One of the racks in our machine room

While perhaps not as impressive as the grocery store-size aisles of a supercomputer like the Earth Simulator (which was recently rated as the world’s most computationally-effficient supercomputer), our servers do suck their share of power. The two most significant energy inputs to this system are the cooling unit and the UPS (uninterruptible power supply).

Imagine your hair blowing in an arctic wind

Just a few hundred pounds of batteries

We have a system that monitors the power used by both the UPS and the cooling unit, and we’ve been performing measurements since February. Recently, we added a routine that calculates daily consumption for the current month. The plan is to put this information on the IMA Dashboard, and we’re using this opportunity to try out Google Charts.

Current IT power consumption for July

One of the key survival tools for anyone on the quest for sustainability is measurement. From the data we have here, it appears that we have a baseline of about 1300kWh per day. Knowing this, when we swap out hardware in the future we will be able to determine how much of an improvement we actually make, as opposed to relying on theoretical figures.

Of course, we’d like to be confident that our measurements are correct. In order to do this, we’re using Kill-A-Watt devices to measure power input on individual components in the rack. If the measurements are correct, the total power calculated from sampling individual units should be approximately equal to the power measured by the monitoring system. If not, we’ll have some investigating to do.

Tapping into a couple servers

Tapping into the main line

Hopefully, our measurements will validate and we’ll soon have live data up on the Dashboard. Until then, check out some other fancy energy usage dashboards.

This device monitors voltage, power, and cost

In today’s post I would like to share some investigations that I have done with a couple of cool energy efficiency tools. Perhaps you’ve heard of vampire energy. I’ve suspected that some of my home appliances are vampires, and a few weeks ago I decided to start doing some investigation of my own by purchasing a consumer-grade energy meter to measure the amount of energy that various electronics around my house consume. I haven’t done much analysis yet, but I have found that during normal use over a month, my entertainment center uses 75.04kWh and costs me $10.20 in electricity. This amounts to $122.40 over the course of a year. When everything is in standby, these undead electronics suck down 36kWh in a month (I need to run this test with the DVR on another outlet). Leaving the system in standby for a year would still cost me about $60.

The software displays your savings to-date

Last week, there was a new device laying on one of our office tables that I couldn’t resist asking about. It’s called an ecobutton™, and of course I couldn’t wait to take it home for some analysis over the weekend (yes, I’m a real party animal). The Web site claims that the the button will put your computer into a deep sleep, achieving power savings beyond what your normal standby mode can offer. I don’t have a desktop machine at home, but I did test it on my laptop from the Museum. The software is simple to set up. After entering a few numbers that it uses to calculate your savings, it’s ready to go. It’s also simple to use. When you walk off to a meeting, simply pressing the button puts your machine into power saving mode. The instructions say that the computer will awaken when the button is pressed again, but on my laptop I had to press the power button.

For my experiment, I made sure to disable all of the power management settings and only enable the mode for each measurement. Ideally I would run each measurement for an equal period of time, but this was a weekend project among other errands that had to get done (in as few trips as possible, of course). My hypotheses going in were that screensaver mode wouldn’t be much better than normal mode compared to “no monitor” mode, and that normal and deep standby would both be very close to zero consumption on a laptop.

So, let’s take a look at the results (Average Wattage = Consumption * 1000 / (Time / 60)):

Indeed, running in screensaver mode saved 5W on average, while no-monitor mode saved 16W. I ran the system in normal standby overnight, because when I checked it after about two hours it read 0kWh with an instantaneous wattage of 0W that seemed to occasionally spike to 2W and I wanted to see if anything would accumulate over a longer time period. Not surprisingly, deep standby also ran at 0W.

The lesson that we learn here is that deep standby is just as good as normal standby for our laptops, so the main benefit is that standby mode is easier to explicitly trigger with the button (rather than having your machine wait until Windows power management kicks in, or selecting it from the menus). As a side note, you can also use the pulsating glow of the button to practice stress-relieving breathing exercises (maybe there’s a Mac vs. PC commercial idea in there). I want to run this analysis on a desktop machine as well, to see if the button can save more energy in standby as the company claims.

Now, actual savings ultimately depends on the power management settings configured on our computers and requires a few assumptions and more calculations to estimate. It seems to me that the ecobutton™ could be a fun way to help us save a little extra energy (hopefully enough to offset the cost of manufacturing and shipping the buttons), and keep conservation in mind. As one of the Eco crew team members at the IMA, I’m on the lookout for technologies like these that can help us in being good stewards of the environment (hopefully while saving a little money and having a little fun as well). Stay tuned for more discoveries!