Research And Optimization Of Energy-aware Dynamic Virtual Machine Management

In recent years,with the rapid development of cloud computing and big data,the amount of data has increased dramatically,and the scale of data center is increasing day by day.These data centers consume a lot of power every year,which will not only increase the operating cost of enterprises,but also have a negative impact on the environment.The energy efficiency of data center is low and dynamic consolidation of virtual machine(VM)is an effective way to improve resource utilization and reduce energy consumption.However,only emphasizing on minimizing energy consumption will lead to performance degradation and cannot ensure that cloud computing providers comply with the Service Level Agreement(SLA)signed with users.Therefore,the starting point of this paper is to reduce the energy consumption and improve the resource utilization of data center by VM dynamic consolidation on the premise of SLA.VM consolidation includes VM migration and VM placement,and this paper focuses on these two aspects as follows:(1)An adaptive multi threshold virtual machine migration mechanism AMT is proposed.Based on the consideration of CPU utilization and SLA,this paper designs different migration strategies for different types of hosts.First of all,a compound threshold is designed for the overload host,which is CPU utilization and SLA violation rate.The threshold is adjusted dynamically according to the fluctuation of load to improve the accuracy of migration triggering.On this basis,a Q-learning based migration trigger mechanism for underload hosts is proposed.The underload hosts are further divided into light underload hosts and heavy underload hosts.The light underload hosts will be used to place the migrated VMs,so as to improve resource utilization and reduce migration times;while the heavy underload hosts will be shut down after migrating all the VMs above,so as to minimize the number of active hosts and save energy.In addition,it is necessary to consider which virtual machines to migrate on the overload hosts.This paper proposes a new virtual machine selection strategy(BM),each time selecting the most suitable virtual machine to reduce the number of migrations and keep the host with high CPU utilization.The experimental results show that our proposed algorithm can optimize resource utilization and reduce energy consumption of the data centers while minimizing the SLA violation rate and the number of migrations.(2)A priority based mechanism PBFD for virtual machine relocation is proposed.The existing VM Placement strategy only considers minimizing the number of active hosts,and does not consider the relationship between power consumption and processing capacity of hosts.The common algorithm to solve the problem of VM placement is heuristic multi-objective optimization algorithm.In essence,this kind of algorithm merges multi-objective into a single objective and so it is difficult to achieve real multi-objective optimization.To solve these problems,firstly,SLA priority is set for the host.According to the partition results of Q-learning algorithm,the light underload hosts are set as high priority,and the normal hosts are set as low priority.When selecting the target host,priority is given to search in the light underload hosts set to ensure the quality of service.On this basis,energy efficiency priority is considered,and the one with the highest energy efficiency priority is selected as the target host for migrated VM.Experimental results show that compared with other VM placement algorithms,the PBFD algorithm proposed in this paper achieves better results in energy saving and SLAV,while effectively reducing the number of migrations.The experiment of this paper is to create a data center in the cloud computing simulation software Cloud Sim,expand the relevant interfaces,implement the virtual machine consolidation algorithm we have proposed,and complete a comparison experiment with other series of virtual machine consolidation algorithms.