What, exactly, do you and your procurement department use to calculate the cost of running a server? Chances are you've gone beyond the hardware pricetag to figure in all the service, software and maintenance costs. But does your math include the facilities capital and operational expenditures that surround that box?
Ken Brill, founder and executive director of the Uptime Institute, which held its Lean, Clean & Green IT Symposium in New York this week, said many organizations still overlook the facilities costs associated with their data centers. This, of course, is a big focus of Uptime's ongoing research and was a theme that Brill drove home during his conference set-up keynote.
According to Uptime's latest data tracking this sort of thing, Brill figures that the capital expenditures associated with a $1,500 to $2,500 spend on a server are between $8,000 to $16,000 in the true "infrastructure" needed to support it. (We're talking cooling systems, physical rack infrastructure, power supplies, power distribution systems and so on.) The energy consumed by a single server can easily cost $400 to $500 per year, he says.
Other mind-blowing statistics that Brill uses to sensational effect:
- If the industry continues on its current path of consumption without substantially changing its habits, by 2015, it will consume 7 times the energy it does now. That's because we're still adding about 1 million servers to our collective IT infrastructure every year.
- The bottom third of data centers in terms of efficiency will consume 18 times the energy in 2015 that they did in 2000.
- The consumption of a single large data center is equivalent to the energy used by 20,000 people. (I don't know where he got this one, readers, but it attracted a bunch of attention when I tweeted this information earlier this week.)
- In short, data centers are THE most energy-intensive asset for any business that is not in the manufacturing sector.
Uptime recommends some specific actions this year for data center managers that can help stall these trends. Here are Brill's three top agenda items:
- Kill "comatose" servers that aren't serving a useful purpose and put in place a formal, ongoing decomissioning strategy.
- Virtualize 40 percent of your applications as a start (look for a 5 to 1 collapse in your virtual to actual server ratio, at least).
- Buy power supplies and fans that are 20 percent more efficient than existing systems in order to cut the facilities energy used to support your servers. Urge your procurement team to look at the "watts saved per dollar" if they balk at the upfront investment.
I've got more notes to sort through about the conference so there will be at least a couple more posts. Meanwhile, you can read my coverage of what Google had to save about improving data center efficiency.