Biophysical Society Thematic Meeting | Hamburg 2022

Biophysics at the Dawn of Exascale Computers

Poster Abstracts

31-POS Board 31 GROMACS IN THE CLOUD: HIGH THROUGHPUT LIGAND SCREENING ON A GLOBAL SUPERCOMPUTER Carsten Kutzner 1 ; Christian Kniep 2 ; Austin Cherian 3 ; Ludvig Nordstrom 4 ; Helmut Grubmüller 1 ; Bert L de Groot 1 ; Vytautas Gapsys 1 ; 1 Max Planck Institute for Multidisciplinary Sciences, Theoretical and Computational Biophysics, Göttingen, Germany 2 Amazon Web Services, Amazon Development Center, Berlin, Germany Biomolecular simulation is a challenging example of a compute-heavy scientific application that spans the whole range from high performance computing (HPC) to high throughput computing (HTC), depending on the questions addressed. We assess costs and efficiency of state-of-the-art high performance cloud computing and compare the results to a typical in-house cluster for molecular dynamics (MD) simulations carried out with the GROMACS MD software. We set up an HPC cluster in the Amazon Web Services (AWS) cloud that incorporates various different instances with Intel, AMD, and ARM CPUs, some with GPU acceleration. Using representative biomolecular simulation systems we benchmark how GROMACS performs on individual instances (for HTC) and across multiple instances (for HPC scenarios). Thereby we assess which instances deliver the highest performance and which are the most cost-efficient ones for MD. We find that, in terms of total costs including hardware, personnel, room, energy and cooling, producing MD trajectories in the cloud can be about as cost-efficient as an in-house cluster given that optimal cloud instances are chosen. Further, we find that high-throughput ligand-screening can be accelerated dramatically by using global cloud resources. For a ligand screening study consisting of 20,000 independent simulations or 200 microseconds of combined simulation trajectory, we used all the hardware that was available in the cloud at the time of the study. Using more than 4,000 instances, 140,000 cores, and 3,000 GPUs around the globe, our simulation ensemble that would normally take weeks to complete on a typical in-house cluster consisting of several hundred nodes, finished in about two days in the cloud. We demonstrate that the costs of such and similar studies can be drastically reduced with a checkpoint- restart protocol that allows to use cheap Spot pricing and by using instance types with optimal cost-efficiency. 3 Amazon Web Services, Singapore, Singapore, Singapore 4 Amazon Web Services, London, London, United Kingdom

56