Without supercomputing, research today would not be where it is — and companies would also struggle to deploy large AI applications without high-performance computing (HPC). Currently, Austria's supercomputer VSC team is working, among other projects, on the new high-performance computer MUSICA, which will begin operation in 2025. Andreas Rauber, head of the VSC Research Centre, discusses in an interview the costs and benefits of the new supercomputer, as well as the challenges and advantages that high-performance computing brings in general.
Interview by Bettina Benesch
There isn’t a single experience — it’s been so many fascinating things. And something that excites me for a while stops being exciting once the issue is resolved. What I love about the VSC is that the user groups come from so many disciplines, most of which I have no knowledge of. I find it utterly fascinating to see what they compute and the kinds of questions they seek to answer.
One group attempted to predict the properties of an alloy and identify where fracture lines might occur — ranging from entire material blocks down to the molecular structure.
There are also numerous simulations focusing on climate research. For example, there are efforts to deduce soil moisture levels from satellite images.
This is a complex topic, and AI introduces entirely new demands that we need to integrate into HPC operations. The complexity arises from the much larger number of users who want to run a diverse array of models, software, and code on HPC systems. We’ll be dealing with vastly larger datasets, especially when people start working with massive image data collections. These challenges require new software architectures and bring entirely different security requirements.
In classical HPC, jobs are submitted into a pipeline and processed sequentially. Once the results are ready, the researcher is notified. But if someone needs real-time usage — for instance, for visualisations they want to manipulate interactively — they require exclusive access to a specific amount of computing resources.
This involves a different type of resource management. For example, if a researcher reserves an hour but finishes in 30 minutes, they should be able to end their session so the resources can be made available to others. Conversely, if they need an extra 15 minutes, it should be possible to extend their session. These are all relatively simple problems individually, but in a large, complex system, they accumulate and need to be solved in a highly automated way — especially since we aim to ensure high system utilisation for maximum energy efficiency.
One frequently discussed topic is HPC’s energy consumption. The question for me is: what does HPC enable us to do to improve energy efficiency in other areas? Yes, it costs money, but it’s money well spent to solve other critical issues. For example, simulations can save an enormous amount of effort in actual experiments. Before going into a lab and working with chemicals, or before starting large machines to build turbines, simulations can save significant time and ultimately energy.
Another major challenge for us is making the data centre and the computing process itself efficient. We strive for high utilisation rates while cooling efficiently with open-loop cooling systems and reusing waste heat—for example, for greenhouses. The trend is definitely towards more efficient structures and larger collaborations with secondary users.
“
HPC costs money, but it's money well spent to solve problems. For example, simulations can save an enormous amount of effort in actual experiments.
„
Yes. At the moment, it’s also raising excitement levels, like any new project. It’s heart-warming in that sense too. Essentially, HPC works well when it’s truly high-performance computing: as much computing power as possible, used collectively by as many users as possible. There’s a very strong trend towards jointly operating such infrastructures, ensuring quality, having the necessary personnel, and collectively providing a large system for everyone to use. This approach works very well in Austria.
Both will remain in operation for the time being. There’s a six-year cycle: procurement for the VSC-4 replacement will begin in 2025/2026, followed by the VSC-5 replacement in 2027/2028. Around 2030, we’ll start planning MUSICA’s replacement.
A procurement process usually takes a year because it’s a bit more complex than adding a laptop to your online shopping cart. For MUSICA, for example, spatial adjustments need to be made, the cooling system upgraded, new power lines installed, and much more.
“
There’s a very strong trend towards jointly operating HPC infrastructures. This approach works very well in Austria.
„
The overall system comprises MUSICA’s compute component and a high-performance storage component. There are also infrastructure costs for spatial modifications. In total, we’re looking at an investment of €36 million for a supercomputer distributed across three sites in Austria.
Not particularly large — it consists of a few racks. The VSC-5 is surprisingly compact as well. But this can be misleading: while the computer system itself takes up only a few cubic metres, that’s just the computing core. There’s also the entire cooling infrastructure, including water pipes, pumps, heat exchangers, circulation pumps, and roof pipework to dissipate the heat. This infrastructure is much larger. Essentially, it’s an entire building — there’s a floor beneath the computer dedicated to cooling systems with turbo cores to generate additional cooling.
Short bio
Andreas Rauber is the head of the VSC Research Centre. He studied computer science at TU Wien, then worked as a postdoc in Italy and later in France. In the early 2000s, Andreas returned to Vienna, where he built a research team at TU Wien. His projects focus on data analysis, machine learning, information retrieval, and research and data infrastructures.
About the key concepts
Believe it or not, High-Performance Computing (HPC) is actually a relatively old concept: the word "supercomputing" was first used in 1929, and the first mainframe computers appeared in the 1950s. However, they had far less capacity than today's mobile phones. The technology really took off in the 1970s.
HPC systems are used whenever the personal computer's memory is too small, larger simulations are required that cannot be run on the personal system, or when what was previously calculated locally now needs to be calculated much more frequently.
The performance of supercomputers is measured in FLOPS (Floating Point Operations Per Second). In 1997, a supercomputer achieved 1.06 TeraFLOPS (1 TeraFLOPS = 10^12 FLOPS) for the first time; Austria's currently most powerful supercomputer, the VSC-5, reaches 2.31 PetaFLOPS or 2,310 TeraFLOPS (1 PetaFLOPS = 10^15 FLOPS). The era of exascale computers began in 2022, with performance measured in ExaFLOPS (1 ExaFLOPS = 10^18 FLOPS). An ExaFLOPS equals one quintillion floating-point operations per second.
As of June 2024, there were only two exascale systems in the TOP500 list of the world's best supercomputers: Frontier at Oak Ridge National Laboratory and Aurora at Argonne National Laboratory, both in the USA. In Europe, there are currently three pre-exascale computers, which are precursors to exascale systems. Two European exascale systems will be operational shortly.
VSC (Vienna Scientific Cluster) is Austria's supercomputer, co-financed by several Austrian universities. The computers are located at the TU Wien university in Vienna. From 2025, the newest supercomputer, MUSICA (Multi-Site Computer Austria), will be in use at locations in Vienna, Linz, and Innsbruck.
Researchers from the participating universities can use the VSC for their simulations, and under the EuroCC programme, companies also have easy and free access to computing time on Austria's supercomputer. Additionally, the VSC team is an important source of know-how: in numerous workshops, future HPC users, regardless of their level, learn everything about supercomputing, AI and big data.