Opname van podcast met Valeriu Podreanu (l) en Sanne Koenen (r)
Podcast

What you need to know about... supercomputing

Supercomputing is indispensable in much research in the Netherlands. What is supercomputing and what does the national supercomputer Snellius bring to our scientists? And why was it very good news that Dutch scientists recently also have easier access to the European supercomputer Lumi?

 

Podcast outline

1. Introduction (0:00 - 0:30)

  • Introduction of the SURFshort-podcast by Sanne Koenen.
  • Guest: Valerio Condreanu, Teamlead High Performance Computing and Visualization at SURF.
  • Topic: Supercomputing.

2. What is supercomputing? (0:31 - 2:05)

  • Supercomputing is described as a collection of powerful computers working together over a strong network and storage.
  • It helps reduce the time needed for calculations that would otherwise take years on a normal computer.

3. Use of supercomputers in research (2:06 - 3:45)

  • Historically, supercomputing has been widely used in computational chemistry and materials science in the Netherlands.
  • In recent years, artificial intelligence (AI) has also been increasingly used, with AI models accelerating traditional simulations.

4. The Dutch supercomputer Snellius and the European system Lumi (3:46 - 6:15)

  • Snellius is the national supercomputer of the Netherlands.
  • LUMI is a larger supercomputer in Europe, to which the Netherlands has access through government investment.
  • LUMI provides access to a larger number of GPUs than Snellius, which is important for large-scale research.

5. The importance of scale and speed (6:16 - 7:30)

  • On LUMI, researchers can train complex AI models thanks to the large number of GPUs.
  • Access to LUMI can be arranged through NWO and SURF in a few weeks, instead of waiting six months through the European application process.

6. Competition with Big Tech  (7:31 - 9:00)

  •  While LUMI is powerful, the systems of big tech companies like Microsoft or Meta remain much bigger.
  • It is not necessary to compete with big tech, but academia should focus on using existing resources more efficiently.

7. Accessibility and ease of use (9:01 - 10:30)

  • Supercomputers are accessible to researchers, but require some experience, such as knowledge of the command line.
  • Work on accessibility is underway by developing web interfaces that allow researchers to access them via their phones.

8. Scalability between university clusters and national systems (10:31 - 11:45)

  •  Researchers working on university clusters should be able to easily scale up to national systems such as Snellius.
  • Interoperability between systems is an important goal here. 

9. Investments in supercomputing in the Netherlands (11:46 - 12:30)

  • The Netherlands invests relatively little in supercomputing compared to other European countries, e.g. Finland.
  • SURF has developed a strategy to work on large-scale scientific research and the required budget.

10. Sustainability and energy consumption (12:31 - 14:00)

  • It is important to make supercomputers more efficient and consume less energy by using specialised computing architectures.
  • Use of GPUs can reduce energy consumption by performing certain calculations faster than CPUs.

11. Tips for further exploration (14:01 - 15:00)

Supercomputing: Powerful Computing for Science and Innovation

Supercomputing plays a crucial role in scientific research and technological development. In the latest episode of our SURFshort podcast, Valerio Condreanu, team lead for High Performance Computing and Visualisation at SURF, discusses the possibilities and challenges of supercomputers. What makes supercomputing so important for the Netherlands, and how can researchers make the best use of it? This summary highlights the key topics covered in the podcast.

What is supercomputing, actually?

"Supercomputing is the use of multiple powerful computers that work together via a high-speed network and shared storage," Valerio explains. These systems collectively perform complex calculations that are too large for a single computer. Think of simulations for climate research or training AI models. Supercomputers such as Snellius and LUMI enable the execution of these computationally intensive tasks in a fraction of the time.

From Snellius to LUMI: Access to European Computing Power

The Netherlands has its national supercomputer, Snellius, but researchers also have access to the European supercomputer LUMI. "LUMI is one of the most powerful systems in the world and offers a large number of GPUs," says Valerio. According to him, this access is essential for conducting large-scale AI experiments. "Thanks to investments from the Dutch government, researchers can use LUMI within weeks, rather than waiting months through the standard European application process." Read more about access to supercomputers on the Snellius and LUMI pages.

The Importance of Speed: Why Fast Access to Computing Power is Essential

Valerio emphasises that fast access to computing power can make a significant difference. "Research that would normally take months can be completed within weeks using Snellius or LUMI." This speed is crucial for scientific breakthroughs, especially in competitive research fields such as AI and genetics.

Can We Compete with Big Tech?

Although LUMI is powerful, the systems of major technology companies like Microsoft and Meta are still much larger. "We do not necessarily have to compete with big tech," says Valerio. "What matters is that we make optimal use of our resources. We need to develop smart algorithms that are specifically tailored to our systems." According to Valerio, making more efficient use of existing infrastructure is just as important as having the largest systems.

The Accessibility of Supercomputing: Improving Ease of Use

"Supercomputers are not necessarily difficult to use, but they do require some technical knowledge, such as working with the command line," says Valerio. To lower barriers, SURF is developing web interfaces. "We want researchers to have access to computing power even from their phones." By improving accessibility, supercomputing becomes more widely usable, even for researchers without extensive technical backgrounds.

Investments and Challenges: Is the Netherlands Falling Behind?

According to Valerio, the Netherlands invests relatively little in supercomputing compared to other European countries, such as Finland. "This affects the capacity we can offer," he notes. That is why SURF is working on a strategy to make better use of available resources and enable large-scale scientific projects. "We are focusing on the efficient use of the systems we already have."

Sustainability in Supercomputing: Tackling Energy Consumption

Reducing energy consumption is a major challenge in supercomputing. "GPUs can be much more efficient than traditional CPUs for specific tasks, which helps reduce energy consumption," Valerio explains. He also mentions improved cooling systems as a way to enhance the sustainability of supercomputers. "We must continue to innovate in energy efficiency."

Valerio’s Tips: Learn More About Supercomputing

For those who want to learn more about the ambitions and future of supercomputing in Europe and the Netherlands, Valerio recommends reading reports from PRACE and the Dutch Research Council (NWO).

Every month, we bring you up to speed on technological developments in education and research with a new SURFshort in just 15 minutes.

Listen to more podcasts

Valeriu Codreanu

Valeriu Codreanu