ERECT: Energy-efficient reactive scheduling for real-time tasks in heterogeneous virtualized clouds

摘要

High energy consumption in large-scale cloud data centers has become a burning issue, and efficient task and resource scheduling is an attractive way to cut down their energy consumption while providing satisfactory services for the customers. Unfortunately, existing scheduling approaches do not fully exploit the heterogeneity of real-tasks and physical hosts for maximum energy savings, while guaranteeing the timing requirements of real-time tasks. To solve the above problem, in this paper, we firstly develop a novel scheduling architecture that transforms the dynamic scheduling problem into multiple static schedules. Then, we propose an energy-efficient reactive scheduling algorithm, namely ERECT, to schedule the real-time tasks and computing resources in virtualized clouds. The proposed algorithm ERECT fully consider the heterogeneity of the real-time tasks and the hosts. In addition, when adding and deleting the virtual machines (VMs), the optimal operating frequencies and energy efficiencies of heterogeneous hosts are exploited to achieve energy conservation. Finally, in order to demonstrate the effectiveness of our approach, extensive experiments are conducted to compare ERECT with two base-line scheduling algorithms in the context of Google traces. The experimental results show that ERECT outperforms those two existing algorithms in terms of guaranteeing tasks' deadlines (up to 14.06%) and energy saving (up to 9.81%). (C) 2017 Elsevier B.V. All rights reserved.