Supercomputing [Wyoming Tribune-Eagle, Cheyenne]
(Wyoming Tribune-Eagle (Cheyenne, WY) Via Acquire Media NewsEdge) Nov. 13--CHEYENNE -- Less than a month after the doors opened on a supercomputer here, the University of Wyoming has cut the ribbon on its own lightning-quick machine.
Friday marked the official opening for "Mount Moran," a supercomputer designed and built specifically for UW researchers.
It is nowhere near as quick as the National Center for Atmospheric Research's "Yellowstone," which is located in the North Range Business Park west of the city.
But it is substantially faster than what UW had before, and it will allow a broader range of researchers to access it.
Since Yellowstone was built primarily for atmospheric and geophysical research applications, UW researchers hoping to use it must belong to a select set of research fields. And even then, they only get a certain percentage of its operational time.
While Mount Moran is smaller, it also will let researchers in other fields do work on it.
"One of the intents is so that people can have a friendly environment where they can run their codes," said Tim Kuhfuss.
He is UW's director of research support for information technology.
"We're also building the support infrastructure n the help desk, training manuals and consulting n that allow researchers to use it," Kuhfuss added.
"That's been one of my goals: I want to get the disciplines to start thinking about supercomputing. The best way to do that is to have great expertise."
Kuhfuss said some of the researchers running data through Mount Moran will use it as a stepping stone to gain time on Yellowstone.
One of NCAR's requirements to get on Yellowstone is to show that data models can "scale up" from small systems to larger ones without slowing them down.
Other researchers may use Mount Moran as a bridge to other supercomputers around the country that specialize in crunching data in different fields. As an example, Kuhfuss pointed to Alex Buerkle, an assistant professor of evolutionary genetics.
"(He) is doing gene sequencing," Kuhfuss said, noting that discipline isn't an approved use on Yellowstone. "But there are a lot of national facilities similar to Yellowstone's that are for that, and he's able to jump onto those."
Kuhfuss said other non-Yellowstone-approved subjects that can be studied on Mount Moran include pure math, chemistry, material sciences and the study of quasars.
Mount Moran is capable of performing about 25 trillion computations a second, which is about 60 times slower than Yellowstone's 1.5 quadrillion. But it still is substantially more powerful than the previous system UW had in place.
"Prior to this we had a little over 20 small clusters of computers, and what the researchers would do is run their jobs on those. But they were kind of compute-bound," Kuhfuss said.
By concentrating its power into one computer instead of multiple nodes, Mount Moran also will use much less electricity, providing substantial savings to UW, Kuhfuss said.
Like Yellowstone, Mount Moran was built by IBM. It was installed at UW's Advanced Research Computing Center over the past month and a half.
At a cost of about $1 million, it comes with about 350 terabytes of storage, but both its space and speed can be upgraded.
In fact, Kuhfuss said one nice thing about Mount Moran is that UW researchers have a stake in improving it.
Those who intend to use it are encouraged to contribute to buying more processor cores and storage so that its capabilities can continue to grow, benefitting the whole research community.
"You buy nodes and you get to use those nodes whenever you want them," Kuhfuss said. "But when you're not using them, they go into the community pool and anyone can use them.
"We're thinking we're going to put almost a $1 million a year into it, and that's primarily contributions from researchers."
Mount Moran will never come close to the size and strength of Yellowstone, but Kuhfuss said he's more interested in improving researchers' access to quality data modeling than trying to set records.
"We'll be very forgiving for codes that don't run well and help the researchers to fine-tune them," he added. "But I'm not going to play the game of getting it big just to get it big. We'll size it up to whatever makes the most economic sense."
UW professors already have begun using Mount Moran to run data computations and modeling with more expected to sign on in the coming months.
"I've got 30 accounts set up right now," Kuhfuss said. "I'm thinking we'll probably level off around 70-something once the word gets out."
(c)2012 Wyoming Tribune-Eagle (Cheyenne, Wyo.)
Visit Wyoming Tribune-Eagle (Cheyenne, Wyo.) at www.wyomingnews.com
Distributed by MCT Information Services
[ Back To IBM News 's Homepage ]