If you've been following this blog at all, you've probably noticed I have a bit of an interest in datacenter energy efficiency issues. As part of my regular work week I spend time browsing web sites and searching for new and unique approaches to running datacenters in a more energy efficient manner. I even wrote an eBook on that very topic last year, and I try to keep my knowledge current as to trends in the community.
So it was in this frame of mind that I skimmed through the coverage of yesterday's Green:Net 2010 conference and found a presentation by Google Energy Guy Bill Weihl titled "Innovating to Reduce Energy Consumption of Computing." Now that's the kind of stuff I like to see, I thought to myself; a forward looking company talking about innovations used to reduce power consumption. This should be worth my time to sit through the video presentation.
Well, the most interesting thing about the presentation was that Google is looking to get into the power generation business, in sort of a sideways fashion, by requesting permission from the Federal Energy Regulation Commission to buy and sell power on the public market. This makes a lot of sense for a business model like Google's and I hope that they get the permission they seek.
But the "innovation" part of the presentation was a complete letdown, at least from the perspective of knowledgeable datacenter administration. The "innovation" consisted of some basic, commonsense steps for reducing datacenter energy consumption:
- Separate hot and cold
Google uses inexpensive plastic curtain hot aisle containment, with end caps, a technique that has been in general use for the last few years and that has many vocal advocates
- Run hotter
Running your datacenter hotter allows better PUE numbers; Microsoft talked about using 90+ degree temperatures to get their pods down to a PUE of almost 1.
- Turn off the A/C
Use alternate cooling and only turn on the chillers when you need them
Some of the coverage of the event mentions that these are Google techniques for smaller computing rooms and not proprietary datacenter cooling information. Or that it doesn't discuss techniques like Google's per-system UPS model, which requires proprietary hardware builds (and which had a little discussed, but very notable major design failure less than 3 months ago).
I'm sure that Mr. Weihl gave this presentation in good faith, but frankly, if you're a major player and are going to give a presentation with this title, I think the attendees actually deserve some information of real innovation in energy efficiency and not a rehash of commonsense techniques already in use.