The Truth About PUE | Green House Data Blog

Happy Earth Day! As we take today to reflect on environmental impact, let’s take a look at some solid data to see how much energy and carbon dioxide emissions are really saved by data center energy efficiency and renewable energy use. The recent memo from the Greed Grid reintroducing their metrics for data center efficiency provides a great jumping off point to estimate the environmental impact of an average data center.

With recent headlines like Apple’s shift towards renewables and Google’s funding of wind farms, not to mention our own efforts at Green House Data to improve efficiency and overall Power Usage Effectiveness (PUE), many in the data center industry are curious about the actual data at hand. If a company improves 2% of its overall carbon footprint through efficient or renewably powered data centers, what does that actually mean? Is it a large impact or just a PR opportunity?

Summary of Data Center Emission Data and Method

We took a theoretical 10 MW facility and assumed it was operating at capacity for simplicity of math and comparison. We measured this facility’s emissions at 1.8 and 1.2 PUE to see how improving operations would affect emissions.

Using the Carbon Usage Effectiveness (CUE) metric, which is a combination of PUE and the Carbon Dioxide Emissions Factor (CEF) of a location, we discovered:

• the electricity grid region of a data center has an enormous impact on total CO2 emissions
• dropping from 1.8 to 1.2 PUE can save anywhere from 36,000,000 to 100,770,000 pounds of CO2 emissions annually
• the regions with the highest carbon dioxide emissions factor had 65% higher emissions than the lowest
• a 1.2 PUE facility in the least efficient grid region still emits 46% higher emissions than a 1.8 PUE facility in the most efficient region — that’s 42,485,427.84 kilograms more each year despite being 1/3 more efficient!

eGrid Subregions – Find Your Carbon Usage Effectiveness

The United States electricity grid is divided into various subregions called eGrid regions (Emissions and General Resource Integrated Database). This database tracks air emissions rates, net generation, resource mix and more. The EPA provides this information and also the geographical division of eGrids, which we subsequently placed on a Google Map:

If you select the region of your data center facility location, you can see the Greenhouse Gas Emissions Factor and CEF to help calculate your own CUE (Carbon Usage Effectiveness).

CEF is the Carbon Dioxide Emission Factor, measured in kilograms of CO2 emitted per kilowatt-hour (kgCO2eq/kWh), as described by the Greed Grid. The EPA reports the Greenhouse Gas Emissions Factor (GGEF) of each eGrid subregion. The formula for CEF is as follows:

CEF = ( Greenhouse Gas Emissions Factor / 1 BTU) / 1000

The Greenhouse Gas Emissions factor is measured in kg CO2 emitted for each MBtu, so dividing it by the value of 1 BTU in Watt Hours (which is 0.293 wH) describes the amount of CO2 emitted in kg per Megawatt-Hour. Dividing that amount by 1000 provides the CEF in kWh.

As an example, the Rocky Mountain eGrid is RMPA and has a GGEF of 254.6387. Any facility located in the RMPA grid region has the following CEF:

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

The Annual Emissions of a 10 MW Data Center

To find the total emissions in kg of a 10 megawatt data center facility, we first have to calculate the Carbon Usage Effectiveness (CUE) at both 1.8 and 1.2 PUE.

The CEF is multiplied by PUE to get CUE, which is multiplied by annual energy use in kWh to find the total emissions of a data center in kg of CO2.

CEF * PUE = CUE
CUE * Total Annual Energy Draw = Annual Emissions

For our 10 MW facility, if it were in the RMPA grid region, the calculation would go as follows:

PUE (10 MW Facility in RMPA Grid Region) Carbon Usage Effectiveness Annual Emissions (kg)

1.8 PUE
0.86907 kgCO2e/kWh * 1.8 PUE = 1.564 CUE
(1.564 CUE * 8,765.81 hours per year) * 10,000 kW = 137,126,485.77 kg annual emissions

1.2 PUE
0.86907 kgCO2e/kWh * 1.2 PUE = 1.043 CUE
(1.043 CUE * 8,765.81 hours) * 10,000 kW = 91,417,657.18 kg annual emissions

The chart below shows the annual emissions of a 10 MW facility in each subregion at 1.8 and 1.2 PUE.

Although PUE is a variable metric and has been accused of manipulation in the past, if we assume a legitimate measurement, lowering PUE from 1.8 to 1.2 can result in millions of pounds of CO2 saved from the atmosphere.

This chart shows the amount of CO2 saved in each grid region (calculated as [Emissions at 1.8 – emissions at 1.2] * 2.20462 lbs/kg). Once again, this is for a 10 MW facility.

The data points to two interesting conclusions, only one of which is really controllable by data center operators: (1) lowering PUE delivers dramatic reductions in carbon footprint and (2) the electrical grid region will significantly impact the emissions level of data centers.

For many companies including Green House Data and our favorite headline-grabbers like Google and Facebook, another weapon in the fight against emissions is renewable energy sources. The big companies are constructing their own renewable generation or investing in large scale privately owned wind farms and solar fields, removing themselves from the grid subregions entirely. Other companies purchase Renewable Energy Credits, meaning they are still impacted by the efficiency of their local, dirty grid; but at least are making an investment to reduce the ultimate carbon footprint of data center operations.

In either case, it will be interesting to see if data center operators large and small begin to measure their Carbon Usage Effectiveness and total emissions on a yearly basis. Of course, these calculations are for regular operation pulling off the standard grid only and do not take into account factors like diesel generators, office supplies, executive travel, etc. But the above charts, maps, and formulas can at least help get operators on their way to measuring carbon footprint of data centers.

Posted By: Joe Kozlowicz