Research Of Virtual Machine Placement In Cloud Data Center

In recent years, with the rapid development of cloud computing, more and more users choose to deploy their applications on the cloud data center. As a result, the cloud data center needs to manage a great number of virtual machines. How to choose the target server to the virtual machine, namely virtual machine placement problem has been widespread concern.On the basis of in-depth study and summarize the existing achievements in re-lated fields, in this paper, around minimize the communication cost between virtual machines, reducing the waste of physic machine’s resource and minimize the cost of virtual machine migration of virtual machine placement problems, we mainly carried out the following research work:Firstly, according to the characteristics of the data center network topology and communication characteristic between virtual machines, we build communication cost matrix for virtual machines. At the same time, we model the server resources. Those are the basic work for virtual machines placement policy.Secondly, we propose a hierarchical virtual machine placement framework, we combine the virtual machine initialization and runtime placement to solve the problem of virtual machine placement. In the initialization virtual machine placement phase, we determine the placement policy due to the type of task of tenants and consider the virtual machine belong to the same tenant as a whole, ultimately place virtual machines on the server. According to the operational data of servers and virtual machines, we find the virtual machines that need to reposition and find the target servers of those virtual machines due to the virtual machine runtime placement algorithm.Finally, this paper designs and implements a virtual machine placement component which consists of two modules:data collection module and virtual machine placement policy module. Data collection module is developed based on open source software OpenStack’s monitoring components Ceilometer. We extend Ceilometer so that it can collect the required data of virtual machine placement policies. And we use the virtual machine placement policy we proposed in this paper as the virtual machine placement policy module. Then we integrated the component to our experiment platform. Exper-iments show that the proposed virtual machine placement algorithms achieving good results.