Research On Virtual Machine Resource Scheduling Technology Of OpenStack Cloud Platform

Cloud computing technology has received extensive attention in many countries and has achieved rapid development.With the increase in the number of cloud users and the ever-increasing scale of cloud services,the scale of cloud data centers has also increased dramatically.How to better allocate the resources of physical machines and improve the resource scheduling mechanism of virtual machines to improve the resource utilization,improve resource balance,and reduce the energy consumption of the entire cloud system has become a hot issue in the field of cloud computing.Currently,the OpenStack cloud platform can manually migrate virtual machines through the Dashboard management interface or the nova live-migration command,lacking complete autonomous migration of virtual machines and the selection mechanism of target physical hosts.In addition,the built-in destination host selection algorithm in OpenStack is single,only after being filtered by the host and sorted according to the amount of memory remaining,the host with the largest memory remaining amount is preferentially selected as the target physical host.To solve the above problems,this thesis firstly studie s the internal resource scheduling mechanism of OpenStack and analyzes the shortcomings of OpenStack's built-in scheduling algorithm.Then,it studies the scheduling strategies of popular virtual machines at home and abroad in recent years.Finally,an ant colony algorithm based on the initial allocation strategy of the improved virtual machine is proposed.The algorithm can dynamically integrate the virtual machine.The simulation results on the CloudSim show that this algorithm can effectively balance multiple conflicting objectives,which not only improves the resource utilization of the entire cloud platform,but also reduces the power consumption of the entire system.After the cloud system runs for a period of time,the load imbalance occurs between physical servers.Some physical servers increase its load as application increase.Since the resources of each physical server are limited,the increase in the load will lead to a drop in the user's service quality.Some physical servers will run at lower loads as applications decrease.These large number of low-load servers cause a serious waste of power.Therefore,a reasonable virtual machine dynamic scheduling algorithm is needed in OpenStack to supplement the deficiency of the original algorithm.To solve this problem,this thesis proposes a dual-threshold dynamic scheduling algorithm for virtual machines,and then elaborates on the four aspects of system resource monitoring,selection of resource scheduling opportunities,selection of virtual machines to be migrated,and selection of physical migration machines.The migration strategy uses time prediction to select the timing for the migration of the virtual machine,and uses a multi-objective optimization algorithm to select the appropriate physical machine as the target host for the migration.In order to avoid the clustering effect,a physical host selection algorithm based on probability selection was designed.Finally,the simulation on the CloudSim shows that the algorithm has a significant increase in power consumption,user SLA violation rate,and resource balance.