HPC
NAISS
NAISS, the National Academic Infrastructure for Supercomputing in Sweden, is, roughly speaking, responsible to allocate computing times on Swedish HPC resources to scientists working at Swedish academic institutions. Except maybe for Dardel, the clusters might be a bit smaller than the ones accessible via EuroHPC, but you don’t have to compete with researchers from all over Europe. It’s the successor of SNIC (you might still find this name in some of their URLs).
https://www.naiss.se/resources/
Among the NAISS resources, Dardel is located at KTH, more specifically at the PDC Center for High Performance Computing (https://www.pdc.kth.se/).
Other useful system available via NAISS include Berzelius and Tetralith run by NSC (https://www.nsc.liu.se/systems/).
NAISS also offers a cloud solution potentially interesting for large sets of small simulations, analysis jobs or AI/ML projects (https://cloud.snic.se/).
EuroHPC
EuroHPC can grant you access to the most powerful supercomputing clusters currently available in Europe. For all these clusters, a pre-defined fraction of the computing times is distributed by EuroHPC, whereas the remaining times are assigned by a national organization. The national organizations usually have less competitive calls, and they usually require only one of the project partners, who formally has to file the application, to be working at a national academic institution. If you work in a multi-national project, please consider going via the national organization before filing a EuroHPC application.
Nevertheless, several group members have successfully won Regular Access Call proposals at EuroHPC. For more information on their call system, have a look at:
https://eurohpc-ju.europa.eu/access-our-supercomputers/eurohpc-access-calls_en
Here is a list of some of the clusters available via EuroHPC calls:
LUMI (https://www.lumi-supercomputer.eu/) is the largest CPU/GPU cluster in the North and relies on AMD hardware (i.e. you need to use the GROMACS hipSYCL port if you want to run on GPUs there). It comes with the advantage that computing times are assigned via NAISS (SNIC), too.
Leonardo (https://leonardo-supercomputer.cineca.eu/) at CINECA is the largest NVIDIA-based GPU cluster in Europe (i.e. you can use the more established GROMACS CUDA port there).
Mare Nostrum (https://www.bsc.es/) run by the Barcelona supercomputing centre is a very large CPU/GPU cluster and relies on NVIDIA GPUs.
Moreover, people in this group have made good experiences with these smaller clusters:
Karolina (CPU/GPU (NVIDIA)): https://www.it4i.cz/en/infrastructure/karolina
MeluXina (CPU/GPU (NVIDIA)): https://docs.lxp.lu/
Discoverer (CPU): https://sofiatech.bg/en/petascale-supercomputer/
CSCS
Outside Sweden and EuroHPC, group members (especially from the Delemotte group) have successfully secured and made good experiences with allocations at the Swiss National Supercomputing Centre (CSCS):
Storage
Published scientific data must be stored for decades and should ideally be easily accessible to everybody. An easy solution is to transfer your data to databases where you can publish your datasets (with a DOI of their own). Such databases are, for example:
- OSF (Open Science Framework): https://osf.io/
- Zenodo: https://zenodo.org/