Technical scribblings RE Harmonizing Global Metrics for Data Center Energy Efficiency

Posted by John Stanley

A couple of weeks ago, several industry and government groups from the US, Europe, and Japan announced that they had come to some basic agreements on “the guiding principles of data center energy efficiency metrics.” There are some very encouraging technical aspects to what was decided, which I’ll describe in this post.

(Two weeks is an eternity in blog time, so I won’t bother re-hashing the press releases yet again. If you want the basics, please read about them via The Green Grid,  IT World, and Environmental Leader. Even better, read the different documents from early February stating what was agreed to, which can be found via EPA ENERGY STAR, another Green Grid post, or Rich Miller’s recent post.)

Here are a few encouraging technical points:

1) The first encouraging point is that the agreed-upon guidelines  for PUE define “total energy” in the numerator to include all forms of energy, not just electricity. For example, if you run a natural gas-fired absorption chiller in your data center, you might use much less electricity than your neighbor with an electric compressor chiller, but you should still count that gas energy as overhead. You’re paying for it, and it’s sure not going to your IT equipment or directly serving your IT customers.

2) The second encouraging point is similar: For data centers that are part of larger buildings, total energy in the PUE numerator should count all data center cooling, lighting, and other support infrastructure. Some data centers in larger buildings forget that their cooling systems consume lots of chilled water from the central building chiller, driving up chiller electricity use. If you forget this chiller energy, and just count electricity on the circuits directly powering the server room, then your PUE will look better than it really is.

Being diligent about counting all relevant energy is something Uptime Institute has long advocated, so I’m glad to see that #1 and #2 made it into the global guidelines.

3) The third encouraging piece was that total energy in PUE numerator should be calculated using source energy. When we talk about electricity, we often talk about source energy as opposed to site energy. Site energy is the amount of energy consumed at the site where it’s used, in the form of electricity. Source energy is the amount of fuel energy needed back at the power plant to get that electricity to the site. Source energy is much bigger. Of all the fuel energy put into a power plant, only about 33% of it is turned into electricity–the other 2/3 turns into waste heat. Then, ~6-7% of the generated electricity is lost as it makes its way down the power lines to its destination. So, for every 10 kBTU (~3 kWh) of electric energy used by a data center, you need to burn about 10 / (1-.07) / .33 = 32 kBTU worth of fuel.

Source energy is the correct way to compare electric and non-electric energy consumption. Image that you just put 10 kBTU of natural gas energy into your gas-fired chiller, while your neighbor just used 10 kBTU of electricity in his/her electric compressor chiller. The site energy is the same, but your neighbor has just consumed about 3x as much natural gas if you trace the electricity back to the natural gas-fired power plant.

Source energy is what we care about from an  environmental perspective, since it’s BTUs of coal and gas that cause pollution. Electricity is just the middle-man. Now we’re getting somewhere.

-

So, good progress by the industry, and I’m glad we’re getting buy-in for universally accepted practices across the globe.  I only want to have to write blog posts about one set of standards.