Migration-less Energy-Aware Task Scheduling Policies in Cloud Environments

Abstract

Burgeoning market for cloud applications has brought significant interest for cloud providers. They look for methods to maximize their profit margins by executing higher number of jobs while minimizing the operational costs and environmental footprints. Energy consumption of servers plays a crucial role in operational costs. There have been different methods to deal with energy consumption from hardware level to higher layers of abstractions such as compilers and operating systems. For example, several resource-scheduling policies have been proposed to not only regulate energy consumption but also guarantee Service Level Agreement (SLA). One of the prevalent techniques in reducing the total energy consumption in data-centers is through consolidation of virtual machines (VMs). In this technique, one or several VMs migrate from a physical node to other nodes, which in turn incurs a significant overhead not only on a server but also on the network infrastructure of a cloud. To address this problem, we propose a VM scheduling algorithm based on the unsurpassed utilization level to come up with optimal energy consumption while meeting a given QoS. In other words, our proposed algorithm aims to regulate execution speeds of VMs on a host with a result that the host works at its optimal energy level. In fact, a host is scheduled to run its allocated tasks faster to reach the optimum level of utilization instead of migrating its tasks to other hosts. We also propose several task scheduling policies to adjust execution speeds for real-time tasks in each VM. The simulation results show that the proposed task scheduling policies not only reduce the total energy consumption of a cloud by 19%, but also have profound impacts on turnaround times of real-time tasks by 27%.