Student Cluster Competition


During my parallel programming class I heard of an opportunity for students to build and configure a high performance computing cluster. I applied to join the six-person team and (surprisingly) was accepted. Over the next few months we practiced running applications on Texas Advanced Computing Center (TACC)’s systems, designed the hardware configuration of our cluster, and learned about Linux system management and control. My main tasks were running and tuning High Performance Linpack, writing scripts to automatically build, run, and generate matplotlib graphs that analyzed the performance of several applications. I also created a power monitor system using SNMP communication, which stores data in a graphite database and visualized with grafana. In the Fall our hardware came in, and we assembled and configured our cluster of 12 nodes by using Open HPC to bootstrap a few representative nodes, then cloned out their images to the rest of the cluster. On November 11th we flew out to Salt Lake City to the SC16 Conference where we competed against teams from across the world for 48-hours straight. My primary task during this time was learning to build and use the molecular dynamics application GROMACS, which was the ‘mystery application’, so we didn’t know about it until the start of the application. We placed 4th out of 14 teams overall, which given that our team was made up entirelly of new competitors I was very pleased with.