Competitors Represent Growing Diversity, International Interest
This year’s competitors for the Student Cluster Competition (SCC), showcase a mix of new and returning talent. From of a pool of 30 applicants, 13 teams were selected to compete, where they will be joined by last year’s SCC winners, Tsinghua University (China), along with the Asian Supercomputing Challenge (ASC19) champions from National Tsing Hua University
(Taiwan). The winners of the ISC19 SCC, being held later this month in Frankfurt, Germany, will round out the SC19 SCC lineup.
Meet the Teams
Among the teams selected are six brand-new groups, while some of the returning teams have included members from new educational institutions as part of their HPC squads. Notably, diversity in the membership among the teams continues to grow.
- Tsinghua University, last year’s champions, promises to field a potent team and to continue their legacy at the SCC. Their baseline architecture is a multi-node cluster of Intel Xeon CPUs and NVIDIA V100 GPUs. However, depending on the applications to be run during the competition, their setup is subject to change to something different, such as ARM, field-
programmable gate arrays (FPGAs) or a related hybrid.
- Twitter: @Tsinghua_Uni
- National Tsing Hua University, The ASC19 winners, remain a perennial contender at the SCC as well, achieving the highest LINPACK scores at SC07 and SC14, and winning the competition at SC10 and SC11. They plan to bring a five-node cluster with Intel Xeon CPUs and NVIDIA V100 GPUs as they vie to sustain their performance from ASC at SC19.
- Facebook: @nthu.tw
- Team RACKlette from ETH Zurich (Switzerland) is new to the competition. Although the team participated in previous ISC competitions, this will be its first time at SC. The team members represent five countries and two continents and are supported by ETH Zurich and CSCS-Swiss National Supercomputing Centre. They plan to bring a four-node cluster with Intel Xeon CPUs and NVIDIA V100 GPUs.
- Team GeekPie_HPC from ShanghaiTech University (China) has participated in the ASC and ISC competitions but are new to SC19 SCC. The group derived its name from the GeekPie_ student association for computer science enthusiasts at ShanghaiTech. Their base architectural design consists of four Intel Xeon/NVIDIA V100-heavy nodes, but they also are considering building an ARM-based HPC cluster or using FPGAs.
- The Nanyang Technological University (Singapore) team is another perennial powerhouse at all SCC events. They won the competition at SC17 and achieved record-breaking LINPACK scores at SC17 and SC18. In seeking to claim the SC19 SCC championship, the team plans to bring a cluster with one NVIDIA GPU-heavy node and four Intel Xeon CPU-heavy nodes.
- The Joint Baltic Team (Estonia/Latvia) also is new to the SC19 competition. With team members from the University of Tartu (Estonia), Riga Technical University and Riga Technical College (Latvia), the Estonians have experience from this year’s ASC and ISC competitions, while the Latvians are entirely new to cluster competitions. The team plans to bring a cluster with AMD Rome CPUs and NVIDIA Tesla V100 GPUs.
- Team Segfault from Friedrich-Alexander-Universität Erlangen-Nürnberg (Germany) is another recurrent competitor at the SC competition. They plan to bring two NEC SX-Aurora Tsubasa nodes, each containing two 12-core Intel Xeon Skylake processors with 192 GB memory, plus eight type-10B SX-Aurora Tsubasa vector engines. The team believes its novel architecture with 16 total vector engines will be competitive versus the V100 setups expected in other teams’ clusters.
- Purdue University (USA) has competed at SC, ISC and/or ASC since the first SCC at SC07 and every year thereafter. One member of last year’s all-female team is serving as a team advisor for SC19, while another is a returning competitor. In addition, Purdue sponsored an all-high-school team at SC17, and a member from that squad has since joined the “big league” team for this year’s SCC. Purdue plans to bring a Dell cluster consisting of four nodes with Intel Xeon CPUs and NVIDIA V100 GPUs.
- While Oak Ridge National Laboratory has supported University of Tennessee (USA) teams in past competitions, this year they have elected to host a cross-institutional team with students from the University of Tennessee, Pellissippi State Community College (USA) and Maryville College (USA). The team is coached by two ORNL staff members who are SCC alumni. The team plans to bring a cluster consisting of one GPU-heavy node with NVIDIA V100 GPUs, four CPU-heavy nodes with Intel Xeon CPUs, and a low-power Intel NUC head-node.
- Wake Forest University (USA) marks its second foray in the SCC at SC19. During SC18, the team was resilient in the face of adversity: when the rack for their cluster did not show up, they found a used one for sale on Craigslist and rented a U-Haul truck to pick it up and haul it back to the exhibition hall. This year, the team is planning to employ a single-node system with eight NVIDIA V100 GPUs.
- Team Warsaw (Poland) is another repeat competitor, drawing students from the University of Warsaw, Warsaw University of Technology, Lodz University of Technology, University of Wroclaw, and Military University of Technology. These universities have established a robust pipeline of students learning about HPC, including graduates who are ineligible for the competition but can share their wisdom, novices who are coming up to speed for future competitions and the team that submits the proposal. They plan to bring a four-node cluster with Intel Xeon CPUs and NVIDIA V100 GPUs.
- Inspired by her participation in the SC18 “Experiencing HPC for Undergraduates” program, one student spearheaded the efforts for the North Carolina State University (USA) to field a team for the SC19 SCC. These first-time competitors have support from Cisco to build a three-node system with Intel Xeon CPUs and NVIDIA V100 GPUs for the competition.
- Peking University (China) returns to the competition after a one-year hiatus. The team has members originating from provinces across China. They propose a system consisting of Intel Xeon CPU-heavy nodes and NVIDIA V100 GPU-heavy nodes.
- Shanghai Jiao Tong University (China) is new to the SC19 SCC, but they are experienced, participating in and winning several ASC and ISC competitions over the years. This time, they will see how their experience translates to the SC19 competition. The team plans to bring an eight-node cluster with Intel Xeon CPUs accelerated with NVIDIA V100 GPUs.
- Finally, the University of Washington (USA) team may be new to the SCC, but they are by an advisor to a previous championship team (Indiana University, SC08). The team’s vendor partner, Amazon, also employs an SCC alumnus who serves as their liaison. The team to bring a three-node cluster with Intel Xeon CPUs and NVIDIA V100 GPUs.
About the Student Cluster Competition at SC
Held annually during the SC since 2007, the SCC is the original student cluster competition. In this non-stop, 48-hour showdown, teams of six students who have not yet attained their undergraduate degrees compete to complete a workload consisting of real scientific applications without exceeding a 3000-watt power limit on small supercomputing clusters they have designed and built.
Rebecca Hartman-Baker, SC19 Student Cluster Competition Chair
Rebecca Hartman-Baker leads the User Engagement Group at NERSC, where she oversees engagement with the NERSC user community to increase user productivity via advocacy, support, training, and the provisioning of usable computing environments. Rebecca is the SC19 Student Cluster Competition Chair, and previously coached two Australian teams for the cluster competitions at SC13 and SC14. She also coached two all-female teams representing NERSC for competitions at ISC16 and ISC17.