ITaP pros help give student supercomputing teams a challenge at the SC13 supercomputing conference

November 26, 2013

This time of year, ITaP staff member Stephen Harrell would normally be riding herd on Purdue’s student supercomputing team as it competed at the annual SC supercomputing conference.

However, Purdue sent a team to the international student supercomputing competition in Germany earlier this year and so sat out the U.S. version, nicknamed the Cluster Challenge, which is held annually in conjunction with the SC conference, the largest high-performance computing showcase in the world.

SC 13 was held in Denver last week and the student competition went on as usual with U.S. teams traveling from schools spread from Massachusetts to California to participate, as well as international teams from Asia, Europe and Australia.

Harrell, a high-performance computing specialist for ITaP, was there, too.

He was looking to beat the students at their own game.

Well, not really. But Harrell and former ITaP staff member Andy Howard, a repeat member of Purdue’s Cluster Challenge team when he was a student, set up the supercomputer that a group of high-performance computing professionals used to challenge the student teams after the students were done challenging each other.

“We basically took it from bare hardware to a running machine,” Harrell says of the four-node cluster loaned to the pros for the event by supercomputer maker Cray.

In addition to Harrell’s role at SC13, Verónica Vergara, a scientific applications analyst at ITaP Research Computing (RCAC), served on the student cluster competition's organizing committee and was its science lead. She selected the applications used in the student competition.

The Student Cluster Competition Pro-Am Challenge was a new event at SC13, built around the same format as the student cluster competition. The pros, like the students, had to set up their own supercomputer. They and the student competitors were then given a surprise application to run, with the winner determined based on how fast they could get their machine to run it in a limited amount of time.

Harrell was asked to participate because of his long involvement with the student cluster competition at SC. While he and Howard are proud of their abilities and did everything they could to set up a winning machine, Harrell thought it would actually be exciting for the students if they won.

“That’s what the Cluster Challenge is about,” Harrell says. “It’s about getting kids interested in high-performance computing and letting them have a chance to do it.”

In the end, the student teams combined their resources and expertise to take down the pros. Harrell sees that as a big win. He says it allowed the students to work together across teams, where they normally wouldn't, and challenged them to find solutions for combining their disparate hardware and software stacks on the fly.

Originally posted: November 26, 2013