Staff from the Department of Energy’s Oak Ridge National Laboratory returned this year to mentor students in the 2023 Winter Classic Invitational Student Cluster Competition. The virtual event invites student teams from historically Black colleges and universities and minority-serving institutions to take on high-performance computing challenges from mentor organizations. This year’s competition featured 12 student teams and mentors from ORNL, Amazon Web Services, Hewlett Packard Enterprise and NASA.

The Winter Classic Invitational, which debuted in 2020 and is held annually from January to April, is organized and led by Dan Olds and Addison Snell of Intersect360. The competition has a unique format designed to level the playing field for all teams. Unlike other student computing competitions in which students are required to source and build their own computing clusters, the Winter Classic Invitational’s mentors provide all the necessary compute, training and technical assistance to each team to ensure that everyone has an equal opportunity to succeed. The competition is also entirely virtual to enable teams from anywhere in the country to compete.

“It’s cool to see that there are opportunities for students to get into high-performance computing through this route when they don’t have the traditional on-ramp,” said John Holmen, an HPC engineer and ORNL mentor.

Holmen is one of five staff members from ORNL’s National Center for Computational Sciences who offered their expertise for this year’s competition. The mentor team was led by group leader Verónica Melesse Vergara and HPC engineer Dan Dietz and also included Suzanne Parete-Koon and Elijah MacCarthy. Together, they organized access to the ORNL compute environment, provided students with an introductory course in HPC and developed a multipart challenge that used the exascale-class computational astrophysics application AthenaPK.

“This year, we combined the simulation and visualization, so students got to see the full pipeline of running the code, analyzing the data and visualizing it to understand what is happening,” said Melesse Vergara.

The HPC crash course is a staple of ORNL’s mentorship strategy. Versions of the course have been used for over 10 years to introduce students and non-technical professionals at ORNL to the basics of HPC. Recently, Parete-Koon and colleagues in the NCCS have updated the course with a modular, customizable curriculum that covers HPC fundamentals such as job scheduling and programming in C and Python. For this competition, Parete-Koon led students through training and hands-on practice problems to acquaint them with Ascent, the single-cabinet test system with the same architecture as ORNL’s IBM AC922 Summit supercomputer, and prepare them for the official challenge.

To provide the students with a realistic HPC experience, Holmen and MacCarthy designed this year’s ORNL challenge around AthenaPK and the VisIt visualization tool.

“Our challenge let the students see what HPC can be used for. It’s not just about speeding stuff up and getting the best performance,” said MacCarthy.

Based on adaptive mesh refinement, a technique used to refine specific areas of a grid simulation to obtain accurate results more quickly, AthenaPK can simulate various astrophysical and cosmological phenomena. The 12 teams used this technique to explore the propagation of a blast wave, a common occurrence in astrophysical settings, and discover a secret hidden by Holmen and MacCarthy.

“We tried to make it fun for them, and we gave them an unidentified image — the ORNL oak leaf — and had them experiment with different mesh resolutions and grid configurations to try to resolve the image,” said Holmen.

An oak leaf rendered on a grid of varying density

Student teams used adaptive mesh refinement to resolve the image of the Oak Ridge National Laboratory oak leaf. Credit: ORNL

The students handily met the challenge, with one team successfully scaling the application to use all the GPUs on the node and even creating a video of their visualization. The range of skills required by the ORNL challenge stood out for several students, including Nicholas Anderson, who competed on the Matadors team from Texas Tech University.

“I very much enjoyed the ORNL challenge portion of this year’s competition. Having to visualize what we were doing was a great addition that none of the other challenges had us do,” Anderson said. “I liked how ORNL’s challenge had us doing a little bit of everything since we had to configure, build, run and visualize using the HPC system that was generously partitioned for us.”

For the mentors, the victory lies in introducing students to HPC and building skills and experience that they can then apply in their careers.

“It’s not just about the challenge they worked on but all the other skills, too. I’m glad that we could contribute to this aspect of their careers and education,” said MacCarthy.

UT-Battelle manages ORNL for the Department of Energy’s Office of Science, the single largest supporter of basic research in the physical sciences in the United States. The Office of Science is working to address some of the most pressing challenges of our time. For more information, please visit