Scaling CFD a lot by breaking down a workflow, to speed things up

Brendan Bouffler
Date : March 3, 2022
Categories : AWS ParallelCluster ,Amazon NICE DCV ,Life Sciences
Tags : cfd ,cpus ,covid 19 ,dcv ,domain decomposition ,ec2 ,gpus ,hpc ,high performance computing ,lustre ,meshing ,parallel cluster ,schedulers ,solver performance ,storage ,clusters ,hierarchical ,openfoam ,redistribute par ,scotch ,snappy hex mesh ,virtualization ,vizualization ,techshorts

Sometimes when we’re looking for performance we lose the forrest for all those trees - we miss the huge improvements we can do to unglamorous parts of the overall workflow, while we obsess on the pieces that look hard. It’s an engineer thing, I think.

In today’s Tech Short, Neil Ashton shows us exactly this kind of example from the world of CFD (using OpenFOAM, but this lesson applies generally) - and shows us how to break the problem down in order to speed it all up - and pretty easily, too.

The workshops we reference in the discussion are all listed here: https://aws.amazon.com/hpc/cfd/

If you have ideas for technical topics you’d like to see us cover in a future show, let us know by finding us on Twitter (@TechHpc) and DM’ing us with your idea.