SIParCS 2018- Dixit Patel
Performance Analysis and Optimization of Weather Research Forecasting (WRF) Model
The Weather Research and Forecasting (WRF) model is an open source mesoscale numerical weather prediction (NWP) system extensively used in atmospheric research, operational forecasting and educational settings. In this study, WRF's performance is compared against various compilers and MPI libraries along-with several optimization options that are available. Hybrid parallelization of WRF was also tested by varying number of MPI tasks and OpenMP threads with different processor binding strategies. The simulations use datasets obtained from NCEP with different grid sizes. For e.g. Hurricane Katrina data was used at 1km and 3km resolution. The benchmarks were performed on NCAR's current HPC platform, Cheyenne, which uses Intel Xeon E5-2697V4 (Broadwell) processors. WRF is a highly scalable model and the analysis shows that all cases scale similarly. The scaling analysis provided would also benefit users in selecting the best settings currently available to run WRF on Cheyenne and understand how the model's performance would vary at different node counts. This helps to estimate the number of core-hours that would be required for different problem sizes. The results are also compared against Yellowstone supercomputer previously used at NCAR.
Mentors: Davide Del Vento, Brian Vanderwende, Negin Sobhani