From March 22–26, 2021, high-performance computing (HPC) experts and enthusiasts from academia and industry came together to exchange ideas and best practices, share presentations, and ultimately help shape the future of HPC in Europe.
While such activities are fairly standard for industry conferences, the EuroHPC Summit Week Conference took place as a fully virtual event one year after having needed to be cancelled due to the COVID-19 pandemic.
Despite the unfortunate circumstances, event organizers did everything possible to make the event as social and interactive as possible. Each session offered robust chat function that gave participants options to ask questions during live Q&A sessions and set up virtual meetings. The conference organizers also scheduled multiple “networking breaks” each day of the conference.
Staff members from the Gauss Centre for Supercomputing (GCS) and its associated centres participated in several sessions during the conference. Their presentations mainly focused on forward-looking initiatives in Europe in the realms of software development and optimization, Pan-European collaboration and information sharing, the development ofs made-in-Europe technologies, and the roadmap for Europe to reach the next major supercomputing milestone—delivering an exascale machine on the continent capable of at least one billion billion calculations per second.
On Monday, GCS Managing Director Dr. Claus-Axel Müller participated in a session focused on EuroHPC’s current infrastructure investments while also presenting a roadmap for the future. During the session, representatives from across Europe introduced the first round of machines supported by EuroHPC funding—the LUMI supercomputer in Finland, Leonardo supercomputer in Italy, MeluXina machine in Luxembourg, VEGA system in Slovenia, Karolina machine in Czechia, PetaSC system in Romania, and Deucalion machine in Portugal.
Afterwards, Müller, who also serves as the chair of EuroHPC infrastructure advisory group (INFRAG), presented the INFRAG’s recommendations for the next phase of EuroHPC funding and system acquisition. The proposals include increasing funding not only for HPC systems, but also for the associated data management and transfer architecture being deployed across Europe. Recommendations also include making additional funds available to help centres hire experts in porting and optimizing applications for increasingly complex computing architectures, and to implement a collective European cybersecurity concept, among other objectives.
Later in the day, High-Performance Computing Center Stuttgart’s (HLRS’s) General Manager Dr. Bastian Koller and Jülich Supercomputing Centre’s (JSC’s) Dr. Estela Suarez sat in on a session focused on developments in Europe’s research and innovation activities. Koller presented three projects funded in the last year that HLRS is coordinating—EuroCC and the associated CASTIEL projects as well as the FF4EuroHPC project.
EuroCC and CASTIEL are both focused on helping to better coordinate national HPC activities in Europe with Pan-European initiatives, including training activities. Additionally, the project aims to develop competency maps for each of the participating countries, creating a catalog of HPC experts and their associated expertise. FF4EuroHPC is a continuation of the Fortissimo project, which focused on increasing the amount of industrial users capable of using HPC resources to speed up their R&D activities. The goal of the new projects is ultimately to strengthen European economic competitiveness through faster time-to-market for products and other innovations.
Koller also gave a presentation during the nearly full-day industrial track on Tuesday, and sat on a panel about the European Centres of Excellence.
The SEA projects build upon the results of the DEEP project series. The initial DEEP project, started in 2011, focused on proposing innovative heterogeneous computing architectures—systems that use both traditional CPUs and some kind of accelerator, typically GPUs. Over the last decade, the project has been renewed and expanded multiple times, and the latest incarnation is actually 3 different projects—DEEP-SEA, IO-SEA, and RED-SEA.
DEEP-SEA focuses on scaling software on heterogeneous architectures, including those following the modular supercomputing architecture (MSA) concept being spearheaded by JSC and its current flagship supercomputer, JUWELS. IO-SEA addresses one of HPC’s biggest bottlenecks—input/output (I/O) challenges for users who need to efficiently move their datasets on and off HPC centres’ resources. RED-SEA focuses on one of exascale computing’s other major looming challenges—improving the interconnects that can quickly share information across nodes on HPC resources.
On Tuesday, during a session on “HPC, Big Data, IoT and AI future industry-driven collaborative strategic topics,” Dr. Jan Martinovič, Head of Advanced Data Analysis and Simulation Lab at IT4Innovations, Czechia’s leading HPC centre, presented the LEXIS EU project he coordinates. The data system work package of LEXIS is led by the Leibniz Supercomputing Centre (LRZ) and one of the primary collaborators on the project, LRZ’s Dr. Stephan Hachinger, contributed to the presentation. The project focuses on building an HPC workflow platform that integrates big data and cloud-based workflows. The team presented its current results during the session.
For more information about the EuroHPC Summit Week, please visit the website.