-mentioned problems,this thesis investigates the researches on task scheduling and virtual machine consolidation in cloud computing environment and proposes some optimization models and algorithms.The main research works of this thesis are as follows:1.To address the first problem mentioned above,this thesis investigates the problem of task scheduling in cloud computing environment considering incentives for both users and cloud provider.First,a multi-objective optimization model,which considers the incentives of both parties,is proposed.Second,a heuristic scheduling algorithm called greedy-based dynamic price scheduling algorithm is developed.The developed algorithm adopts dynamic pricing mechanism,which makes the prices of all computing nodes satisfy the market price rules of commodity.The experimental results show that,in most cases,the developed algorithm could satisfy the incentives for both parties better.2.To address the second problem mentioned above,this thesis investigates the problem of workflow scheduling in cloud computing environment by considering the probability that computing nodes may fail during execution.First,this thesis deduces the calculation formula of both the expectations of workflow makespan and workflow execution cost.Second,a multi-objective optimization model is proposed to minimize both of the two expectations.Finally,a min-min based cost-time weighting algorithm is developed.The experimental results show that,as expected,failures of computing nods and fault recovery indeed have impact on the two expectations,and the developed algorithm can tradeoff the two optimization objectives better.3.To address the third problem mentioned above,this thesis investigates virtual machine consolidation by taking load balance into account.First,a multi-objective optimization model is proposed,in which the optimization objectives are minimizing the number of active computing nodes and balancing the loads among these computing nodes.Second,a greedy-based load balance consolidation algorithm is developed.The experimental results show that the developed algorithm can reduce the number of active servers and achieve the best load balancing level at the cost of a few more migrations.4.To address the fourth problem mentioned above,this thesis investigates the problem of migration cost aware virtual machine consolidation.First a virtual machine consolidation model with multiple constraints is proposed.Second,a migration cost aware virtual machine consolidation algorithm is developed.To reduce the migration costs of consolidation process,the developed algorithm chooses the virtual machine to be migrated with the minimum value of cost factor.The experimental results show that,to some extent,the developed algorithm can reduce the number of active servers,as well as the migration cost.