Researchers at Carnegie Mellon University have created a structure that would allow computers to process large amounts of information using less electrical energy. The new research could eventually reduce energy costs for colleges, universities, and other institutions that must process and store large amounts of information.
The research relies on the idea that fast processors burn up more energy than they should, said David Andersen, a leader of the research team and assistant professor of computer science at Carnegie Mellon. Instead of using one fast processor to get a task done, setting up a system of slower processors can achieve the same task, using less energy. The structure the researchers have found is called “Fast Array of Wimpy Nodes.”
Fast processors are like fancy sports cars that consume a lot of gasoline, but using a group of slower processors is more like a small energy-efficient car, Mr. Andersen said.
Programmers at colleges and universities must invest some time figuring out how to set up that system of processors initially.
“Let’s say that you have a program that gives you a list of what students are taking what classes,” Mr. Andersen said. “What you’d have to do is figure out a way to divide the list up,” so that different processors are responsible for sections of the information, he said.
For institutions that keep large amounts of data, such as information on students or employees, the programmers’ time could pay off in the long run, Mr. Andersen said. The structure could cut energy costs in half, he said, and reduce hardware costs by 25 percent.
Michael Kaminsky, a senior research scientist at Intel Labs Pittsburgh, was the research team’s other leader.