Stories

Dr Matthew Evans, BEWARE Research Fellow at UCLouvain & Matgenix, explains how they are numerically testing thousands of materials to determine their properties, based on target properties, without having to synthesize them. Combined with machine learning, these simulations can help to greatly accelerate the search for the best materials to achieve specific goals.

In his PhD, Mats Denayer (General Chemistry Research Group of Vrije Universiteit Brussel) developed a new computer-based protocol to predict whether polymers will dissolve in certain liquids (solvents). Proven valid, the program can be used for polymer recycling. Mats' research fits within the context of transitioning to a circular economy through the development of (polymer) recycling processes.

Finding means to fight climate change is a major target of today's scientific research. To this aim, it is essential to limit greenhouse gas emissions, for instance by replacing fossil fuels with alternatives such as biogas or hydrogen, or by combining power and heat generation. Maintaining complete and stable combustion in such unconventional conditions is not straightforward. In this user story, Alessio Pappa explains how his research can help design better combustors for sustainable energy production.

Whether it's designing more efficient wind turbines ⚡, faster cars ?️, or more fuel-efficient aircraft ✈️, computational fluid dynamics (CFD) is an essential tool for researchers. To obtain accurate results, it is sometimes necessary to describe complex structures with great precision, which can require very substantial computing resources; hence, the use of supercomputers. However, implementing CFD methods efficiently on HPC infrastructures is not straightforward, as a lot of communication is inherently required between the different processing units of the supercomputer. In this user story, Pierre Balty, from Ph. Chatelain’s lab, explains where this limitation comes from and how they work on efficient parallelisation methods, pushing back the limits of what is possible in CFD.

Upscaling your resources from Tier-2 to Tier-1 or Tier-0 can be a huge advantage and accelerator for your research ??! Tim Lebailly, PhD student at PSI, aims to improve current machine algorithms for AI. He went from Tier-2 to Tier-0 infrastructure for his research. According to Tim, moving up from Tier-2 to Hortense (Tier-1) was very straightforward: “It’s very similar hardware. Only the scheduler is different, but that’s a detail.”  Tim also had to conduct numerous experiments in parallel. If he attempted this while using Tier-1, he used the entire GPU partition exclusively, causing some jobs to remain in the queue for a long time. The next logical step for Tim was to apply for compute time on LUMI, the fastest Tier-0 supercomputer in Europe. Tim: “Thanks to the scale of LUMI, I only utilise a small fraction of the supercomputer's capacity, enabling me to schedule all my jobs simultaneously. In that regard, the user experience is really nice.” Want to know more? Read all about Tim's upscaling journey by clicking the title above or via www.enccb.be/usupscalingtimlebailly.

Matchmaking is one of the services offered by VSC: finding the right partner to help solve research questions of users. In this case, VSC introduced Puxano to the Data Science Institute of UHasselt (DSI). DSI brings together over 150 researchers working in the broad field of data science across faculties and research groups at Hasselt University.

In this interview, Damien François, system admninistrator at the CISM of UCLouvain and trainer for the CÉCI, explain what Slurm is and how it can help researchers and system administrators better manage their HPC resources.

Besides computing infrastructure, VSC also offers a wide range of services to its users. One of these services is helping users to improve their software so it can run more efficiently on the VSC infrastructure. In the case of Mpacts, VSC recommended sorting the particles so that particles close in (simulated) space are also close together in computer memory. This programming technique made the software more efficient and, thus, faster.