/
#12 Ensemble simulation workflow for UQ
#12 Ensemble simulation workflow for UQ
Abstract
Comprehensive parametric sensitivity analysis typically requires a large number of simulations, especially when the “short ensemble” strategy is used instead of traditional long-term simulations. We developed a hybrid method that makes use of the CESM/ACME model’s multi-instance capability in combination with the batch-script-based task bundling to create machine capability jobs at LCFs. We also used cross-system job submission to optimize the use of computational resources. First experiences at OLCF show that it was possible to finish thousands of 3-day simulations within 24 hours of wall clock time counting from job submission to completion. The poster describes the workflow and suggests possible future improvements of CIME.
, multiple selections available,
Related content
Parallel Ensemble Simulations for ACME Performance and Verification
Parallel Ensemble Simulations for ACME Performance and Verification
More like this
#W01 ACME Ensemble Results Explanation and Reproducibility with ProvEn
#W01 ACME Ensemble Results Explanation and Reproducibility with ProvEn
More like this
Short simulations for efficient model evaluation and tuning: looking for robust sensitivities
Short simulations for efficient model evaluation and tuning: looking for robust sensitivities
More like this
Short simulations for efficient model evaluation, tuning and calibration, Part I: Strategy and Framework
Short simulations for efficient model evaluation, tuning and calibration, Part I: Strategy and Framework
More like this
Bundling runs to improve throughput on Mira
Bundling runs to improve throughput on Mira
More like this
#A03 Parametric sensitivity and optimization in ACME-V1 atmosphere
#A03 Parametric sensitivity and optimization in ACME-V1 atmosphere
More like this