Towards an Adaptive Multi-Power-Source Datacenter

Abstract

Big data and cloud computing are accelerating the capacity growth of datacenters all over the world. Their energy costs and environmental issues have pushed datacenter operators to explore and integrate alternative energy sources, such as various renewable energy supplies and energy storage devices. Designing datacenters powered by multi-power supplies in the smart grid environment is becoming a promising trend in the next few decades. However, gracefully provisioning various power sources and efficiently manage them in datacenter is a significant challenge. In this paper, we explore an unconventional fine-grained power distribution architecture for multi-source powered datacenters. We thoroughly investigate how to deliver and manage multiple power sources from the power generation plant outside of the datacenter to datacenter inside. We then propose a novel Power Switch Network (PSN) for datacenters. PSN is a reconfigurable multi-power-source distribution architecture which enables datacenter to distribute various power sources with a fine-grained manner. Moreover, a tailored machine learning based power sources management framework is proposed for PSN to dynamically select different power sources and optimize user-demanded performance metrics. Compared with the conventional single-switch system, evaluation results show that PSN could improve solar energy utilization by 39.6%, reduce utility power cost by 11.1% and improve workload performance by 33.8%, meanwhile enhancing battery lifetime by 9.3%. We expect that our work could provide valuable guidelines for the emerging multi-power-source datacenter to improve their efficiency, sustainability and economy.