CFD training for the hydropower sector

Supercomputing for first-time users

CFD applications require much computing power and are most efficiently run on high-performance computing (HPC) systems with large throughput local storage, low latency networks, and optimized CPUs. Because the participants of the innovation camp had never used HPC before, an additional workshop on HPC and using HPC systems such as VSC clusters was an important part of the training.

It soon became clear, however, that the traditional HPC workflow would pose an entry hurdle for the participants. Conventionally, one would have to create a model on a local machine, check the mesh for inconsistencies, set the appropriate boundary conditions and parameters and only then send the resulting file to the supercomputer, write a job script, submit the job, transfer the outcome back to the local machine in order to visualize and interpret the results.

To cater for a convenient work environment on the HPC systems, the technical team provided
GPU-accelerated rack workstations on VSC-4 for the training. This enabled the participants to run their simulations entirely on the supercomputer. Thanks to the GPUs in the rack workstations, the graphical output was fluid. It was the first time that the rack workstations were used in a larger training event, and the VSC staff did everything possible for it to go smoothly.