Greening the data center: the essential lie of PUE

The most common measure being touted today for the “green-ness” of a data center is power usage effectiveness (PUE). In this post, I’m going to talk about what PUE is, what I think it’s really measuring, and where I think it all goes terribly wrong.

PUE is a very simple metric, simply divide the total amount of power entering the data center by the amount of power that is actually operating IT equipment. A PUE of 1.0 means that all the power is operating IT equipment, a PUE of 2.0 means as much power is going to other uses as is being used to operate IT equipment. This is a deceptively simple idea.

What this metric is really measuring is the data center’s effectiveness at removing waste heat. What this doesn’t tell you is how effectively the data center is using the power for actual work. Idling machines, horribly inefficient old machines, it doesn’t matter to PUE just as long as you have a reasonably efficient way to remove the waste heat. This leaves the data center operator wondering what went wrong as they go bankrupt while the PUE looks great!

While using virtualization positively impacts the bottom line by getting more work out of existing equipment, it doesn’t gain PUE at all. Whether you leave machines idling or remove them, the overall ratio stays basically the same, even though the change can reduce the actual operating cost drastically. In fact, reducing the number of operating machines can raise PUE since many cooling systems are inefficient until they reach a certain level of load.

So, if reducing heat load and improving the efficiency of the IT systems doesn’t produce a gain in PUE, what does? The sole thing that produces a gain in PUE is an increase in cooling efficiency. More cooling for less power consumed is the secret.

Is there any place that PUE can be used properly? Only in helping to evaluate changes in a single facility (notice I said HELPING). Using PUE as a sole or even primary metric is terribly misleading and totally useless for comparing two different facilities, since it tells nothing about the differences that really impact the bottom line between those facilities.

When it comes right down to it, there’s only one sane metric for efficiency and that’s power in vs amount of work out. Nothing else makes sense.

Vern, SwiftWater Telecom

Data center, web hosting, Internet engineering


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s