Data Centers Energy-saving Research Based On Virtual Machines Scheduling And Adjustment

Cloud computing marks the milestone of information technology development. It allows users to purchase computing resources based on users’ demand as purchasing water and electricity. Being led by Amazon, many IT companies have launched their own cloud computing and computing centers containing thousands of compute nodes. Meanwhile, the energy consumption issue of data centers is becoming a hot button. A significant consumption of energy not only leads to more operation costs but also emits large quantity of carbon dioxide which causes environmental problems. The power consumption used in the data center accounts for 1.1% to 1.5% of the total electricity consumption worldwide in 2010. We forecast a growing trend of ratio in the future.This thesis first summarizes energy-saving technologies used in data centers. Then I subdivide the existing energy-efficient technologies into several levels. The scope of this study is energy saving technology at the data center level based on virtual machines scheduling and adjustment. Then I model energy consumption of computing virtual machines scheduling which placement virtual machines to servers in data center, proposed minimizing energy consumption by minimizing total busy time of all selected servers. A 2- approximation algorithm named LSF is proposed to address this NP hard problem and it’s proved be effective to this problem by simulation experiments.This thesis finally focuses on OpenStack energy-saving problem, including researching functional models related to energy-saving based on virtual machines adjustment such as virtual machines migration and automatic scaling group. Followed by that, I discuss the direction of energy-saving researches in OpenStack with these models. A model to save energy consumption of data centers will be designed and implemented through a virtual machines migration. To address the imperfectness of OpenStack native virtual machines scaling policies in terms of energy saving, this thesis designs a policy named ExactUtilization to save energy consumption, which is proved effective based upon simulation experiments.