Power-Aware Real-Time Virtual Machine Schedulers in Discrete DVFS Systems

Power-aware computing has become an important problem not only in embedded systems but also in large-scale data centers in Cloud computing environments. Many recent studies have focused on power-aware virtual machine allocation or consolidation. However, few research has been done about implementation of power-aware schedulers in VMM (Virtual Machine Monitor). In this paper, we provide a prototype design for power-aware schedulers in RT-Xen hypervisor. The proposed system is based on compositional real-time scheduling framework which provides a method to abstract each virtual machine into a single component. The proposed prototype includes how to transmit VM real-time requirements into VMM and dynamic frequency control based on those requirements. Throughout the experiments, we analyze the proposed prototype and show the performance evaluation.