The National Institute for Computational Sciences

The Cluster Challenge

Competition Will Test UT Team in a ‘Hands-on Race to the Finish’

By Scott Gibson

The University of Tennessee’s SC13 Student Cluster Competition team has dedicated nearly a year to working with UT mentors to gain real-world experience in high-performance computing. Their abilities will soon be challenged in a contest involving students from around the world. Kneeling, the team’s primary mentor, Stephen McNally of the National Institute for Computational Sciences (NICS). Front row, from left Austin Carter (Oak Ridge High School, Oak Ridge, Tenn., student), Daniel Barry (Hardin Valley Academy High School, Knoxville, Tenn., student), Jeramy Harrison (UT Knoxville undergraduate). Back row, from left, Ben Olson (UT Knoxville undergraduate), Rory Fournier (UT Knoxville undergraduate), Adam Howard (UT Knoxville undergraduate). Image: Christal Yost, NICS.

Sometimes preparing for a contest offers a victory all its own, one of newly acquired confidence and knowledge. That’s true for the team of students that will represent the University of Tennessee when it participates for the first time in the annual international Student Cluster Competition at the SC13 supercomputing conference in Denver, Nov. 17–22.

Four UT undergraduates and two high school students from East Tennessee earned their places on the team roster by impressing staff members of UT’s National Institute for Computational Sciences (NICS), a supercomputing center, with both their self-starter attitudes and technical aptitudes during an interview process involving numerous applicants.

Since that time nearly a year ago, the team has been hard at work, winning a spot in the competition with a proposal they wrote themselves, selecting vendors from a number of possibilities, building their own small computing cluster within competition specifications, and gaining real-world experience in high-performance computing (HPC) as interns at NICS.

The competition is designed to introduce the next generation of students to the HPC community. During the last couple of years, it has drawn teams from around the world, including Canada, China, Costa Rica, Germany, Russia, and Taiwan. This year, UT’s team is one of eight from the U.S., Germany, China, and Australia selected to compete in the event’s Standard Track.

“We received 13 proposals for the Student Cluster Competition and planned to select six teams to compete, but the caliber of the proposals was so high we decided to go with eight teams,” said Student Cluster Competition Chair Dustin Leverman of Oak Ridge National Laboratory (ORNL). “Though the selection process was very difficult, it makes me confident that it will be a good competition this year. It’s always exciting to watch the students as they apply what they’ve learned in this hands-on race to the finish.”

When the event begins, the teams of undergraduate and/or high school students go non-stop for 48 hours, assembling a small cluster on the SC13 exhibit floor and striving to demonstrate the greatest sustained performance across a series of applications.

Helping the students get ready for the competition has been a center-wide effort, with a host of staff members chipping in to assist the students, said NICS HPC Operations Group Leader Stephen McNally, the team’s primary mentor. McNally, who initiated UT’s involvement in the competition, said NICS Director Greg Peterson has provided strong support to the project. “He has been completely on board from day-one, and has been very helpful in setting up contacts, in getting us the initial space on campus to use when we needed it, just being supportive of the whole effort overall,” he said.

NICS Security Officer John Wynkoop joined McNally in interviewing the students, assembling the team, soliciting vendor support, and leading the initial meetings. Computational scientist Vince Betro of the Application Acceleration Center of Excellence was instrumental in the role of liaison for systems and applications matters with Intel, one of the team’s vendors; the other is Cray. Both vendors have been “very good to work with,” McNally said.

“We’ve given the students some situational challenges leading up to the competition, and as much as possible, we’ve tried to emulate the environment,” McNally said. “The tricky part here is that no one from our organization has been involved in the Cluster Competition before, and so we’ve had to assume a fair amount of things. Essentially, we have instructed the students to be prepared for anything and everything at the competition. Actually, a lot of the teams that have participated in the past don’t know what to expect, either, because a mystery application will be involved.” The team will be required to run four scientific applications, three of which have been announced and one purposefully not revealed.

During SC13, teams will assemble, test, and tune their machines and run the HPC Challenge benchmarks until the starting bell rings on Monday night, Nov. 18, at the Exhibit Opening Gala, when they will be given the competition data sets for all four applications. In full view of conference attendees, teams will execute the prescribed workload while showing progress and science visualization output on large high-resolution displays in their areas. Teams race to correctly complete the greatest number of application runs during the competition period until the close of the exhibit floor on Wednesday evening, Nov. 20.

The showcase portion of the competition allows teams to show off what they’ve learned and what they can do with the equipment. Veteran HPC experts will be present to judge the visualizations and to interview each team on their cluster and application knowledge.

The winning team in each track will be determined based on a combined score for workload completed, benchmark performance, conference attendance, and interviews.

Since becoming members of UT’s Cluster Competition team, the students have spent most of their project time at ORNL, but have also had meetings on UT campus and remotely with one another on off-days. During the summer, they were at ORNL every day serving as interns, not only preparing for the Cluster Competition but also seeing HPC concepts in action in a real-life setting.

Team member Adam Howard, a UT undergraduate, said he and the other students have been taught what he describes as a “holistic process” encompassing how to set up a small computing cluster, recognize parallelizable problems, and troubleshoot hardware. He said he’s also discovered “what works in the HPC space, and what doesn’t.”

Daniel Barry of Hardin Valley Academy, Knoxville, Tenn., said he had never used LINUX and had a great time learning it. He added that he was drawn to the competition because he was really interested in computational science and research.

“I thought supercomputing was a cool word,” said UT undergraduate Ben Olson, describing his first response to hearing of the Cluster Competition. He credits his experience as a member of the team as enabling him to gain programming skills and understand how to “utilize massive resources.”

Howard said that what he found to be most valuable were the “people, networking, mentorship, and guidance.”

When the time to compete arrives, it will be the next stage in what has been a very positive experience for the students and their mentors. “Regardless of what happens, they’ve invested a great deal of effort and learned a lot along the way,” McNally said. “Based on what they’ve said to us, they’ve come to the realization on their own that they’ve really increased their knowledge. That tells me we did the right thing—we did our job.”