Research On Apaptive Mechanism Of Virtual Machine At Runtime In Cloud Computing

Cloud Computing pools software and hardware resources based on virtualization technology, and provides service to user on demand. It has advantages such as high scalability, high reliability and flexibility, and improves the efficiency of resource utilization. Because of this, it has become a research focus in information technology area. However, the workload of virtual machine and application request will change with time. When difference between pre-allocation and the actual request amount is too large, it will have a negative impact on the quality of service. Although we can manually resize or migrate virtual machine, however, considering with large scale datacenter, these work cannot be done just by manpower. Based on this, this paper will focus on how to automatically reallocate resources in cloud environment based on the running state of virtual machine, so as to ensure the on-demand allocation feature at runtime.This paper presents an adaptive management mechanism of virtual machine, so that the allocated resource capacity will match the actual meet. This mechanism is composed of three parts:(1) Resource monitoring. It will collect and store the realtime performance data of monitor target, from physical host to virtual machine, which is the foundation of subsequent modules. With the aid of Nagios monitor software, it exposes the monitor configuration details and performance data through REST API.(2) Alarm analysis. It will match the performance data with predefined alarm rules. Because the analysis is based on a certain range of time, it will minimize the negative impact of workload fluctuation brought by virtual machine.(3) Decision making and execution. This part is responsible for selecting appropriate physical host as the migration target of virtual machine. We complete the resize or migrate operation with the help of OpenStack programming interface. It also provides several available provision strategies, and reduces the switching cost to minimum.In addition, this paper designs a simulation scheme based on CloudSim simulation framework. It compares all the related provision strategies from three aspects:energy consumption, number of migrated virtual machines and service level agreement violation. Finally, we verify the functionality of adaptive virtual machine management system. The result demonstrates that when virtual machine’s workload changes at runtime, the system can respond in time, generate reasonable alarm information, and adjust dynamically. This reflects the characteristics of on-demand allocation at runtime.