Undergraduates gain experience using Dell HPC solutions

AUSTIN, Texas, Nov. 24 -- University of Texas at Austin students last week competed in the SC10 Student Cluster Challenge in New Orleans and came in second overall, including winning the HPL (High Performance Linpack) benchmark award.

HPL is the most common performance benchmark for supercomputing systems, and the student team exceeded one teraflop on HPL for the first time in the competition while staying below the 26 Amp constrained power budget. A teraflop is a measure of computing speed equal to one trillion floating point operations per second.

The 46-hour, non-stop competition challenged students from around the world to build, maintain and run the most-cutting edge, high performance computing (HPC) architectures out of commercially available components, using no more power than that of three coffeemakers.

The students -- Bethany Barrientos, Vladimir Coxall, Alex Heinzmann, Jason Kilman, Loren Micheloni and Phillip Verheyden -- are from the Computer Science and Applied Mathematics departments, and worked under the mentorship of Byoung-Do Kim and Carlos Rosales-Fernandez, both research associates in the High Performance Computing Group at the Texas Advanced Computing Center (TACC) at The University of Texas at Austin.

"Six months ago, the students didn't have any experience in HPC and were overwhelmed by the magnitude of the project, but they've come a long way and shown a lot of enthusiasm," Kim said. "They quickly learned what they needed to master and worked as a team throughout the project. We're looking forward to competing again next year and keeping the Texas pride high at SC11."

"We were the first team in the competition's history to break a teraflop under the 26 Amp limit," said Phillip Verheyden, the team captain and a student majoring in computer science. "It's a testament to our tireless work in power testing and optimizing the benchmarks. Going from zero knowledge to second place in an international competition showed our dedication and our abilities, but was also highlighted all the support we received from our advisers in the process."

National Tsing Hua University from Taiwan was the overall winner with the highest aggregate points in the High Performance Computing Challenge (HPCC) set of benchmarks, throughput and correctness of business-world applications, and interviews. HPCC consists of seven benchmarks, including HPL, and is intended to provide a realistic measurement of modern computing workloads. The benchmarks are scalable, and can be run on a wide range of platforms, from single processors to the largest parallel supercomputers.

Verheyden said: "The Taiwanese team had the exact same configuration as us, but used one fewer node than we did. We did a lot of power testing prior to the competition and saw just how far we could stretch that power envelope to eek as much performance out of the system as possible."

Dell sponsored the SC10 UT Austin/TACC team by donating Dell PowerEdge R410 servers with low-power Intel processors and low-power memory to meet the strict energy limitations of the competition. Dell team members provided guidance and instruction to the team on hardware and HPC applications.

"SC10 is the biggest international meeting of the year in supercomputing, and having our first-ever student team win HPL and place second overall is amazing," said Jay Boisseau, director of TACC. "We embrace opportunities to work with and train the next generation of computational scientists, and our partnership with Dell enables that commitment."

John Mullen, Dell's vice president and general manager for higher education, said, "Dell has been a supporter of the supercomputing conference Student Cluster Challenge with our energy-efficient PowerEdge servers and team member volunteers since it first began four years ago. We congratulate the UT Austin/TACC student team for achieving best benchmark performance and for bringing a high level of excitement and energy into the research community, the competition and conference overall."

-----

Source: Texas Advanced Computing Center