"Forecasting weather is like a 24-hour marathon - it never stops and it requires technology to keep pace", stated Prof. Dr. Thomas Schulthess, director of CSCS. "And the complexity is growing in weather and climate forecasting and modeling as we put advanced workloads to the test. Our partnership with Cray continues to provide us the compute resources necessary so customers like MeteoSwiss can tackle the biggest challenges in meteorological research. This new system will make it possible for them to more accurately predict the weather."
MeteoSwiss continually invests in innovation and development for the benefit of the economy and public safety. With CSCS' purchase of its first two Cray CS-Storm systems in 2015, MeteoSwiss became the first major national weather service to use a GPU-accelerated supercomputer to run production numerical weather models. And with its latest CS-Storm - designed for organisations that require the performance and efficiency of today's latest GPU accelerators - MeteoSwiss will continue with the GPU approach.
"Complete, extremely detailed forecasts is our vision for the future", stated Philippe Steiner, Head of Numerical Prediction at MeteoSwiss. "Switzerland's unique orography adds a level of complexity to forecasting that requires us to develop advanced, data-intensive modeling methods and techniques. With this latest technology from Cray, MeteoSwiss will be able to run increasingly complex, high-resolution models for higher-accuracy forecasts and nowcasts, as well as to develop additional weather forecast products."
MeteoSwiss found success with its existing Cray supercomputers and selected this new CS-Storm to provide the additional computational power required to process increasing volumes of weather observations and produce higher fidelity forecasts. The CS-Storm system was also selected for its ability to run numerical weather forecasts within a reduced energy footprint - as compared to competing solutions, and for the reliability the platform provides MeteoSwiss when running critical workloads.
"The new Cray CS-Storm at CSCS equips MeteoSwiss with the computational performance required to power next-generation weather forecasts. Cray is proud to play a part in helping support the amazing work MeteoSwiss does to keep Swiss citizens safe, informed and protected through accurate and timely weather forecasts", stated Ilene Carpenter, Earth Sciences Segment Director at Cray.
CSCS' new CS-Storm is configured with 18 compute nodes, each with 8 NVIDIA V100 GPUs and 2 Intel Xeon Gold 6134 CPUs, and includes two Cray ClusterStor L300 storage systems. The CS-Storm was accepted in April 2019 and will become fully operational in 2020.