BioExcel Webinar #34 - MDBenchmark: a tedious task – made straightforward

Molecular dynamics software is optimized to yield the highest performance on available resources. The details of why computations perform better or worse depends on various parameters. As an example, common biomolecular simulations stop to scale linearly when too much computational power is provided. Optimization of available parameters is a tedious task and is perfectly suited for automation procedures.
MDBenchmark is a tool that provides a straightforward way to perform scaling studies of molecular dynamics simulations across a different number of nodes, as well as different parameters. Given an input file, e.g., a TPR file for GROMACS, MDBenchmark handles submission of jobs to any given queuing system, retrieval of performance data and creation of a plot for an easy estimation of the most efficient usage of computational resources.
In this webinar, we will present how to use MDBenchmark, starting with the installation process, going over benchmark generation, submission and analysis. Furthermore, we will provide explanations on how to set up MDBenchmark and customize submission scripts, to perform scaling simulations on any high performance cluster. We will also discuss the future vision for the project and how anyone can get involved in the development.