The green computing movement has gotten some traction; I’m glad. That was one of my earliest campaigns, a cause I fought before I had a voice. And until a few years ago, I was frustrated that others were overlooking the obvious: Corporations don’t need to become champions of the environment to push for cooler, quieter, more efficient electrical equipment. I’ve said that they just have to look at their monthly electric bill. That was naïve; how can a business tell what portion of its electric bill goes to computing and storage? There must be a way to figure out whether there really is a cost savings.
There is so little consensus on the proper method for calculating the value of green computing that it’s become an opportunity for vendors. AMD makes an effort to give us a ballpark tally of the dollars we’re saving, or wasting, by running, or not running, Opteron servers instead of Intel’s. It’s fuzzy science, to be sure, but there’s enough fuzz to go around. Intel prides itself in power conservation at the component level, but it refers to “typical” power utilization rather than the worst-case numbers it gives system engineers to guide their cooling system design. At least AMD is still publishing that worst-case number.
If you don’t feel like playing Dr. Electron, you can leave some of it to me because I’ve now got an arsenal for measuring the effectiveness of green computing. I’ve come up with simple, objective formulas for calculating green-computing savings.
If you do like working in the details, there are three variables to measure: power consumption, heat, and noise. The closer you can get to an accurate measure of all three, the better your odds of charting your success at bringing down your total IT costs. But there are two shortcuts you can employ to get you on the right track until I start making with the science.
One shortcut is to measure any one of the three variables because they’re all related. Power consumption has the most appeal because amps can be converted to watts, then watts over time to kilowatt-hours, and from there to real operating cost. But the cost of cooling complicates matters. The single most power-hungry machine in my facility is the air conditioner. It spends some of its energy cooling the equipment and some of it making the place livable. I could cancel out the human element by letting the room warm to a level where the equipment won’t melt down, but I would. If I set up that scenario periodically, I could safely take cooling out of the equation.
If I were to turn the thermostat up first, I might not have to measure power draw at all to plot a savings trend. The less efficient a device is, the more heat it generates. Fans will have to spin faster to keep the system going, and faster fans always mean more noise. Noise is easy to measure with a single handheld device, and you get the added benefit of measuring the habitability of the environment.
Turning the AC off entirely yields another simple measure: How fast does it take to heat up the room? You can get a rough measure of the heat output of a single device by checking the temperature at its exhaust.
None of these processes yields hard and certain dollar savings, but they are adequate for tracking the progress you’re making.
Or you could ignore this whole column and go with the lazy man’s way of tracking power savings: Count the total number of power supplies.