While we have been discussing the merits and flaws of PUE, another computational model is making great strides in the data center sphere: Computational Fluid Dynamic, or CFD. However, the whole concept is totally different.

Because where PUE focuses on bringing a straightforward, easy to use calculation on data center efficiency, the core value of CFD is accurate tracking of the air flow. In cooling, air flow is (of course) vital and probably one of the most difficult things to track.

You can compare it with the weather forecast: while there are very knowledgable institutions at work, they are rarely spot on in their predictions because of the complex nature of wind, pressure, humidity etc. That makes CFD so tricky: air flows are dependent on many things, which is why software to track CFD is so damn expensive. It starts at 7500 dollar for a year license. That is no small sum for a start up.

So should you go for it? That is the question Data Center Journal poses. The basic answer the Journal brings is: maybe, if you are running a large facility. Smaller data centers should stay well clear, since 'cheaper' solutions actually sacrifice the one thing that would make CFD so interesting: accuracy.

I agree wholeheartedly. Why would you even invest in something that is meant to give you accuracy, when going for the budget option actually denies you your basic need? Is CFD worth it anyway, as thermic camera's can give us a good insight in where the hotspots lie?

Another thing that puzzles me is this: isn't CFD originally designed to track fluid flows? What is it that makes data center hot air/cold air fluids all of the sudden?