Research On Resource Allocation Strategy Optimization Based On Openstack

For years, with the rise of cloud computi ng, the scale and quantity of Data Center are increasing dramatically. Due to the huge power consum ption of the data center ’s equipment, how to reduce the ener gy consumption of a data center server host and the entire equipment of a data center has become a hot issue in IT industry.With the development of virtualization t echnology, of which the virtual m achine migration technique can be used as a soluti on to the ener gy consumption of the cloud data center. When the utilization of a server resources is so low that energy consumption is much higher than its ability to p rovide services, it would result in waste of resources. There are a lot of servers in th e idle state in a lar ge-scale data center every day, if we can avoid such a situation, the total ener gy consumption of a data center will optim ize largely. In addition, when the utilization of a server resources is so high that wil increase the energy consumption of the server, and cannot meet the requirements of some virtual machine of its to run, it w ill reduce the server’s energy consumption increased by high load and relieve the shortage of the se rver resources through the virtual m achine migration.Secondly, the virtual m achine migration is im plemented in Open S tack only through the comm and nova live-m igration currently, or the manual m igration function in Open Stack management page. It lack of the autonomy of the virtual machine migration timing and the tar get virtual m achine migration selection. Moreover, the algorithm of the target host selection in Open Stack’s Nova-Scheduler module is single, which just select th e target host preferen tially according to the amount of m emory remaining sorting after the host filter.So as to solve above problem s, based on the research and analysis of the popular virtual machine migration strategy both at home and abroad in recent years, this thesis proposed an optimized virtual machine migration strategy combined with Open Stack. It firstly analyzed the existing resource allocation of Open Stack, and then analyzed its live migration strategy, seriously the analysis of the Open S tack’s Scheduler m odule. By optimizing the virtual m achine migration strategy and the Open S tack’s Scheduler module, it seriously improved the positioning of the tar get host in the Scheduler algorithm. Because th e Scheduler does not consider the invalid m igration when selecting the target host, so that the target host selection strategy put forward in this thesis, by setting an appropriate thresh old and com bining time series prediction technology, avoided the invalid migration and saved its energy.Finally, this thesis simulated a cloud data center using Cloud Sim sim ulation software. Through the modification of Cloud Sim relevant classes, on which achieved the virtual m achine migration algorithm proposed in this thesis. The ener gy consumption data of the virtual m achine migration has been associated in experim ents. By analyzing these data, it proved the algorithm can optimize the energy consumption of a data center largely.