Tsinghua Student Supercomputing Team becomes the Student Cluster Competition (SC18) Overall Winner

On November 15, the Student Cluster Competition (SC18) concluded in Dallas. The supercomputing team from the Department of Computer Science and Technology (DCST), Tsinghua University, won the championship with a total score of 88.398 (full score of 100), 11.518 score higher than Nanyang Technological University of Singapore who won the second place. With this, Tsinghua team has won all three top international student cluster competitions in 2018, which are ASC, ISC and SC, and achieved another grand slam after 2015. It is also the 11th championship won by the team in these three top international student cluster competitions.

The Tsinghua team who took part in the competition are composed of 6 undergraduate students: Yu Jiping, Yu Xinjian, He Jiaao, Zheng Liyan, and Zhao Chenggang, who are all from DCST and Lou Chenyao, from the Institute for Interdisciplinary Information Sciences. Technical support was provided by Cao Huanqi, Feng Guanyu and Wang Miao, graduate students of the Institute of High-Performance Computing in DCST, who have participated in many international competitions. The advisors are Associate Professor Zhai Jidong and Postdoctoral Researcher Han Wentao from DCST. Meanwhile, five undergraduate students: Tang Shizhi, Chen Yu, Gu Yuxian, Chen Shengqi and Zhang Chen, also participated in the training.

Undergraduate students from 15 universities all over the world formed teams to take part in the competition. During the competition, the teams were required to design and build their own computing clusters and compete in the performance of 6 applications. The power of the cluster must not exceed 3000 watts during the entire 48-hour competition. The applications are High-Performance Linpack (HPL) benchmark, High Performance Conjugate Gradient (HPCG) benchmark, parallel deep learning, particle transport simulation, earthquake simulation, and a mysterious application announced on-site, which is weather forecasting model. Meanwhile, sessions including interviews, attending expo and poster design were also set up, where the team members were required to introduce their optimization on applications to the judges in the competition, as well as writing a report which is eligible to be published in an academic journal within 48 hours.

Writer: Wilson Khor
Editors: John Olbrich, Guo Lili