A Sampling of Bright Ideas : Simple Tweaks? |
|
Go Back to Home |
DETAILS
![]() |
If you owned an early home computer or worked on an early
terminal at the office, you know how much they’ve changed:
They’ve gotten smaller, and they’ve gotten richer—loaded
with software capable of doing many more things, laden
with what we used to call “bells and whistles.” The problem
with all these cool capabilities is that they’re piled
in layers on top of software that was laid down long
ago, and the result is very inefficient use of energy.
Erez Zadok, associate professor in the Department of Computer Science at Stony Brook, says that 1 percent to 2 percent of U.S. energy consumption is devoted to data centers, those air-conditioned rooms where big companies store their computers, blinking and humming and drinking up electricity day and night. If you add the power consumption of desktop and laptop computers, that accounts for another 1 percent to 2 percent of U.S. power. “That’s a lot, and mostly while computers are just idling,” he says.
Because all of the software developed over the years has been built by succeeding generations of engineers, often on top of glitches and flaws, “no one has looked at the big picture, how it all works together,” Zadok says. That’s what he’s doing now—looking at software to see where the inefficiencies are, and he has found that with “simple tweaks” to servers, he can almost always get improvements, “up to three times, four times, even nine times better.”
Obviously, that would be good for big companies spending big bucks on their data centers. With a ninefold increase in efficiency, they would save on power to run the computers and on air-conditioning as well, because the machines would run cooler. But sharply boosting software efficiency would also be good for students, businesspeople, and other ordinary folks. Imagine having your laptop battery last twice as long because your software is draining less energy from it.
Zadok’s three-year research project in his File System and
Storage Laboratory is funded by the National Science Foundation
to the tune of $720,463. Year One was devoted to measuring the
power drain of various usages. This past summer [summer 2010]
he began Year Two, devoted to writing more optimal software.
What he hopes to end up with in Year Three is a sort of toolkit that
will allow people to upgrade the efficiency of their data centers.
Some companies have moved their machine rooms to cool
climates, so they don’t have to spend so much money to air condition
them, Zadok said. He noted that Google, which runs
24/7, has moved one of its data centers to Northern Ireland.
Another firm puts its computer room in the high plains of
Colorado; to cool the computers, the windows can be opened.
Relocation may be one option, but wouldn’t it be smart to eliminate the inefficiencies in the first place? That’s Zadok’s goal.
More information on Dr. Zadok's work can be found here.
Article created on: November 2010

