HLRS managing director Bastian Koller explains three new EuroHPC projects to support HPC-related collaboration at national and European level


17 Aug 2020 Almere - Bastian Koller from the High-Performance Computing Center in Stuttgart (HLRS), Germany, is managing three new EuroHPC projects, namely EuroCC, CASTIEL, and Fortissimo for EuroHPC.Primeur Magazineinvited him in its 4m below sealevel virtual studio to explain what the three projects are about.

Primeur Magazine: Can you very shortly introduce yourself and say what your role is in HLRS?

Bastian Koller:I am the managing director of the High-Performance Computing Center (HLRS) which is part of the University of Stuttgart. On the one side I am the managing director but on the other hand I am responsible for the project work and also coordinating some activities, especially the political relevant ones.

Primeur Magazine: There are three new projects that you are coordinating. They were approved by the EuroHPC Joint Undertaking and they will start shortly. Can you just give a brief overview of those three projects?

Bastian Koller:Yes. Within EuroHPC there was the wish to work on national competence centres. There were two of the three projects which deal especially with the set-up of national competence centres on HPC within the frame of EuroHPC. One is EuroCC which is a research and innovation action with currently 36 beneficiaries in 33 countries. Each of them is between 1 and 2 million euro of budget to work and to set up the frame of a national competence centre in their local area, in their nation.

Then we have a second project which is CASTIEL which is a coordination and support action. CASTIEL is a project which wants to establish the European level collaboration between the single national centres. The idea is not to say: "Okay, here is your money for the national competence centre and good luck with it". Instead, the idea is to bring people together to understand that we have different levels of maturity in the different nations and that these nations can benefit from each other in one or another aspect. The idea is to bring them together within the CASTIEL project.

The third project is called Fortissimo for EuroHPC or FF4EuroHPC. This project is a follow on for the previously and successfully executed Fortissimo project activities which were led by Edinburgh. Unfortunately, they couldn't do this now because the BREXIT has a certain impact on them. HLRS took over the lead of this. What we want to do here is to bring HPC closer again to small and medium enterprises (SMEs). We want to make business experiments with them which result from open calls for projects, so that at the end we can provide success stories and to show the real impact of HPC in the application in the real world industrial environment.

Primeur Magazine: That is a short overview of the three projects. If we dive a little bit deeper into EuroCC, can you tell a little bit about it? How many partners are there and what will they exactly do during the coming years?

Bastian Koller:EuroCC consists of 33 nations. In total, there are 36 partners. If you look at all the linked third parties coming with these partners, we are over 100 participants. What they do is, they all follow the same approach for two years to set up the framework for what could be a national competence centre. This starts from identifying the competences in each of the countries. Then they continue to work and understand the needs of industry in their respective field but also from other potential customers like in the public administration, like the society in general, and to understand really what are the strengths of the country and what are the weaknesses. This sounds hard but they seek to identify where they are not so mature yet so that they could benefit from setting up this national competence centre and to foster activities to strengthen a global availability of HPC and obviously related technology services. So, it is not only plain HPC but also takes into account Artificial Intelligence (AI) or data analytics as being part of the overall workflow.

Primeur Magazine: You mentioned 36 partners in 33 countries but if we look at EuroHPC there are only 32 Member Countries. Does this mean that there are also countries outside EuroHPC that are in the project?

Bastian Koller:Yes, we have some Associated Countries. Some of them recently joined. It is quite an agile project. When we submitted the proposal, we had set money aside to onboard new members of EuroHPC, which really happened now. The last two ones joining were Iceland and Montenegro coming into the game. We are set up here and we are open and flexible to onboard also other countries further on, if they join. The only thing is that we have to take into account that there will be no budget from the EuroHPC JU side for these two years but we could integrate them into the discussion with some contracts. In this way, they could also prepare for a potential follow-on phase.

Primeur Magazine: Another topic is of course that a lot of countries already have national HPC centres and some countries like, for instance, Germany already have three while others don't. How do you handle that very big range of different kind of national centres?

Bastian Koller:We gave a clear and strict workplan or topics to address for this where we identify which are the common topics for each of them, even again taking into account that the level of maturity is quite diverse. As you say, in Germany we have the three centres. We have the Gauss Center of Supercomputing as an umbrella. Germany has invested a lot of money into HPC over the last years. It is similar for other countries like, for instance, France. On the other hand, you have a lot of the new Members States which don't have that experience in one or another area but in other domains like the algorithmic work and the research behind that.

