A Scheduler Considering Characteristics of VM for Mitigation of Unstable Performance
Bongwoo Bak, Jihun Kang, HeeSeok Choi, Jaehak Lee, Heonchang Yu, Kwang-Sik Chung · 2018
As virtualization technology evolved, IaaS providers became able to provide virtual machines(VM) with increased utilization of physical resources. The most important aspect of providing such a cloud service is to ensure the time it takes to process the work. However, in the current virtualization environment, since several VMs share limited physical resources, competition between VMs is occurred and results in deterioration of the network performance. There is also the disadvantage that performance is unstable and unpredictable in handling tasks due to irregular I/O operations. In this paper, we propose a scheduler to stabilize performance and improve network performance in virtual environments. We monitor and classify the characteristics of each VMs, prioritize I/O operations, and schedule them over CPU jobs. As a result of this study, IaaS providers are able to guarantee a1 stable service to users.