The Grill: Jonathan Koomey

Computerworld - Professor, scientist and energy efficiency expert Jonathan Koomey, who recently finished a term at Lawrence Berkeley National Laboratory, got the full attention of the IT community in 2007 when a research paper he wrote revealed that power consumption by data centers worldwide had doubled in just five years.

Here, he discusses how the industry has responded to those runaway costs, why cloud computing is better for the environment, and why you should think twice about where you locate your next data center.

Has the growth curve in data center power consumption moderated since your paper came out three years ago? We don't really know. We have some inkling that things have slowed down in part because of efforts to improve the efficiency of IT equipment and because of the economic slowdown.

By 2007, some data centers were already running out of power and having cooling problems. Did your study really come as a surprise? The study provided a quantitative estimate that encapsulated the problems people were seeing in their own facilities. It showed that this was an issue in the aggregate. People didn't know just how big it was.

There's been criticism regarding the amount of electricity that large cloud computing data centers are consuming -- and where that power comes from. Is that fair? No, I don't think it's fair at all. Cloud computing data centers are much more energy efficient than standard in-house data centers, and moving bits instead of atoms is a very good thing. The electricity used for the cloud computing data center allows you to do a lot of things in the economy that are much better from an environmental perspective. People don't have to go somewhere -- with videoconferencing, you're moving bits instead of atoms.

Those things turn out to be much more important to the environment than just the electricity used by the data center. Cloud data centers are allowing us to accomplish our goals in a way that is less damaging to the environment.

A public cloud has energy efficiencies because it's based on a shared-service model and benefits from economies of scale. Do internal clouds have the same benefits? Anytime you can separate the virtual computations from the physical infrastructure, it will reduce costs and energy use. If you did that internally, you wouldn't have quite the diversity of users that Google has, but you would have those other effects, the economies of scale and the fact that you have the separation of the virtual and physical servers. For example, that means you wouldn't have to have dual power supplies in your servers but could just have the software route around a server that died. The cloud is redefining reliability, which also saves resources.

You've said that a carbon tax on data centers is inevitable and that it could negatively affect some data centers. Why? We need to reduce our emission of greenhouse gases by 80% to 90% by the middle of this century. In order to do that, governments need to impose a price on carbon, because if you have a price, then the markets will filter that through and change behavior in an efficient way. We're going to have to have a price on carbon. That means there will be an effect on the direct electricity use by data centers. That creates a business risk for companies that site data centers in places with high coal use in the electricity sector, for example.

But electricity from a coal-fired plant can be much cheaper than other types of power sources. Would such a tax wipe out any savings from locating in areas that use coal-fired power plants, or would it still be cheaper to operate there versus, say, in Manhattan? It depends. The average industrial price for electricity is about 7 cents per kilowatt-hour in the U.S., so a 2-cent tax on top of 7 cents is roughly a 30% increase. It's not going to wipe out the difference between operating a data center in New York or Wyoming. My point is only that this is a business risk that you need to factor in when deciding where to put a data center.

You talk about misplaced incentives that work against efficiency in data centers. What are the biggest ones? This one of silos, of having the IT and facilities budgets separate. The IT folks don't have an incentive to spend even an additional dollar on a more energy-efficient server because the savings don't accrue to them.

That's the same issue people were talking about three years ago. Has nothing changed? I'm sure some companies have changed, but most haven't. If you changed that, a lot of other problems would go away.

How will data centers evolve over the next three years? There's going to be an economic trend pushing people more toward cloud computing. Internal data centers are going to see more and more pressure on them. With very few exceptions, they're going to be forced to change because their costs are going to be higher -- a lot higher -- than the cloud model.

Beyond what we see with cloud computing, how much more efficiency can we expect to squeeze out of data centers? We have a vast potential for improving the efficiency of this IT equipment, and by "vast" I mean by many orders of magnitude. We're very far from the theoretical limit. It's just a matter of getting smarter about how we deliver these services.

Read more about Data Center in Computerworld's Data Center Topic Center.