The need for massive computer power to process big data is rising exponentially in both industry and the research sector, and high performance computing (HPC) has been on Luxembourg’s agenda for the past decade. Well aware that Europe was lagging behind in this field, and that the necessary investments would exceed what a single country could do on its own, the Luxembourg government initiated discussions with the European Commission resulting in the launch of a pan-European HPC initiative.
In 2017, Luxembourg and six other countries signed an agreement to participate in the joint European and national effort to build and deploy a world-class computing and data infrastructure in Europe with participation of all EU member states and associated countries. The EuroHPC joint undertaking was set up to pool all financial means available for developing top-of-the-range exascale supercomputers for processing big data. Recognising the leadership shown by Luxembourg during the process, the EU decided to locate the EuroHPC in the Grand Duchy.
In early 2019, the EuroHPC launched a call for proposals to select eight sites all over Europe that would host petascale supercomputers. Luxembourg was obviously keen to host one of the HPCs. “We wanted to propose an innovative computer architecture adapted to the needs of the Luxembourg economy and our strategy for data-driven innovation,” says Jean-Marie Spaus, HPC Project Coordinator at the Ministry of the Economy.
Following discussions with the user community from the outset, the ministry brought together an innovative team that developed a modular supercomputing architecture. Lxembourg’s proposal was highly rated during the evaluation process, ranking second overall. “Our innovative approach and the business case behind was highly recognised,” notes Mr Spaus.
Open for business
Located in Bissen, literally next door to Luxembourg’s Mobility Innovation Campus, the HPC will be available to start-ups and industrial companies as well as universities, public research organisations and administrations. “This is quite unique as most HPCs are pure research centres,” says Roger Lampach, CEO of multi-tier and multi-tenant data centre facility LuxConnect that will host the HPC. “35% of the capacity is reserved for research, and we will commercialise the remaining 65% to companies inside and outside Luxembourg.” Expecting the HPC to be operational at the end of 2020, LuxConnect is starting the recruitment of hardware, software, application and user support specialists with relevant expertise in fields such as Industry 4.0, smart mobility, health technologies and space.
Our objective is to turn Luxembourg into a kind of HQ for data owned by companies.
Luxembourg has a solid basis for making the HPC a success. “We have the knowledge of data processing, high-speed connectivity, top-tier data centres and computing power and cloud computing elements on top,” says Mr Spaus. “Our objective is to turn Luxembourg into a kind of HQ for data owned by companies. We want to bring enormous amounts of data here, process it and create information with real economic value.” The HPC is making this possible.
Artificial intelligence laboratory
Big data processing is also the basis for developing artificial intelligence (AI), another Luxembourg key focus area. The demand for AI-related technologies is growing, notably in the fields of finance, healthcare and space resources research, and in early 2019, Luxembourg set up the first European partnership on artificial intelligence with global AI leader NVIDIA. It was put in place in the context of the governmental initiative Digital Luxembourg that aims to strengthen Luxembourg’s status as a digital pioneer. “This partnership will give our researchers access to state-of-the-art equipment and, in particular, allow them to benefit from NVIDIA’s experience to treat real-world problems by using AI,” said Stéphane Pallage, the rector of the University of Luxembourg, at the launch of the lab.
Senior research scientist Radu State heads the research group at the university’s Interdisciplinary Centre for Security, Reliability and Trust (SnT) that hosts the laboratory. “Together with representatives of Digital Luxembourg, we started to reflect on using graphics processing units (GPUs) produced by NVIDIA for AI and machine learning,” he remembers. Digital Luxembourg organised an initial workshop with NVIDIA that was met with much interest and led to the idea of a large-scale collaboration.
Focus on key sectors
NVIDIA, the university’s HPC centre, the SnT, the Luxembourg Centre for Systems Biomedicine (LCSB) and the Luxembourg Institute of Science and Technology (LIST) have initially committed to designate a team of six people to work on joint projects. The SnT hosts an engineer from NVIDIA who plays a coordinating role. “One of the first things I think we can do together is engage in machine learning and deep learning projects funded by the EU’s Horizon 2020 programme for research and innovation,” says Dr State. “I also really look forward to extending the collaboration to private companies.”
Working with the private sector is a priority for the SnT, and almost all of the 34 researchers in Dr State’s group are funded by industry projects. “We work in a very applied way with local businesses,” he explains. “The NVIDIA technology is very interesting for Luxembourg’s key sectors such as autonomous driving or space technologies. There is, for example, a company here in Luxembourg that is building a rover for exploring the surface of the moon. NVIDIA’s GPUs have the very high image processing capacity needed to determine where to land and to operate the robot on the moon.”
I really look forward to extending the collaboration to private companies.
The university now offers training sessions, open to anyone, which give participants hands-on experience of working with NVIDIA technology. So what results does Dr State hope to see coming out of this AI collaboration? “I’m very business oriented. For me, the initiative is successful if in the next 5-10 years we manage to develop 20-30 companies specialised in machine learning that can show positive economic results.”