Augmented Data Center Infrastructure Management System for Minimizing Energy Consumption

Abstract

The significant continuous increase in the power consumption of data centers, brought about by the adoption of cloud computing, is a major social problem. Moreover, the demand for data centers is expected to grow owing to the increasing requirement for data processing related to Internet of Things (IoT) applications. Consequently, reducing the power consumption of data centers has become an urgent and major issue. Modification of the operational settings of equipment residing in a data center leads to fluctuations in the power consumption of other equipment due to their complex interdependencies. Therefore, to reduce the total power consumption of the data center, it is not sufficient to reduce the power consumption of each piece of equipment, but one should consider the dependencies among the data center equipment. In other words, a coordinated control of the operational settings accounting for interdependencies among equipment is required to reduce the total power consumption of data centers. In this study, a data center energy management system focusing on the coordinated control of equipment in a data center is proposed. In the proposed system, the operational settings are determined using power consumption prediction that is based on the machine learning methods. We design a data center control system focusing on the task assignment of servers and air conditioner settings, and this system determines their settings based on a genetic algorithm. The proposed system is evaluated by experiments in our data center testbed and the evaluation results indicate that the proposed system reduces the total power consumption of the data center.