Power and thermal-aware virtual machine scheduling optimization in cloud data center

Abstract

Virtual machine (VM) consolidation technology is a commonly used energy-saving method in data centers. Most of the existing methods are committed to consolidating VMs to a small number of servers to improve the utilization of server resources and reduce the total energy consumption while preventing hotspots. However, energy and thermal-aware VM scheduling is a multi-objective optimization problem. Most of the existing related work cannot provide adequate means to adjust the impact of energy consumption and temperature on VM scheduling. Hence, this paper proposes a power and thermal-aware VM dynamic scheduling scheme (PTDS) for cloud data centers. The proposed PTDS dynamically adjusts the VM consolidation scheme by real-time detecting the server temperature and resource utilization rate under the premise of considering the thermal cycle effect of the data center computer room. Power and Thermal Objective Ant Colony Optimization (PTOACO) is proposed in the VM placement. PTOACO improves the defect that the ant colony algorithm easily falls into local optimization and adds control parameters to adjust the bias between sub-objectives. We performed extensive experiments by using real PlanetLab and random workloads. The performance results were compared with several advanced schemes regarding total energy consumption, hotspots, SLA violation rate, etc. The experimental results demonstrate that PTDS reduces energy consumption by 26.69% on average compared with other advanced schemes and ensures a meager SLA violation rate while avoiding hotspots.