High-performance computing (HPC) is vital for the advancement of scientific and engineering research. But HPC is expensive, and every part of the world handles management and access to this valuable resource in its own way.
In Europe, PRACE (Partnership for Advanced Computing in Europe) enables high-impact scientific discovery by offering world class computing for academia and industry through seven HPC systems operated by its members.
We spoke with Núria López, professor in chemistry at ICIQ and chair of the PRACE Scientific Steering Committee (SSC), to find out what is on the horizon for PRACE and European computing in the new decade.
What is the job of the PRACE Scientific Steering Committee?
We look ahead and try to identify what scientific trends are becoming important and which kinds of research PRACE should support. We also oversee the access committee that allocates hours on the PRACE supercomputers.
So, on one side, we propose new actions, such as training, and on the other side we determine that the scientific practices are carried out in the best way for everyone involved.
What are some of the biggest issues you’re facing right now?
With the Scientific Case for Computing in Europe 2018-2026 we are looking ahead to exascale and pre-exascale machines. With faster computers, we know that some of our existing algorithms will work, or nearly work. For those applications, it will be a matter of fine-tuning to obtain higher resolutions or more precise results.
But for other applications, we aren’t completely ready. We know that big data is coming, and we have to figure out how we’re going to deal with it—we need development there.
Take biology, for example. Just a few years back, the routine analysis for a DNA sample took a very long time, but now you can have it in 1-2 days or less. This increased speed actually creates increased computational demand.
This demand will impact the type of supercomputers that will be needed. Having machines with different types of architectures will help us provide services to many communities—including ones that may not even be using HPC now but will be in the future.
There are some multi-disciplinary areas that I’m worried about because there could be a big explosion in demand that we don’t see coming. We don’t know what they will be needing. But if we get the machines and we get the right tools, this is an area where the success will be very, very big.
How do you deal with anticipating the needs of a field that may not yet exist?
One thing is that, beneath all the layers, there is math. We have some common problems that all rely on a mathematical way of doing things. I think this is one of the beautiful things about using supercomputers.
What we do need to do is improve the fundamental math in order to push our algorithms farther. It’s a mistake to constrain yourself only to your own field. High-performance computing scientists are very versatile in this way because they see that there is math inside their problem. So if someone has similar math, they say ‘Let’s try it’. The problems may look different on the outside, but at the end of the day we are facing the same problems.
Your tenure as chair ends in 2021. What do you hope to accomplish by then?
2019 has been a transitional year, with plenty of challenges regarding the integration of the EuroHPC Joint Undertaking with the services PRACE has been providing in recent years. I will be happy by the end of my term as chair if the situation becomes a bit more clear and we can bring the spirit of science into the core of the EuroHPC joint undertaking.
Can you explain a bit more about EuroHPC joint undertaking?
The European Commission has recognized that there is a real need for super-computation in Europe, and a need to support this though strong investment over the long term. This is very encouraging.
But to get the best from the procurements of these new machines that will be coming, we have to adapt them to all of the problems that we know we want to solve. The software we have now will have to be adapted. So we need part of the money to be going not only to the physical machines, but also to software and training.
How have diversity efforts been going under your watch?
In the last year, we have raised the number of women who participate in our training course by 4-5 percent. I think this is extremely positive. I’m also very happy that we keep increasing the number of women applying for computational resources as principal investigators. On the other hand, I would like to see the chief director of one of the computing centers be a woman, but that is going to take a little more time.
And on our scientific steering committee, we are 50-50. And we are 50-50 in the access committee, as well. So that gives us an opportunity to promote diversity. All these initiatives that come from the bottom-up, they take a while. But I believe that we will only make more resilient teams if we address diversity—in all areas.
PRACEdays20 /EuroHPC Summit Week 2020 will take place in Porto, Portugal from 23 to 27 March 2020.
- The race to exascale
- What's next for supercomputing in Europe
- Another step on the stairs to exascale