Search eceee proceedings

Policy measures for energy efficiency of data centres

Panel: 8. Products, systems and technologies to decarbonise buildings

This is a peer-reviewed paper.

Authors:
Hans-Paul Siderius, Netherlands Enterprise Agency, The Netherlands
Fiona Brocklehurst, Ballerat Consulting, United Kingdom
Adriana Diaz, Ecodesign Company, Austria

Abstract

In light of the increasing demand for data centre services, and their associated energy consumption estimated for Europe at 54 TWh/year in 2030, policy makers and industry are exploring ways to reduce data centres’ energy consumption and/or greenhouse gas emissions. This paper first presents a set of specific metrics to define and assess energy efficiency of data centres, which could be used to design energy efficiency policies.

The approach for developing these new metrics and improving efficiency moves away from directly quantifying the useful computing or service provided by the data centre, to look instead at avoiding energy losses in data centres, i.e., focusing on quantifying the energy spent not delivering any useful performance. Although power usage effectiveness (PUE) is still the dominant metric used in the sector, it does not capture the magnitude of losses in the IT equipment of a data centre. The authors introduce the server energy loss coefficient (SELC) to assess the losses of IT equipment, focusing especially on the servers.

This paper shows how the SELC can be calculated from standardised server operational parameters. In addition to this new metric, the paper presents a short overview and analysis of policies for energy efficiency of data centres, which range from MEPS (Ecodesign) and labelling to procurement criteria, incentives, and reporting requirements.

Finally, the paper provides an estimate of the associated energy savings from such measures, which in Europe could be 15 TWh/year in 2030. Half of these savings would result from decreasing PUE to 1,20; and the other half would result from lowering losses by increasing the utilization of servers, an accelerated shift to cloud computing, and switching off equipment at periods of low utilization.

Downloads

Download this paper as pdf: 8-001-24_Siderius.pdf

Download this presentation as pdf: 8-001-24_Siderius_pres.pdf