Research On Dynamic Co-scheduling Based On Virtual Machine Status

Virtualization can efficiently manage and consolidate server resources and is a core technology for cloud computing. So improving the perfor-mance of virtual machines is important for virtualization and even for cloud computing. In this paper, we focus on the LHP(Lock Holder Preemption) problem in virtualized environment which lead to performance degrada-tion of virtual machines. Existing solutions improve virtual machine per-formance by coordinate scheduling(co-scheduling) the virtual machines. But they can not apply to all scenarios when workload in the virtual ma-chine changes. Therefore, we propose a solution that dynamically adjust the co-scheduling algorithm for a virtual machine based on its running state and thus effectively improve the performance of all the virtual ma-chines. We classify the running state of a virtual machine based on it-s Pause VM-Exit frequency. And according to the classification of the virtual machine, we apply traditional co-scheduling algorithm and PLE based co-scheduling algorithm to the virtual machines. Experimental re-sults show that our proposed solution can effectively improve the perfor-mance of virtual machines and have significant performance improvement comparing with existing solutions.