The U.S. Environmental Protection Agency (EPA) recently extended its Energy Star program to include rack servers....
By submitting your personal information, you agree that TechTarget and its partners may contact you regarding relevant content, products and special offers.
Can you talk about the challenges the EPA will face in quantifying energy efficiency in the data center?
[The EPA] has been working on this for about 5 or 6 years. The main genesis for doing this came out of a charrette, a whole bunch of experts getting together to work on a problem, done by the Rocky Mountain Institute. They brought people together from the data center industry to think about this problem of energy efficiency pretty much in advance of the current wave of interest. They wrote a whole report where one of the results was to push the government to consider extending Energy Star to servers and other data center related products. I think it's a great idea.
I think it can go further. Half of the power that comes in from the utility companies never even makes it to the server. It's wasted in power systems and cooling systems, the kinds of products that APC makes. We have a huge responsibility to fix that. How would that be measured?
Energy Star has other programs besides these product level certifications where they recognize [efficient organizations]. They try to recognize solutions that are exceptional in their field for performance. It's not benchmarked against others. I think that's a good way to start when you have a difficult problem; just select solutions that you know have done a better job. And as you get better and understand it, you can start laying the standards out in detail.
I think they should start doing that for data center infrastructure; for complete data center designs. If each part is efficient, it doesn't necessarily make the data center efficient. I can still make a bad data center out of efficient parts. The part-level qualifications are nice, but if I take very efficient servers and throw 18 million of them in a data center when I only needed six, I'm going to waste a lot of power. It's not just the parts. You have to look at the complete system. So, something like what LEED (Leadership in Energy and Environmental Design) does?
I totally agree with that, because one of things that people are going to find is what am I going to do with all the junk from the data center after I retire it in 12 years? Can it be recycled? LEED thinks about those kinds of things, recycle-ability, disposability, etc. A lot of people aren't thinking about that today. They will be thinking about it for sure. You can throw a lot of things away in a dumpster and get away with it. As we all know, that is going to be less likely in the future. All of the sudden it's going to be a big deal for people to think about the environmental friendliness of the data center overall, not just energy efficiency.
For example, we're in a massive effort to remove lead from all of our products. Solder is made from lead, we make all kinds of electrical things and they all have solder in them. It's a huge project to try to redesign everything to eliminate lead from the solder in all these products. That's the kind of thing we've been doing at APC now for three years. How can a data center manager improve efficiency on a complete system scale?
If you look at the opportunity to save energy, most of the opportunity is in bad system design, not in the parts. The key to energy efficiency in the data center is right-sizing. The number one driver of waste in the data center, whether it's servers, power systems, cooling systems, is over-sizing whatever it is you're looking at. And it's rampant, the amount of over-sizing that occurs within data centers at all levels.
The reason people oversize is because it's so difficult to scale. So people build a giant thing first and hope that it's going to fill up. But while it's filling up, it's inefficient. A typical person who builds a data center expects it to survive about 12 years -- and that number is increasing. If I expect it to last 12 years, I have to build it for a 12- year forward looking goal. That's the logic being used, but they don't have a clue what they're going to need in 12 years. So what do they do? They know that they've been growing, so they over-design these things, make them huge and start out with a small load of IT equipment and gradually grow it over time. When you do that, it's very inefficient.