Towards High-Fidelity CFD on the Cloud for the Automotive and Motorsport Sectors 2020-01-0665

This paper presents the results from an investigation into the performance of OpenFOAM v1806 on the Amazon Web Services (AWS) Elastic Compute Cloud (EC2) service for a realistic racing vehicle using a high-fidelity hybrid RANS-LES CFD approach. It is shown that AWS can provide the HPC environment to enable greater use of high-fidelity CFD methods by allowing higher core counts to reduce turn-around time. With the correct instance type - which potentially differs between meshing and solving - AWS was competitive against a high-performance Cray XC30 supercomputer, up to 1920 cores and meshes up to 280 million cells. However it is recognised that this Cray XC30 displayed superior scaling whilst containing older generation processors (Intel Ivybridge) compared to the AWS Instances (Intel Skylake). It was found that the influence of instance type was more pronounced during the meshing process within OpenFOAM (snappyHexMesh) where only the C5n.18xlarge instance was able to match the performance of the Cray XC30. This work establishes a set of best-practices and baseline configuration that will be used to look at larger models, larger core counts and to focus on other areas of the CFD workflow including post-processing.