A poor choice of job submission or simulation parameters leads to poor performance and a waste of computing resources. A common mistake is requesting too many CPUs or GPUs to run a simulation as fast as possible. The fastest simulations may be inefficient due to poor parallel scaling. As a result, such simulations use excessive resources and negatively impact priority, resulting in longer queue times and reduced productivity.
Finding optimal MD engines and submission parameters in a complex HPC environment with heterogeneous hardware poses a daunting and time-consuming challenge. We designed this interactive performance guide to simplify optimization of MD simulations. This web portal presents performance metrics of all installed MD software packages. It includes a searchable benchmarks database with simulation details such as hardware descriptions, SLURM submission scripts and input files for simulations.
Currently, the database contains benchmarks for four major MD software packages: AMBER, GROMACS, NAMD and OpenMM. The efforts to test all possible hardware/software combinations are underway.