It’s gratifying when you see the premier organizations in our industry step up to the plate and support a good thing. The good thing? It’s the Winter Classic Invitational Student Cluster Competition.
This competition is the first of its kind. It’s targeting Historically Black Colleges and Universities (HBCUs) and Hispanic Serving Institutions (HSIs), to give them a chance to participate for the first time in these valuable learning events.
In this virtual competition, student teams will be running benchmarks (HPL and HPCG) along with a set of real-world HPC applications (NAMD, OpenFOAM, LAMMPS, and Gromacs) on systems provided by their mentor organizations. The objective is to see which team can best optimize the workloads.
All of these teams (except Florida A&M who competed at SC way back in 2010) are complete newbies when it comes to cluster competitions and, for the most part, big time HPC. But they won’t be flailing around on their own. Each mentoring organization is offering up coaching along with use of their hardware. The students also have faculty coaches who will help guide them along the way, plus a plethora of materials provided by competition organizers.
There aren’t many classes on HPC in universities these days and we’re hoping that our competition will show students that there is this whole other part of the computing industry that they might not have been exposed to yet. One that works to solve the very largest challenges on the very largest and fastest systems in the world.
We extend our deep and sincere gratitude to our slate of mentors for participating in this competition and supporting the student teams. This will be an annual event, so we would encourage other organizations (vendors, labs, HPC shops in commercial/industrial orgs.) to look into mentoring teams for the 2022 competition.
With that, let’s take a look at the mentors for this year’s crop of cluster competition competitors. These organizations are spending time and money to give students an opportunity to learn about and join our HPC community, so give them a big hand.
Google: has stepped up in a big way by supporting three teams for the Winter Classic Competition. These teams include California State University, Santa Cruz, University of Texas, El Paso, and University of Houston. This is Google’s first foray into any student cluster competition and they’re not just dipping their toes in the pool – they’re jumped in headfirst by sponsoring three teams.
The teams will be running on Google’s Cloud Platform, which can offer anything from CPU-only clustered instances to multi-GPU instances with high GPU to CPU ratios. The teams and Google experts are still deciding exactly what configuration(s) they will use. They have time to test and deliberate as their final configurations don’t need to be locked down until early April.
Dell: isn’t a stranger to the world of big-time student cluster competitions. They’ve sponsored student teams in nearly every major event in the past decade, making them the vendor who has provided the longest (and most) support for these competitions – thank you Dell!
Dell started by sponsoring the multiple championship winning University of Texas, Austin teams in 2010. The company also underwrites and facilitates the South Africa’s Centre for High Performance Computing internal cluster competition, which selects the team that will be sent to the International Supercomputing Conference (ISC) event in Germany. These teams have also won multiple championships vs. a highly competitive international field of competitors.
In the Winter Classic, Dell is supporting Tennessee State University and Prairie View A&M University teams. Dell will be partnering with the HPC AI Advisory Council to provide the teams with Dell hardware for their clusters.
Microsoft/Intel: will also be supporting two teams in the upcoming Winter Classic, Prairie View A&M University and Tennessee State University. Microsoft has played a large role in student cluster competitions in the past, providing compute resources for the entire SC20 virtual competition most notably. Intel has always been supportive of the competitions, but mostly from behind the scenes.
The teams will be running their workloads in Microsoft’s Azure cloud, where they can run on a dizzying array of configurations. But, like all the teams, they’ll need to make their final configuration decision by early April. Oh, and one more thing. I’ll tell you this as long as you keep it to yourselves….Microsoft is generously giving each team member a very nice gift for participating in the competition. More on this later, but it’s a big secret, so don’t say anything to anyone.
HPE/Intel: HPE has sponsored several teams in past SC student cluster competitions and they’re a welcome addition to the field of mentors in the Winter Classic. This year, the combined HPE/Intel organization is sponsoring Claflin University from Orangeburg, South Carolina.
Like all of the schools in this competition, Claflin does offer a Computer Science major, but, again, like most of these schools, they don’t have a lot of courses that cover HPC concepts. This is where the competition can help give the students exposure to our industry and encourage them to learn more.
Lawrence Livermore National Laboratory: has become the first national laboratory to mentor a team in the Winter Classic competition – and they’re taking on this responsibility with gusto. They’re mentoring the Florida A&M University and the University of the Virgin Islands teams, providing them with compute resources and an amazing array of coaching and teaching resources.
The LLNL personnel are very responsive to student questions and have laid out milestones and weekly goals for their teams. They’ve put together a concerted effort to get their teams up and running their workloads on the LLNL nodes and are making very good progress. LLNL’s Tony Baylis has been a key driver of the competition and helped to recruit and mentor several of the teams in this year’s field.
We also need to thank the folks at Arm are providing access to their Arm Forge suite to each of the student teams, free of charge. This will help the students characterize, profile, and optimize their applications. This will be particularly helpful to these students as many of their universities don’t offer a lot of classes that cover HPC or how to optimize applications.
Brookhaven National Laboratory also provided funding that allows student teams to compete at this event, so a big thank you to them.
From now until April 19, the students will be working with their mentors to learn how to use the hardware, run the applications, and optimize the codes. On April 19, the coaches and mentors exit and the students begin the final competition week. This is where it’s all on the line.
On Friday, April 23, the students have to turn in their final results and then undergo judges interviews the following week. The judges will be looking to see how much the students have learned, how well they understand their hardware and software, and how well they worked together as a team.
Finally, on April 30, the winners will be announced and the individual awards handed out at our online Awards Gala. Details to be announced later.
Next up, we’ll meet the teams in our patented up close and personal video interviews. Don’t miss it and stay tuned for more Winter Classic Cluster Competition coverage.