Research On Virtual Machines Dynamic Consolidation System Based On A Kth-Order Combined Markov Model

The rapid development of cloud computing technology has brought a new way to manage data center, and many technology companies are accelerating to set up their own data center, which also brought a lot of electrical energy consumption. At present, the worldwide electric energy consumed by a data center has been rising, while the resources of data center such as CPU utilization is often less than half, which resulted in a large number of servers is idle. Cloud computing platform is actually a virtual machine management platform, and computing resources of data center are provided by the virtual machines, so many busy virtual machines can be integrated into partial server, and other idle servers can be switched to a low-power mode, thereby reducing energy consumption of data center.VM integrated system typically includes two cases, and one is to choose an appropriate time to migrate out a virtual machine to reduce energy consumption and the number of virtual machine migrations; the others includes the selection of migrated virtual machine and the search of placement virtual machine. Traditional virtual machine migration trigger strategy is mainly based on the current load of host or migrating virtual machine periodically. Virtual machine migration trigger strategy based on time series forecasting is faced with the problem of low prediction accuracy, and the current research on virtual machine dynamic consolidation on real cloud platform also does not take into account its influence on the cloud platform system itself.To solve the above problems, we deeply study the All-Kth-Order Markov model and consider the influence of CPU state between different moments, and propose a Kth-Order Combined Markov Model. The Kth-Order Combined Markov model is used to predict the stats of CPU load in next period of time. The dynamic consolidation system designed in this paper can be run on an open-source cloud platform like OpenStack. The system is a three-level model, and it adds a host as a management node, which is used to locate a suitable virtual machine placement and send request of virtual machine migration. Finally, we design detail experiments for virtual machine migration trigger strategy based on Combined Markov Model to compare with the traditional method. The results show that the strategy proposed in this paper can decrease the number of virtual machine migration and data center power consumption, a large number of experiments also show that the system designed in this paper can manage virtual machine consolidation on real cloud platform effectively.