The idea is to ensure that they all follow a certain plan; that they work on road maps; that they understand what are the things they already have; what are the deltas they have to close; and then really go into that direction. The level of maturity is very interesting but also gives us in conjunction with the CASTIEL project the possibility to do new programmes, like for instance exchanging trained trainer programmes, exchange people, giving courses in other countries which they don't have now available but want to set up.

We want to boost the developments there to better understand also on their side what they really need.

Primeur Magazine: You mentioned CASTIEL. Perhaps it is good to now explain a little bit more about the CASTIEL project.

Bastian Koller:This project has a long history of changing names during the proposal preparation where we somehow always hit the target of some country being not happy because, if you have so many countries, you have the language differences. You, as a German, have an idea of a word which means something very nasty in another language. So, CASTIEL was the final decision. It is a coordination support action. We have diverse activities here. The whole thing is set up in order to provide a frame for the national competence centres to get into exchange with each other and set up working groups on certain domains.

We have identified three main domains of working groups. One is the training twinning mentoring area; one is the industrial interaction area; and the other is about how to create awareness and how to disseminate the results. Indeed, at the end what you need is, if you have a country investing money into these activities, people in the society also have to understand what is the benefit of it.

Primeur Magazine: Is the CASTIEL project only serving the competence centres or has it a wider goal?

Bastian Koller:It will have a wider role. The main purpose obviously is to serve the competence centres. The other thing is that with EuroHPC, there is now a really big strategy implemented. We have connections to diverse areas like the Centres of Excellence in HPC, like PRACE. PRACE is part of the CASTIEL consortium, for instance. We tried to interact and interlink the available expertise and knowledge because, at the end what I don't want to have, is a two-year project reinventing the wheel which has been invented already by things that reinvented the wheel before. We have a lot of excellent activities in Europe which now have to somehow get into contact.

One part of CASTIEL is also to create the awareness of what is there, so taking the competence map from the national competence centres, making like a big competence map over Europe but also understanding which other activities are there; what is already funded; and where they could benefit from each other. The uptake then is obviously the decision of the national competence centres themselves. We will not dictate it. It is really a coordination and support action to rather support and provide them access to this. From my experience, I know all the project developments which have been part of projects where I'm involved in but sometimes, I am surprised indeed. Some projects are very good in delivering results but not really good in spreading the message, and some other activities are more spreading the message but are not up to the point. The interesting part is to get into contact and to foster the developments in Europe.

Primeur Magazine: Perhaps, it is good now to turn to the third project. I will let you pronounce the name so I can remember it and we will go a little bit into deeper detail.

Bastian Koller:The third project is FF4EuroHPC. FF stands for Fortissimo because it is a continuation of the Fortissimo activities. Fortissimo I and II were in the area of manufacturing and enabling, within open calls, a business experiment. The idea was that some centres come together with industrial partners. The main focus is on small and medium enterprises (SMEs) to help them boost their developments and improve their business, so that they can see what is the benefit of applying HPC, AI, HPDA, and whatever is needed to their respective business. This is somehow like an automatism that by having these success stories, we get the attendance from other players in this field because if they see that a competitor has some advantage, obviously they are interested in getting the same advantage done on their side.

Now with FF4EuroHPC, we continue this tradition. We plan for two open calls within the next two years. Of course, we will also have a close link to the EuroCC and CASTIEL activities because ideally, we can help support the national competence centres to provide perfect bids. It is a completely open process. There will be an open call; there will be evaluations of this call; and the best proposals will be selected. The idea is to focus on the success stories' creation. In terms of budget you see this very well reflected because it is a two million budget for the core partners. There are six of them coming from the old Fortissimo consortium mainly and eight million are all for the open calls themselves. There is a lot of money into this indeed to create the success stories.

Primeur Magazine: There are of course other initiatives to support SMEs with HPC technology and also companies working on that in Europe. You also have, for instance, initiatives like Ubercloud which are doing similar things. Will you collaborate with those?

Bastian Koller:We are always open for collaborating. They are now not part of our initiative but they could come to us. I am always open for discussion rather than saying "I am the best one". Obviously, if you always pretend that you're the best one, you try to get all the money but in the end I want to have the impact and the impact means also to see and learn from the others and on the other hand to be open to the others. At a certain point in time, when you have greater business, there are protective measures but in the end, I think we should be as open as possible to interact and to share experiences.

