An energy-efficient virtual machine scheduler based on CPU share-reclaiming policy
Qu Xilong, Peng Xiao · International Journal of Grid and Utility Computing · 2015
With the development of cloud computing, resource virtualisation technology has been widely applied for deploying large-scale IT-infrastructures in many high-performance data centres. However, using virtualisation technology will also lead to some extra performance costs, especially when the cloud system is in the presence of mixed-workloads. In this paper, we present a novel virtual machine scheduling approach, namely share-reclaiming policy, in which the scheduler allows virtual machines to obtain extra CPU shares if they were frequently blocked by I/O interrupted recently. By this policy, short-term I/O-intensive tasks will have more chances of being scheduled so as to compensate their performance losses caused by I/O operations. Extensive experiments are conducted by using various benchmarks; the results show that the proposed policy outperforms existing scheduling algorithm in the terms of energy-efficiency, especially when the virtualised system is in the presence of intensive mixed-workloads.