Work

Evaluating computational performance metrics in climate modelling: insights from CMIP6

Year: 2024
Type: conference-paper
Abstract: The Coupled Model Intercomparison Project (CMIP) is one of the biggest global efforts aimed at understanding the Earth’s climate by undergoing a multi-model analysis. In it’s sixth phase, CMIP6 [1], a total of 190 different climate experiments where used to simulate approximately 40000 years which produced 40 petabytes of data in the process. This work presents the primary findings for the collection of a common set of performance metrics specially designed for climate modelling, the Computational Performance Model Intercomparison Project (CPMIP). These metrics were systematically collected from production runs conducted within CMIP6, predominantly from institutions affiliated with the IS-ENES3 [2] consortium. Through this comprehensive data collection effort, we contribute to the establishment of a robust database for future community reference, thereby setting a benchmark for evaluation and facilitating essential multi-model, multi-platform comparisons crucial for advancing climate modeling performance. Given the diverse array of applications, configurations, and hardware employed, further endeavors are imperative for standardization , for instance, by using climate and weather benchmarking codes such as the HPCW [3].
Authors: Palomas, Sergi, Acosta Cobos, Mario César, Utrera Iglesias, Gladys Miriam
Language: English
Open Access status: green
OpenAlex ID: W7113253183
FWCI:
Cites: 1
Cited by:
Related to: 10