Power Model Of Live Migration Based On Elastic VM Pool

As infrastructure of information system, data center has been employed increas-ingly widely with the development of e-government, finance, telecommunications, etc.However, data center also has high power consumption and environmental pollution,which has been concerned by all over the world. With the rapid adoption of virtualiza-tion technology, it has been considered as one of the key technologies to build greendata center, and improve its energy efficiency. Recently most resources and energymanagement solutions are based on virtualization technology. However, in practice,the deployment of virtualized power management solutions is lack of systematic re-search on power model of it, so the management solutions cannot effectively integratedwith data center. Therefore, the study of power characteristics in virtualized environ-ment is important for virtualized power management.The topic comes from National Natural Science Foundation program. This projectintends to build a large-scale green data center test bed based on virtualization tech-nology, assess QoS, power efficiency for different transactions, and make software andhardware solution. In order to make the integrated solution Elastic VM Pool effec-tively deploy, this topic study the each main index of Virtualized data center and powerconsumption of virtual machine live migration.The paper's main achievements include:1. Energy monitoring visualization prototype platform: Build Energy monitoringvisualization prototype platform of the whole data center, according to the strategyof power capping, dynamically adjust the power. Collect heterogeneous multi-sourceresources, energy consumption, cooling and other data, in order to set up intelligentdata monitoring and collection platform.2. Single node power model: Each single node to the data center server as moni-toring units. Employing the real-time monitoring data of each unit, study every index that affects the server power consumption, such as CPU usage, frequency, and mem-ory, etc. Find the mathematical modeling.3. Power cost of VM live migration in clouds: Virtualization technology makesenergy management through physical resource consolidation and resource dynamicalallocation, etc. The study on VM live migration can effectively guide the deploymentof live migration technology. Finally, we try to modeling the power usage of VM livemigration.By studying on each power affection index of virtualization technology, the powermodel of single node in the data center predicts the power within 3% of the real power,and the deviation between the power model of live migration and real is less than 5%,which provides a good quantitative basis to power management based on virtualizationtechnology. Additionally, experimental data has also showed that, the power overheadof migration is much less in the scenario of employing the strategy of consolidationthan the regular deployment without consolidation.