Posted by John Stanley
Let’s start with the punch line: big data centers get all the attention, but only focusing on making the big ones efficient will fail to curb national data center energy consumption.
Now the details.
If you do a web search for “new data center” (with quotes), the stories you read are about the next monster facility that’s coming online. You get articles about a 90,000 sq ft data center opened by Liquid Web, the 147,000 sq ft facility for Facebook, or the 500,000 sq ft facility by Apple. If you search for “data center environmental impacts” (no quotes), one of the first hits is an article entitled “Big data centers = big environmental footprints,” which also mentions big data centers.
But according to the often-cited 2007 Environmental Protection Agency (EPA) report on data centers, only about 40% of the electricity that data centers consume nationally is used by “Enterprise-class” facilities, defined as those being over 5,000 sq ft. So, if you only count the truly huge facilities (perhaps 100,000+ sq ft), then these would likely count for significantly less than 40% of national data center energy use. (For reference, 100,000 sq ft is about the size of a Wal-Mart.) Instead, most data center electricity use is in the countless smaller facilities, from the medium-sized sites just under 5,000 sq ft all the way down to the server closets of 200 sq ft or less.
This is important.
This is important because the majority of the hype, the pending regulations, the efficiency initiatives, and the efficiency consultants are generally targeted at the largest facilities. But even if you could reduce the electricity use of all these large facilities to zero, you’d still miss 60% or more of the consumption.
We as an industry need to find a way to make our small and medium-sized facilities as efficient as we’re trying to make our large ones.
Improving the non-huge facilities won’t be easy, though, since it’s typically the large sites that have the big electricity bills to motivate efficiency and the on-staff expertise to execute it. This could be an argument for organizations to start consolidating their server closets into bigger sites, or moving things to hosting companies and The Cloud.
2 Comments
February 15, 2010 at 12:48 pm
Hi,
Good luck with the new blog.
You raise an interesting point here. Do you have any numbers on how many “huge” vs “enterprise-class” vs small-to-medium data centres there are out there? Globally? By country?
To make facilities more efficient, you need to have decent metering and monitoring in place so you can understand where you are starting from and continue to measure and trend how you are changing over time – real-time. Data centres of all sizes seem to be relatively poorly instrumented in this regard. Do you think the bigger ones are more likely to invest to address this than the smaller ones are?
February 15, 2010 at 4:32 pm
I’m not aware of very much good, publicly available data that directly estimates the number of data centers out there, especially the number of server closets and smaller facilities. The 2007 EPA Report to Congress has a breakdown of US data center kWh use by size class. One way to back out a rough number of sites might be to use these kWh totals, convert to kW, assume a W/sq ft power density, get a total sq ft estimate, and then use the EPA’s definitions of the size classes (a server closet is <200 sq ft, etc.) to get a site count. Very rough, but you could do it.
To your second point, yes, I definitely think larger facilities are more likely to invest in reducing their energy consumption. If some kind of efficiency initiative can save you X% of your power bill, then bigger sites with bigger power bills are chasing larger dollar amounts. This makes it worthwhile for them to put money into tools, equipment, services, and staff time to pursue the savings.