This Purdue team is super — at supercomputing
October 25, 2012
Whether it’s developing new cancer treatments or tracking down the particles that make up the universe, not to mention powering Amazon or iTunes, Nick Molo knows high-performance computing is an essential tool.
So the Purdue junior in computer engineering figures his experience on a team of students building and running its own supercomputer in an international competition can’t help but enhance his prospects after graduation, even if it means keeping long hours this semester and a few sleepless nights.
Molo is part of the six-member Purdue team competing in the 2012 Cluster Challenge, the student competition at SC12, the world’s largest supercomputing conference. The conference is being held in Salt Lake City Nov. 10-16. ITaP is sponsoring the Cluster Challenge team in partnership with HP.
The Purdue team, one of eight selected, will be competing against teams from China and around the U.S. Purdue’s entry is the only one from the Midwest and the Big Ten.
“My short-term goal is to get a good job after graduation that would potentially pay for me to go back to school to further my education,” says Molo, from Joilet, Ill. “The Cluster Challenge already has given me a leg up on the competition. It gives me a great talking point on my resume while also being a great learning experience.”
In addition to Molo, the team includes Cody Breedlove, a senior in computer science from Royal Center; Andrew Huff, a junior in computer science from Cary, N.C.; Kurt Kroeger, a sophomore in computer science from Nashville, Tenn.; Ethan Madden, a junior in computer science from Newburgh; and Tyler Reid, a senior in computer science from Zionsville. ITaP staff members Stephen Harrell and Andy Howard are advising the team, assisted by ITaP staff members Patrick Finnegan and Alex Younts.
With 128 processors, about 30 to 60 times more than a typical personal computer, the students’ entry is a mini version of the Purdue’s new Carter cluster supercomputer. Carter, with more than 10,000 processors, ranks as one of the most powerful research supercomputers in the world. Purdue faculty researchers in aeronautics and astronautics; earth, atmospheric and planetary sciences; economics; genomics; and more use Carter, which is one of five shared community clusters built by faculty partners and ITaP since 2008.
Likewise, the Cluster Challenge teams — limited to undergraduates — have to prepare their machines to run an assigned variety of real research software crunching voluminous sets of sample data as quickly and efficiently as possible. The 2012 applications are used for simulating the actions of atoms, modeling the global climate, probing physics at the very small quantum level and predicting the spread of contaminants underground. Most of this year’s Purdue team helps manage Carter and Purdue’s other community clusters as student workers for ITaP.
The students say their work with ITaP is giving them a good grounding in managing supercomputing hardware. The Cluster Challenge gives them an opportunity to learn more about the software faculty researchers employ on Purdue’s clusters and to get a look at the systems they help manage from a user’s perspective.
“As a student system administrator, I work on keeping the clusters up and running so users can run their applications without interruption,” Kroeger says. “I've always wondered what types of programs researchers use and what their function is. The Cluster Challenge has allowed me to work on many different applications.”
Harrell sees that as one of the benefits the students get from being on the team.
“I think educationally I'd really like them to whet their appetites for computational science,” Harrell says. “To me it's a pretty amazing thing when you think about it, modeling very tiny things or very large things or very hot things. I'd like them to share some of that excitement and maybe consider it as a career.”
At the supercomputing conference, the students share the load monitoring and adjusting their supercomputer around the clock as they work to process as much data as possible over 48 hours. Reid, a member of Purdue’s team last year, says “watching the sun set and rise over Seattle during the first overnight shift” is his most memorable experience from the 2011 competition.
“I wanted to participate in the Cluster Challenge again because it was an awesome experience,” Reid says. “I really enjoyed meeting the students on the other teams and other people interested in high-performance computing. I hope to get a job in the high-performance computing industry or in an industry where I can continue to work on something I'm passionate about.”