Ohio’s New Supercomputer Runs 144 Trillion Calculations per Second

Even though the $1.5 million system is just a few months old, the Ohio Supercomputer Center already is preparing to buy a more-powerful supercomputer by the end of the year.

by Maria DeVito, The Columbus Dispatch / May 26, 2015

(TNS) -- What Ohio’s newest supercomputer lacks in size, it more than makes up for in speed.

It’s called the Ruby Cluster and was unveiled recently at the Ohio Supercomputer Center.

The cluster is about one-third the size of the former champ — the Oakley Cluster, which was installed in 2012. But Ruby is three times as fast as Oakley, said Brian Guilfoos, client and technology support manager at the supercomputer center.

Supercomputers are a faster version of a desktop computer. A calculation that would take a desktop months or years to complete can be done by a supercomputer in hours or days.

The center, located on Ohio State University’s West Campus, has three supercomputer clusters, or groupings of servers. Besides Ruby and Oakley, there’s also Glenn.

All are named for Ohioans: Ruby for Ohio-born actress and activist Ruby Dee; Oakley for famed Wild West sharpshooter Annie Oakley; and Glenn for astronaut and former U.S. Sen. John Glenn.

Each cluster is composed of several racks that contain nodes, which are boxes that hold the processors for the supercomputer. At total peak performance, the Ruby Cluster operates at 144 teraflops, which means it can perform 144 trillion calculations per second.

What makes the Ruby Cluster so much quicker than Oakley are the processing cores and accelerators in each of its nodes. A node inside Oakley has 12 processing cores. A node inside Ruby has 20.

“The processing cores in Ruby are newer and much more compact and, therefore, can exchange information much more quickly than those in Oakley,” said Jamie Abel, spokesman for the center.

That kind of power takes a lot of energy. Ruby uses enough to light up 73 houses.

The center uses half a megawatt of power at any given time to run the three supercomputers and an additional half-megawatt to cool the systems. A megawatt can power 806 homes, said Alan Chalker, director of technology solutions.

The supercomputers use all that energy to perform dozens or hundreds of calculations at a time.

One recent project involved calculating how to make tractor-trailers more aerodynamic to save on fuel costs.

“We can computationally model that rather than actually putting a truck in a wind tunnel to watch what happens,” Guilfoos said. “In this particular instance, we’re modeling air flow over a truck to evaluate the effect of those little flaps you see on the backs of those trucks … to see how much drag they can save and then how much fuel they can save by putting those things on.”

Running a computer program is a lot cheaper than putting a truck in a wind tunnel. Plus, there is only one wind tunnel in the United States that is large enough to accommodate a semi.

“You can spend a couple thousand dollars and run this sweep of simulations, whereas you might spend a hundred thousand dollars to do one test,” Guilfoos said.

Ohio State uses the center the most, but the University of Cincinnati, Bowling Green State University, the University of Toledo and Nationwide Children’s Hospital also use the supercomputers. The center has commercial clients, too, most of which are aerospace engineering consultants.

Even though the $1.5 million Ruby Cluster is just a few months old, the center already is preparing to buy a more-powerful supercomputer by the end of the year, said David Hudak, director of supercomputer services.

“It was really a transitional system between Oakley and the next system,” he said. “It’s amazing how quickly these machines fill up.”

With constantly advancing technology, supercomputers quickly become outdated.

When Oakley was brought online in 2012, part of Glenn was phased out. After the center acquires a new supercomputer in the fall, Glenn will be taken out completely.

The center, which is funded by the state, has $12 million in its budget for a new cluster, which must cover the supercomputer, infrastructure changes and data storage.

“We have to make sure we have sufficient storage to support all of the data that’s needed or generated from the simulations that are run on the systems,” said Douglas Johnson, chief systems architect and high-performance-computing-systems manager. “For every $4 spent on computing, we have to spend a dollar or so on storage.”

Ohio’s center is among 82 academic-based supercomputer centers throughout the country.

“I suspect that the new system will put Ohio in the top five in terms of capacity,” Hudak said.

©2015 The Columbus Dispatch (Columbus, Ohio) Distributed by Tribune Content Agency, LLC.