Powering Advanced Climate Simulations with HPC on AWS for the University of Liège

The university’s research team is developing an advanced climate model called the Modèle Atmosphérique Régional (MAR) to predict climate change impacts with up to 5 km accuracy. To run a future climate projection for Belgium spanning from 1980 to 2100 at a 5 km resolution, the university needs to use 240 CPUs for 4–5 days. With a few supercomputers running on premises, researchers faced delays in accessing HPC capacity to run these critical climate modeling workloads, impacting their ability to meet IPCC deadlines quickly.

To overcome these challenges, the university collaborated alongside Ankercloud, an AWS Advanced Partner, to deploy its HPC environment on AWS. “In a matter of weeks, our technical team deployed the whole environment from scratch and started benchmarks and testing,” says Riccardo De Gasperin, senior cloud engagement executive at Ankercloud.

Ankercloud worked with the University of Liège to set up the new environment using AWS CloudFormation, which speeds up cloud provisioning with infrastructure as code. “With support from Ankercloud, we’re able to run our model on AWS, and it’s working well,” says Xavier Fettweis, professor of climatology at the University of Liège.

By running MAR simulations on Amazon Elastic Compute Cloud (Amazon EC2) Linux-based instances, the University of Liège has enhanced the accuracy of its weather forecasts and the level of detail in its climate projections. Running workloads in the cloud can also lower its carbon footprint by up to 88 percent compared to on premises.

The lab manages its data generated from weather simulations using Amazon Simple Storage Service (Amazon S3), an object storage service built to retrieve any amount of data from anywhere. For observability, it uses Amazon CloudWatch to monitor resources and applications on AWS, on premises, and on other clouds. This cloud-based approach has helped the university right-size and scale resources on demand, reducing operational costs by 50 percent.

Based on this proof of concept, the Laboratory of Climatology recommends that the University of Liège uses AWS across all its research projects to free its teams from the constraints of on-premises resources.

The research team can now meet important deadlines to contribute to IPCC assessment reports and accelerate its efforts to understand and predict the impacts of climate change across Belgium, ultimately supporting policymakers, urban planners, and the public in addressing this critical challenge. “AWS provides a robust set of HPC resources,” says Fettweis. “Working in the cloud is the solution of the future, and it helps us decrease our CO2footprint by performing climate or weather simulations.”