The relentless pursuit of new materials is a core pillar of scientific and technological advancement, spanning applications from energy storage to electronics. In this quest, the Swiss National Center of Competence in Research (NCCR) MARVEL’s successful execution of a groundbreaking computational experiment utilizing the Alps supercomputer underscores the rapid progression in high-performance computing (HPC) and materials science. Through this endeavor, researchers not only showcased the Alps supercomputer’s capabilities but also highlighted the maturity of Swiss-made software tools designed to streamline material simulations.
Inaugurated on September 14, 2024, the Alps supercomputer has positioned itself as a formidable asset within the global sphere of computational power. Operating predominantly from its data center in Lugano, this state-of-the-art facility is managed by the Swiss National Supercomputing Center (CSCS) and represents Switzerland’s leading edge in supercomputing technology. During its acceptance phase, data from selected research groups were funneled through its robust infrastructure, offering a spotlight on the potentials of Swiss engineering in computational materials science.
The computational capabilities of the Alps supercomputer rely heavily on its advanced architecture, consisting of thousands of NVIDIA Grace Hopper nodes, integrated with an extensive array of GPUs and CPU cores. Researchers from NCCR MARVEL were given an early opportunity to leverage this power. Under the guidance of prominent experts such as Giovanni Pizzi, the team engaged in what they dubbed a “hero run” – an intensive computational exercise maximizing the resources of the supercomputer for their own research goals.
On July 17 and 18, a group of dedicated researchers, including Marnik Bercx, Michail Minotakis, and Timo Reents, ran a comprehensive high-throughput computational experiment. Their objective was to merge the impressive capacities of Alps with the AiiDA open-source framework, which automates the laborious calculations often intrinsic to materials science. This collaboration aimed to simulate a wide range of material properties by calculating data for thousands of unique structures in parallel.
As the research team carefully scripted their workflow, they demonstrated the ability of AiiDA to orchestrate these high-throughput calculations, effectively filling the Alps supercomputer with a continuous stream of jobs. The logistics of managing such an operation were considerable, as the AiiDA framework monitored all submissions and ensured that every completed job’s output could be processed and archived for further analysis.
Employing an enhanced version of the Quantum ESPRESSO computational code, the team took full advantage of the Alps’ GPU architecture. The researchers strategically selected structures of medium size—around 40 atoms—as smaller structures would underutilize the supercomputer’s processing capabilities. This meticulous approach underscored their intent to maximize the computational prowess of Alps.
The marathon run culminated in astonishingly high efficiency, achieving a remarkable 99.96% utilization rate of the machine’s capacity. Throughout nearly 16 hours, the team completed approximately 100,000 individual calculations related to the properties of over 20,000 crystal structures stored in AiiDA’s database. This impressive achievement stands as a testament to the efficiency of the software and its capability to drive the supercomputer to its limits.
Following the successful conclusion of the hero run, the wealth of data generated will soon be made publicly available through platforms like the Materials Cloud, a vital resource for advancing materials science research. The implications of these simulations are profound, as they not only enrich existing databases but also foster collaborative exploration of novel materials. The open nature of this data aligns with contemporary scientific ethics, emphasizing transparency and accessibility in research.
Moreover, the results will have notable impacts across various fields, particularly in identifying new materials with potential applications in next-generation energy storage solutions, contributing significantly to sustainable technologies. The collaborative effort exemplified by this hero run demonstrates not only the power of supercomputing but also the promise of innovative computational tools like AiiDA.
As researchers connected through NCCR MARVEL embrace and leverage these high-performance systems, they reaffirm their commitment to uncovering new materials and unlocking unprecedented scientific potential. The Alps supercomputer and the advancements within platforms like AiiDA illustrate a bright future for computational materials science, one where rapid computation and innovative algorithms pave the way for groundbreaking discoveries. In an era defined by technological advancement, the efforts by teams like Pizzi’s at PSI continue to propel the scientific community toward new horizons.
Leave a Reply