Virtual Machine Dynamic Resource Allocation And Placement Algorithm Research

Since the concept of "Cloud Computing" being presented, it becomes the most exciting research focus rapidly in the field of science and technology, and is attracting more and more attention of the industry and the academic. As a new strategic technology, there is not yet a unified definition, but academia and industry is generally thinking that cloud computing is a virtualization technology-based, pay-as-you-go, business model, with flexible expansion, dynamic allocation, resource sharing, and other characteristics of the new networking computing.IaaS (Instruments as a service) is the basic and supportive one of cloud computing services. And the problems of IaaS will be focus by this paper. IaaS shares the physical resource pool by virtualization technology, and users lease the virtual machines from operators to host the applications. With the operator’s business development, it will inevitably lead to a large concentration of instruments resources (such as servers, network, storage, etc.) in the cloud IDC. The physical resource allocation technology will have a direct impact on the overall physical resource pool utilization and SLA (Service-level Agreement).So, this technology is the crucial problem which must take more energy to optimism. If we lack of research for this technology, it will direct impact the effectiveness and concentricity of the cloud computing resources, decrease SLA, waste the computing resource. We believe that, with the cloud computing developing, more and more universities and institute will do more in-depth studies. In traditional cloud computing, it could dynamic increase or decrease resource of one application, but usually by coarse-grained approach which uses virtual machine as the unit. In another aspect, traditional researches on the problem of virtual machines placement, gave more attentions to improve the resource utilization, and ignored the dependence in multi-tier applications. This will lead to network overload after migrate the overload virtual machine.After deep research, this paper presents dynamic resource allocation based-on load prediction algorithm and application-aware virtual machine placement algorithm. Dynamic resource allocation based-on load prediction algorithm can predict the resource usage of virtual machines by monitoring the load of each virtual machine, and dynamic reallocate resource for virtual machines on local physical computer. So this algorithm can more effective and fine-grained utilize local physical resource, decrease the probability of migration. Application-aware virtual machine placement algorithm uses grouping genetic algorithm and considers dependency in different applications. So, it can integrated virtual machine placement more effectively and decrease the use of physical machine for energy saving.