Evaluating Performance and Scalability of Candidate Dynamical Cores for
the Next Generation Global Prediction System at NOAA, John Michalakes,
NOAA
To explicitly simulate moist convection on the global scale and take
advantage of advances in high performance computing, NOAA evaluated five
candidate non-hydrostatic dynamical cores to find a new dynamical core for
the Next-Generation Global Prediction System (NGGPS), which will meet
operational global numerical weather prediction requirements at the National
Weather Service through 2025-30. Based on concurrent evaluations of
computational and scientific performance of the dynamical cores, NOAA
management recently decided to proceed with two of the candidate models, FV3
from NOAA/GFDL and MPAS from NCAR, into the next phase of testing that will
culminate next year in final selection of the NGGPS dynamical core. This
presentation describes methodology, cases, model configurations, and results
of performance and scalability benchmarks conducted during two sessions on
Edison, a 130-thousand processor core supercomputer at the U.S. Department
of Energy's National Energy Research Scientific Computing Center (NERSC),
during March and April, 2015.