Beyond PUE: New Metrics for Data Center Energy Efficiency

Written by Joe Kozlowicz on Wednesday, May 7th 2014 — Categories: Cloud Hosting, Colocation , Data Center Design, Green Data Center

By now you’re likely familiar with PUE, or Power Usage Effectiveness, an industry standard measurement for the energy efficiency of a data center. Despite some claims that PUE is easily manipulated or not enough to judge full environmental impact, many data centers (including Green House Data) are using PUE to measure efficiency.

The Green Grid, a consortium of technology companies who aim to improve data center efficiency, has collaborated with industry groups around the world to develop several new metrics to measure carbon emissions and energy use in the data center, including GEC, ERF, CUE, and DCeP. What are these new measurements, and how does Green House Data stack up?

First, a quick overview of how the standards were developed. The Green Grid is a nonprofit organization founded in 2006, consisting of technology industry leaders including AMD, Dell, EMC, Microsoft, and IBM. In order to address inconsistencies in the application of PUE and other metrics, they gathered together with the US Department of Energy; the European Commission Joint Research Centre Data Centres Code of Conduct; Japan’s Ministry of Economy, Trade and Industry; and Japan’s Green IT promotion council.

Their latest release is an agreement regarding data center productivity and includes the methodology for several metrics, originally developed in 2012. The main focus of the agreement is for data center operators to begin measuring DCeP, or Data Center Energy Productivity.

The goal is to go beyond pure efficiency to reflect how much actual productive work happens in a data center relative to the energy use, green energy sources, and carbon emissions. The more metrics in place to measure these factors, the more ways data center managers can tweak their operations to maximize energy use. Here are the main metrics detailed in the agreement.

Power Usage Effectiveness (PUE)
The classic. PUE measures the total energy use of the data center compared to the energy used by IT equipment. In other words, for every watt used to power IT equipment, how much is used for cooling, lighting, and additional infrastructure? Green House Data averages about 1.25 PUE across all data centers, and our new facility will have a low PUE of 1.14.

Green Energy Coefficient (GEC)
GEC is a measurement of how much of a facility energy use is sourced from green providers like wind, solar, or geothermal plants. This includes “any form of renewable energy for which the data center owns the rights to a green energy certificate or renewable energy certificate”. The amount of green energy purchased or consumed divided by the total energy consumption equals the GEC. Green House Data has a GEC of 1.0 as 100% of our energy use is covered by RECs.

Energy Reuse Factor (ERF)
The ERF of a data center reflects how much energy is exported for reuse outside of data center operations. For example, heat given off in the hot aisle of a data center that is then piped to heat other office areas would be reuse energy. Any energy that is reallocated outside of the data center floor and the infrastructure support (cooling, etc) is divided by the total energy consumption of the facility to find the ERF.

Carbon Usage Effectiveness (CUE)
The most involved of all the metrics, CUE is also potentially the most interesting, as it quantifies the actual emissions involved compared to the work performed. This measurement currently excludes the emissions involved in producing the actual equipment, shipping it to the data center, construction and materials of the data center building, and so forth. It focuses merely on operational energy use and emissions.

CUE is calculated as Carbon Dioxide Emission Factor (CEF) multiplied by PUE. The CEF is found from Energy Star data in the United States and is the kg of CO2 produced for each kilowatt-hour of electricity used. This varies a large amount depending on which grid system is used by the data center, which in turn is entirely based on location.

To calculate the CEF, data center operators should find their energy grid subregion (in our case, the Rocky Mountain Power Area or RMPA) on the Energy Star table describing Indirect Greenhouse Gas Emissions Factors. Then grab the kg Co2e/MBtu measurement and divide that by 1 btu, which equals 0.293Wh. For the RMPA, that would be:

254.6387 kg Co2e/MBtu / 0.293Wh = 869.07 KgCo2e/MWh = 0.86907 KgCO2e/kWh

In other words, Green House Data emits 0.87 kilograms of CO2 for every kilowatt-hour of electricity used based on the efficiency of the Rocky Mountain Power Area electric grid. This measurement is independent of the GEC as all energy credits are eventually fed into the standard power grid. This measurement isn't the most accurate or up-to-date, as the latest EPA chart is from 2010. Admittedly, grid improvements are few and far between.

The CEF calculated above is multiplied by PUE to get the total CUE. Green House Data has a cumulative average CUE of 1.0863. The new facility will have a CUE of 0.99.

Together these new metrics offer a more detailed picture of the efficiency of a data center facility, allowing operators to more closely examine energy use and how to release less carbon emissions. Although most data center providers aren’t widely reporting CUE, ERF, or GEC just yet, the Green Grid and the associated organizations helping to create these metrics continue to refine measurements of data center energy use and impact.

In the future, measurements like Data Center Energy Productivity (DCeP) could allow more detailed reporting on the amount of actual work performed by the servers compared to their energy use. The new agreement does include a formula to measure DCeP, but it is contingent on “tasks initiated” during the assessment window, leaving some ambiguity at hand (of course, plenty of people argue that PUE is also highly ambiguous). Measuring the number of tasks completed can be difficult for service providers as well, as insight into customer environments can vary.

The following section includes some suggestions for Data Center Productivity Proxies, or simpler, more comparable methods of measuring productivity including:

These metrics are all great ways to maximize the use of energy and existing equipment, but can vary dramatically depending on the goal of each data center (research vs. development, for example).

What do you think about the metrics recommended by the Green Grid to measure data center efficiency and carbon emissions? Let us know @greenhousedata on Twitter!

Posted By: Joe Kozlowicz