A Prototype of Online Dynamic Scaling Scheduler for Real-Time Task based on Virtual Machine

Abstract

In this paper, we provide an architecture of power-aware scheduler for real-time virtual machine system using dynamic frequency scaling mechanism This architecture provides that how to manage real-time resource and how to control processor frequency. In addition, we propose two scheduling schemes that utilize slack time to adjust processor frequency without violating tasks deadline. Based on the provided architecture, we implement a virtualization framework with online power-aware scheduler using RT-Xen real-time hypervisor and Litmus-RT real-time OS. Our implementation manages entire of real-time resource and processor frequency depending on system policy. For trasferring real-time requirements, we implement an interface using hypercall mechanism. To evaluate provided system, we analyze performance evaluation in various aspects.