Research On Deploying And Scheduling For Cloud Resources Based On Virtual Machines

Cloud computing has aroused widespread concern as a new delivery and usage pattern of the sharing resources via internet. Virtualization technology enabled cloud platform to allocate the resources in the form of on-demand services and build a dynamic virtual resource pool. However, the increment of the scale of the cloud, the diversity of the cloud services, and the heterogeneous of the resources had brought enormous challenges for the management of the virtual resources. Currently, most cloud packaged the computing and other resources as virtual machines(VMs) to the cloud users, so the optimization on deploying and scheduling strategies for cloud resources based on VMs is worth studying.A good strategy will improve the resource utilization and reliability and reduce energy consumption. So the study about this issue has important academic and practical significance for cloud computing technology.The main challenge of the study is to provide the VMs on-demand and schedule the tasks appropriately while meeting the user requests. For example, the adaptive VMs management for considering both the user dynamic needs and multi-dimensional resource synergies; the VMs dynamic management of cloud for balancing the user requests and other factors on clouds; the task scheduling for considering task deadline and VM load. However, research about the above issues is relatively lack and has just got started.The study in the paper, addressing the above issues, combines the global optimization of VMs deploying and scheduling methods with the user dynamic requests. The research contents and innovations are as follows:1. A self-adaptive VMs management framework, named as AWS-VMMF-DR, is proposed to consider both the user dynamic needs and multi-dimensional resource synergies. It includes three key jobs : building the resource model with weighting the attributes of the cloud nodes with the Analytic Hierarchy Process method; building the integrated load measurement model for the physical nodes; introducing a self-adaptive managing process with high dynamic extension.2. A VMs placement method named as UREBP-DF is presented for balancing the user dynamic requests and the energy consumption. It has two novel algorithms: a placement algorithm named as RCBFDHP. a host activation algorithm named as PFHA. Simulation shows that it could ensure the load balancing to be more effective with considering the dynamic user requests while reducing the energy consumption, comparing with the traditional polling strategy and the BFDSum strategy.3. A dynamic VMs management method named as PEBDM-DR is put forward for balancing the energy consumption and the performance of the platform. It includes three crucial parts: an over/low load positioning method based on the dynamic feedback; a power first under load consolidation method; a relative capacity- based migration algorithm. Simulation shows that it could achieve optimal results on the indicators such as the loss rate of user requests, the total energy consumption of cloud, comparing with the other dynamic management strategies.4. A tasks scheduling method considering task deadline and relative load on private clouds is proposed. Its contributions are as follows: building the task and VMs model; introducing a relative load model; putting forward an improved Min-Min task scheduling algorithm based on the left time of tasks and the relative load between the tasks. The experiments shows that it would improve the load balancing of the cloud and reduce the violation of user requests when the number of tasks increased with the high heterogeneous degree, comparing with other algorithms.Overall, the study closely links to the user dynamic requests guarantee, platform performance and energy balancing optimization, proposes a self-adaptive VMs management framework and VMs deployment and dynamic management method as well as a task scheduling method. The study will contribute to the further development of the cloud computing; bring new solutions for the research on deploying and scheduling of virtual resources on clouds.