Kristin Harty Barkley
FROSTBURG — Frank Barnet was having epic thoughts when he set about naming his new computer.
An associate professor of math at Frostburg State University, he built it over the summer using four Sony PlayStation 3s — each a “node” in what’s known as a “cluster computer.”
“I was going to choose some heroic name,” said Barnet, noting that the type of cluster he built is known as a “Beowulf cluster.” “Something like ‘Odyssey.’”
“And this one would be named Homer,” he said, indicating the first node, “and this would be named Odysseus, and so on.”
In the end, Barnet settled on slightly less dignified names.
“I decided well, yes, I’m going to name this Homer,” said Barnet. “Homer Simpson. And this one is Bart, and Marge and Lisa.”
About 20 students and other math enthusiasts came to campus on a foggy, rainy Wednesday evening to hear Barnet talk about his cluster computer.
Typically built using inexpensive personal computer hardware and free software, a cluster computer is a group of linked computers that performs parallel computing tasks more quickly and cheaply than comparable high-performance computers.
Barnet decided to try to build one after reading about two professors at the University of Massachusetts at Dartmouth who built a PlayStation cluster computer last year to do research on the “properties of gravity waves generated by the merger of two black holes.”
“They wrote a guide — not a very comprehensive guide — but at least a summary of what to do, a road map,” said Barnet, who needed a fast, inexpensive computer to better create computer animations.
“I said, well, that at least looks like something I could do,” said Barnet, who ended up spending about $1,700 on the computer — $400 for each of the PlayStation 3s, plus about $100 for a network switch.
“And if it fails, at least a PS3 computer, you can play games, it has a Blu Ray disc player, WiFi, Bluetooth, all of that.”
Demonstrations Wednesday evening proved Barnet’s experiment to be a success. It took the cluster computer 36 seconds to compute a complex math problem to the billionth degree, meaning it performed about 162 million operations per second. And that was using only four of its 28 processors.
“That’s not too great, actually,” Barnet said. “It’s good. It’s probably about double a typical desktop from four to five years ago. ... Hopefully I will be getting better. I’m pretty convinced it has a lot of potential, and if I add one or two more nodes it actually could do something very interesting from a graphical point of view.”
Barnet plans to use the cluster computer in the classroom with graphics and animations to demonstrate mathematical principles.
“You can compute something algebraically and get some kind of understanding, but if you can visualize it too, you can bet a better understanding of it,” Barnet said.
Fourteen-year-old Daniel Wojnar didn’t have trouble catching on to the concept of the cluster.
“He’s told me about it before, and I looked at it on the Web,” Wojnar said after Wednesday’s presentation. “It’s basically when you have a bunch of processors and running them all together and making this one do this chunk, this one do this chunk, et cetera.”
Former FSU math department instructor Roberta White said Barnet is known for his enthusiasm about technology.
“He does a lot of the interactive, computer graphics displays,” White said. “He’s always trying to bring the technology edge into the upper-level classes and motivate the students beyond just the mathematics involved.”
Contact Kristin Harty Barkley at firstname.lastname@example.org.