Research On Service Component Migration Method Oriented To VM Performance Guarantee

With the dynamic changes of the user request,there is too high or too low problems about the utilization of virtual machines(VMs).When the resource usage is too high,it will result in system crash,poor stability and system halting;while the low usage will bring about imbalance of resources income and energy consumption.Some research found that when the resource utilization is 10%or so,the power consumption would also be more than 60%of that the full load.It is essential for researchers to keep the resource utilization in a certain range.Thus,this thesis research performance guarantee of VM from the perspective of resource utilization.With the development of cloud computing and the demand increase of service,the component system for cloud services emerges.Service component becomes independent deployment unit and functional unit,and multiple components share one VM.On the basis of the previous virtual machine migration and service migration,this thesis proposes service component migration to guarantee the performance of virtual machine,including the migration trigger judgment,selection being-migration components and choose target virtual machine.Firstly judge whether the VM need to migrate component according the resource utilization of VM.If triggered,components are chosen by comprehensive evaluation on basic resource requirement,the correlation degree between component and resource,service quality and service availability.Then employing the improved genetic algorithm to select the appropriate target VM.Finally,the correctness and validity of the components migration method this thesis proposed are verified through the experiment.Contrast experiments have verified the component selection algorithm can choose the appropriate components for VMs violated trigger rules,and target virtual machine selection algorithm can select the accurate target machine for relevant components.Prove that migration method can effectively guarantee resource usage of virtual machines within a certain range.