The original Fortissimo was based on centres which had the experience of working with industry and SMEs already. This is a different kind of working than being a plain academic centre. This, by the way, will also be a challenge for the national competence centres, because part of them wants to interact with industry. This means that you have to have a quite different approach towards dealing with industrial customers than you have potentially with academic customers from universities and from research activities.

Primeur Magazine: That is the third project. If we look at those three projects, how do they fit into the overall picture? The EuroHPC Joint Undertaking has announced that there will be something like eight machines which will be online perhaps by the end of the year but most likely by the beginning of next year, including some very big ones. This is a big part, and then there is a smaller part with research. You also have the more support-like activities that you do. How does this all fit together? How do you see that?

Bastian Koller:Usually, I start by distinguishing between the national level and the European level. The national level is what we will realize with the national competence centres. Here, you might have national resources which can be used, national research which is there, and national industry, which is very important. Ideally, industrial activities are addressed on a national level. It is quite hard for some small and medium enterprises from the Swabian area like in Stuttgart, to go to a centre which is somewhere at the other side of Europe, because there is also a level of trust in here. We need to have the national component here.

Then we go to the European level which is dealt with by the EuroCC and CASTIEL interaction, FF4EuroHPC providing them the industry related activities here. Then you have the machines which are there. They will be utilized as far as we can get access to them, but first of all, we have to prepare the users also on the national level to be able to use these machines. We have to learn from the Centres of Excellence and HPC applications what are the benefits of using these big machines or the pre-exascale systems, and even the petascale systems for industry. I think this fits very well together. The main thing is, just for me, that there need to be some years of continuation now to foster all these things and to integrate the bits which are handled.

Also, which we did not talk about yet, you have a lot of FET HPC type of projects which are coming in. There will be one which will be, for instance, looking at Computational Fluid Dynamics (CFD) and CFD OpenFOAM. If you look at CFD OpenFOAM, this is heavily used by industry. This is what industry needs. If you can give access to this code via the knowledge base, to the national competence centres, they can interact with their customers, using these kinds of codes. This is a perfect benefit for the whole area.

Primeur Magazine: I think that nicely covers the whole area of HPC. Just a tiny question: we are now in the middle of the COVID-19 crisis. Did it have any effect? Did it lead to any changes in your projects or project proposals?

Bastian Koller:Yes. On the one hand obviously, we plan now for less face-to-face meetings than before. I think with 33 countries it would be likely that we have some of the risk areas. That is why we go for virtual meetings. I would have liked to see the people in person to be honest. Some of them I know, some of them I do not know but it is as it is. What we have seen is a certain potential also for the national competence centres to have a look on this kind of COVID-19 related research potentials, for instance.

What we learned from a lot of countries is that there was an urgent request sounding like "we would like to do some statistic analytics, can we use HPC for that?" As I try never to rest too much, I am always thinking in my mind how to set up something like a European COVID-19 central where you could bring together developments from projects addressing the prediction of virus spread or something like this. We already have this in some of the Centres of Excellence, but also taking into account other things, so that we can provide a toolbox for this kind of occurrences and potentially improve the way how we can deal with that by having things ready.

I think the projects directly are not that affected with this obviously. We will have to see for FF4EuroHPC project in how far the SMEs can deal with the economic crisis. I think it is very good for them to get funding for the activities because this is where they can potentially boost their innovation potential. In general, I think HPC will be very important for the next year because it can help the existing businesses to innovate. If you look at the COVID-19 impact, the big players as well as the small players, everybody has an issue.

My fear is there will be somehow an accelerated market adaptation where you have to find out which are really the ones which provide products with an innovative approach to keep standing there. We don't know how big will be the reduction of the overall number of companies that we will see unfortunately but this is how it is. We have to see that we strengthen those which are open to and which can provide innovations. I think that HPC, in conjunction with AI and HPDA and whatever will come up, will be very helpful for this but it needs to fit for the purpose of these players. I think it is a good start having these national competence centres where we can bring in the knowledge how to deal with them, how to support them best, and how to make the maximum utilization of available resources in Europe.

Primeur Magazine: That would also be a very nice closing for the interview, I think. Thank you a lot for this interview.

Ad Emmen