Forecasting data centers power consumption with the Holt-Winters method

Abstract

Data centers are rapidly increasing in the last few years and this trend will continue in the future due to the market demand for virtual infrastructures, cloud services and IoT applications. The whole ICT sector is faced with an energy efficiency challenge and data centers represent a significant share of all ICT-related emissions. Improving energy efficiency has therefore become a critical objective for ICT infrastructures and equipment suppliers. To mitigate the electrical energy demand used both for computing and for cooling, many scheduling and planning methods have been proposed. All of them can benefit from accurate predictions of the workload. In this paper we demonstrate how forecasting tools can remarkably increase the energy efficiency of a data center.