Research On Vm Migration Optimization.based On Combination Of Resource Affinity And Reinforcement Learning

As a new model that relies on the Internet infrastructure to provide users with ondemand computing,network,storage,platform and application services,cloud computing has become the research direction of many scholars.With the widespread application of cloud computing technology,the problem of energy consumption in data centers has become increasingly prominent and severe.Reasonable resource allocation can effectively reduce the energy consumption of cloud data centers.For a virtual cloud data center,the location of the virtual machine determines the number of active hosts.The virtual machine migration technology migrates the virtual machine to a more suitable location by reorganizing the existing resources of the host,making full use of the resources of the host to achieve the purpose of reducing energy consumption.This paper focuses on the virtual machine migration technology,and analyzes and compares the key issues of the current virtual machine migration technology.In view of the current situation that the host resources are difficult to be effectively utilized and the energy consumption of the data center is too high due to the current dynamic and unstable cloud data center,the virtual machine migration process is divided into the selection of migration timing,the selection of the virtual machine to be migrated and the placement of the virtual machine.There are three parts in the selection of the host,which uses the characteristics of reinforcement learning to dynamically interact with the environment to solve the problem of unstable cloud data centers.At the same time,combined with the resource affinity,a series of solutions are proposed for the optimization of host resource utilization and energy consumption.To address the problem of how to optimize the VM migration process,this paper proposes a VM migration strategy based on a combination of resource affinity and reinforcement learning.Considering the differences between the two phases of VM selection and VM placement,we define the selection resource affinity and placement resource affinity between VMs and hosts respectively,which are used to measure whether the utilization of multidimensional resources of hosts tends to the equilibrium state when VMs are moved out from or placed in a host.The similarity of the two phases of VM selection and VM placement,i.e.,both in dynamic and unstable cloud environments,is also considered to model the problem using reinforcement learning for the purpose of reducing energy consumption and resource waste.In addition,during the stage of virtual machine selection,The ratio of utilization is converted into a number from 0 to 100 and used as the state space of reinforcement learning.This strategy shields the specific instance of the virtual machine and avoids the use of reinforcement learning to model virtual machine selection due to the continuous migration of virtual machines.When the action space is constantly changing.When the agent selects the virtual machine to be migrated,in addition to considering the energy consumption reward learned based on historical experience,it also considers the selected resource affinity between the current host and the virtual machine.Finally,the effectiveness of the proposed strategy is verified on the cloudsim simulation experiment platform.