EONS: Minimizing Energy Consumption for Executing Real-Time Workflows in Virtualized Cloud Data Centers

Abstract

Cloud computing is revoluting IT industry, and more and more workflow applications in science and engineering fields are shifting to cloud. However, with the rapid expansion of host volume in cloud data centers, increasing energy and related operating and environmental costs have become a major concern. Therefore, many energy-efficient workflow scheduling approaches have been proposed. Unfortunately, they are typically in low resource utilization and poor energy efficiency, because they allocate workflow tasks to hosts for execution roughly overlooking the fact that a single workflow task can hardly utilize a host's resource fully. To address this issue, we first propose a novel scheduling architecture for a virtualized cloud data center. Based on the scheduling architecture, we develop an energy-efficient online scheduling algorithm, EONS, for real-time workflows. Furthermore, in order to improve the energy efficiency, three strategies for scaling up and down the computing resources are proposed and integrated into EONS to balance weighted square frequencies of hosts. We have compared the performance of EONS with three existing algorithms in the context of various real-world scientific workflows. The experimental results show that EONS achieves a better performance in terms of energy saving and resource utilization while guaranteeing the timing requirements of workflows.