A finer mesh can increase communication, memory and output requirements together. Start with a representative model configuration, a useful forecast window and an agreed accuracy target.
System considerations
Measure domain decomposition, halo exchange and solver time before increasing node count. Keep initial-condition ensembles separate from the scaling of one simulation.
Evidence to bring
- Fixed mesh and model version
- Runtime split between compute, communication and I/O
- Forecast deadline and acceptable error
This is an illustrative research or industry context, not an available World Supercomputing service or a promised scientific outcome.
PRIMARY SOURCES · CHECKED 9 NOVEMBER 2026
NERSC: parallelism →Living documentation; checked 2026-11-09 · Strong scaling fixes total problem size; weak scaling fixes work per processor.
NERSC: job best practices →Living documentation; checked 2026-11-09 · Facility-specific operational guidance; use local site policies for actual jobs.