Power Usage Effectiveness or PUE, is the most commonly used metric in Data Centers today. Data Center industry organizations tout it, major companies brag about their scores, and energy efficiency certifications use it in their ratings.
Despite how thoroughly it has been embraced, though, it’s not without flaws. In my opinion, they mostly stem from the fact that PUE is more complicated than it looks. Although calculating the metric is a relatively straightforward exercise, the resulting score doesn’t tell you anything about the factors that influenced it. Hence why it’s a bad idea to compare PUE among Data Centers.
(I posted a few months ago that I had recently finished writing a book on innovative server environments, titled The Art of the Data Center. Although the book includes detailed profiles of several rooms, I intentionally avoided listing any PUE values. I didn’t want readers to compare scores and – because a Data Center’s PUE constantly changes – I didn’t want to be publishing outdated information.)
I’m reminded of my high school math teachers who always wanted me to show my work. How I got to the answer was just as important to them as the answer itself.
Despite its flaws, PUE has taken hold in the Data Center industry because it does offer some insight. Let’s take a look at how the PUE is calculated, some of the factors that influence it and how PUE should – and shouldn’t – be used.
PUE is transparent….
Once the physical implementations are optimised, the next frontier will be energy efficient algorithms where the work done to produce a given useful or measurable output is optimised.
Great point, Andrew. Optimizing a Data Center’s physical infrastructure and energy consumption is important, but so too is understanding and maximizing the productive work that’s accomplished by that Data Center. If you don’t, you’re just joyriding in that economy car.