Editorial

This newsletter is arriving later in the summer than usual, but for a very good reason: we wanted to provide you with information about the installation of PDC’s new flagship system, Dardel (see “First Phase of Dardel Installed” ). The original plan was for Dardel to be installed in spring this year, but that had to be postponed to late in the summer for reasons beyond PDC’s control.

The new system is one of several significant changes that have happened at PDC this year, which include me starting as the new director of PDC at the beginning of the year. Thanks to strong support from the other members of PDC’s leadership team – namely Gert Svensson, Henric Zazzi and Lars Malinowsky, as well as PDC’s interim director Patrick Norman – this change was organised as a smooth transition. Many thanks to everyone who helped with the transition and, in particular, to Patrick for making himself available as interim director. Patrick will remain closely linked to PDC through joint efforts on modern approaches to scientific software design (see “VeloxChem: Quantum Chemistry from Laptop to HPC” ).

A third change, which will be significant for PDC, is still ahead of us: based on the feedback from an international review panel, the Swedish Research Council (VR) decided to initiate a process to establish a new organization for high-performance computing (HPC) in Sweden. This will be a good opportunity to facilitate the necessary transition of HPC resource providers (like PDC) to becoming e-infrastructure service providers serving research in Sweden while being part of the EuroHPC ecosystem, which is becoming reality as you will see in this newsletter.

In this newsletter we focus, however, on today’s changes, with the arrival of Dardel being the most significant one. This will be a big step for PDC as the system comes with a lot of brand-new technology. This concerns, in particular, phase two of the system, which will be installed in autumn and facilitate a very significant boost in compute performance. While phase one of Dardel is based on nodes with very modern, but in some sense standard multi-core processors, phase two will be based on nodes comprising graphical processing units (GPUs) that act as compute accelerators. Using these GPUs requires applications to be ported using suitable programming models. The use of GPUs for scientific computing is not new, but now the choice of GPU architectures that are suitable for HPC is broadening and, therefore, portability needs to be considered. While Dardel – as well as the EuroHPC pre-exascale system LUMI, which is currently being installed in Finland – use GPUs from AMD, other systems will continue to use products from NVIDIA or plan to use upcoming GPUs from Intel. This newsletter provides you with a primer on GPU computing (see “Preparing for GPU Computing on the Dardel and LUMI Systems” ) as well as information on pilot projects for LUMI (see “LUMI: Successful Swedish Pilot Projects” ) and training opportunities provided by our colleagues at the EuroCC National Competence Centre Sweden (see “Training at ENCCS” ).

To support researchers to exploit such new hardware capabilities, PDC will continue to strongly engage with research communities, for instance, through Centres of Excellence (CoE) that are funded by the European Commission with the goal of getting researchers (and the code they use!) prepared for future exascale computing. The successful work in the BioExcel CoE, which is coordinated by PDC, continues (see “BioExcel News” ) and two new CoEs, namely TREX and PerMedCoE, allow PDC to work closely with researchers who use quantum Monte Carlo methods for research in quantum chemistry and condensed matter physics, as well as researchers working in the area of personalised medicine (see “TREX: Centre of Excellence for Quantum Monte Carlo Applications”  and “PerMedCoE: Exascale-Ready Cell-Level Simulations” ). PDC continues to engage with the brain research community, and we are proud of being part of the EBRAINS research infrastructure that recently succeeded in being listed on the road map of the prestigious European Strategy Forum on Research Infrastructures (ESFRI) (see “Performing Multi-Scale Simulations of the Brain Using MUSIC” ).

We are furthermore intensifying our efforts to make the resources available at PDC easier to use for the broader community of computational and data scientists. This involves simplifying authentication to the HPC services without compromising on security (see “PDC Portal for Improved Login” ) as well as improved support of small-scale use of, and interactive access to, the new Dardel system. Finally, in the next newsletter, we plan to report on a broadened service portfolio after adding a private cloud instance, which can be made more openly accessible than the HPC-based services.

With this overview, I hope to stimulate your continued interest in the PDC newsletter. Let me close by thanking the PDC team for their hard work on preparing for Dardel and preparing phase one of the system to become generally available in October. With the next newsletter, we will provide you with the first results obtained on this exciting new system!

Dirk Pleiter, Director PDC