From the virtual "four meters below sealevel" studio in downtown Almere we had the pleasure to talk with Anders Jensen who is the newly appointed Director of the EuroHPC Joint Undertaking.
Can you tell a little bit about yourself since you are new to the EuroHPC Joint Undertaking?
Anders Jensen: Sure. I am Anders Jensen. I am the newly appointed Executive Director for EuroHPC. I have been in the job exactly one month on October 15, 2020 and I have to say it has been like drinking from a fire hose. A lot of good things have been going on but we have also been through a process to make the joint undertaking autonomous. One of the elements in that was me being in the seat but behind the scenes, a lot more things had to happen. We had to set up systems so that we could implement our own budgets and run our own business. All of that has been going on in the space of the last month. So far, there has been not a single boring day during my first month. I am very excited to be on board and I am very excited about what we are going to be doing, going forward.
Congratulations with your appointment. We also saw that you assembled already quite a team around you. It is not like you have to do it all alone.
Anders Jensen: That is correct. I am very fortunate that there are a lot of good people who have been with this undertaking for a lot longer than me, who has joined the undertaking as well. We have a total of 11 people as we speak and we are continuing to grow that staff. We have new hires coming in over the coming months. There are good plans for us to be able to stand on our own two feet and take the ambitious agenda forward.
We know it is already a long time ago, more than a month ago, but what did you do before joining the joint undertaking?
Anders Jensen: I have built my career basically around three very different pillars in my life. I started out doing a decade of Research and Development (R&D) in which I did the chip design for embedded systems and software for embedded systems. After that, I joined an airline company in Luxembourg. We set up an IT company to serve our airlines. I was there for a decade as the Chief Technology Officer. Finally, I moved to NATO and for the last nine years I have been running the IT of NATO headquarters. So, in summary, I know what it is to do research, I know what it is to do development, I also know what it takes to run a service provider, and finally, I know something about international environments, international organisations, and public procurement.
That sounds all very good. The joint undertaking started out somewhere in March 2017, so where are we today? What is the current status?
Anders Jensen: In its infancy, I believe the EuroHPC Joint Undertaking was set up in 2018. What has happened now is that all of this work has come to fruition with the joint undertaking becoming autonomous. Leading up to that has been a lot of work from colleagues within the European Commission. That has gotten us to the point where we are today. That includes the agenda that we will be discussing also around the machines, the procurements, and so on. While I have had the pleasure of coming in as you can say to take the victory lap and sign some contracts, that is of course the result of all the work that has been done by the Commission before we became autonomous in order to go through the bidding rounds for systems, doing all the evaluations, and all of that being lined up so that once we declared autonomy, I could start signing those contracts on behalf of the joint undertaking. However, during those last two years, a very significant amount of work has been done leading us to where we are today, where I have had the pleasure of signing three petascale systems. During the time that comes you will see more announcements. The aim is still to get to the eight machines by having those signed off by the end of this year.
When we understand correctly, the systems will all be operational early next year?
Anders Jensen: At the end of day, we need to sign the contacts, we need to agree on the implementation dates. Some of them will be earlier than others. The expectation is that as early as March we will have some of the first systems live. Of course those that we have yet to sign, which are also the bigger systems, they take a little longer. Our hope and aim is that definitely everything will be live next year. Some will be earlier, some will be later.
The promise from the Commissioners was that we would have three systems of the EuroHPC in the top five. Will we reach that and when?
Anders Jensen: That is still the ambition. Of course, those are the three final systems of which you will see some announcements in the weeks to come. For those systems, the aim is very much still to be up there.
For the other ones, the total should be somewhere in the top 50?
Anders Jensen: That is correct.
This means that we will have to wait for the November list of the TOP500 next year to see whether we reach that.
Anders Jensen: For the final systems perhaps. Again, I cannot prejudge what the discussions with the companies will be but clearly, the goal has to be that they are in production for 2021.
Of course it is not that important to be on the top of the list. In the HPC community, we are all happy that the Fugaku is now on top because they are doing some very nice research on COVID-19. Each day you see something new. It is this type of research that we hope that will come out of the systems. Apart from the systems, there are also innovation projects. Can you tell a little bit about that?
Anders Jensen: Yes, that is also an agenda item that has been pushed very heavily by both the Commission and the staff that is now within the joint undertaking for the past time. Where we are with that is that we have a couple of projects that have now started in and around Competency Centres. We have 19 projects from our last call and they are being evaluated now. Grants will be issued on those. We have also had our pre-exascale pilot out. We have received responses on those. It is an agenda that is building up. There is quite a lot going on but again, all has been waiting for the joint undertaking to be fully up and running. We are pushing that out. We will have one call that we are putting together now which is really exciting. It is around training and education on HPC. We are working together with the different stakeholders and participating states now to finalize that call. There will be much more information coming out on this as the plot unfolds.
The joint undertaking is a separate legal entity. As you said, there was a lot set up from the start. It was modelled after the ECSEL Joint Undertaking which is more about hardware. What are the lessons learned after two years of joint undertaking?
Anders Jensen: On the lessons learned, I may not be the best placed person to speak about it because having come in a month ago, this is what I have learned. I don't think I am in a good position to comment on ECSEL other than the fact that from what I hear, they are also quite successful. There is one fundamental difference which is that the co-investment in procurement is something that we are doing whereas the ECSEL set-up is more, as I understand, around grants and research. What is very unique to EuroHPC are these procurement actions where we are co-funding machines with participating states or consortia of participating states.
The idea of the joint undertaking is that we put together financial resources from the countries together with financial resources from the European Union which means that we need to have some kind of co-financing scheme. How did that work out during the first two years?
Anders Jensen: Personally I think everyone agrees that has worked out really well. The machines will now be delivered but they will be bigger than if we had invested on our own. I think what has been created with EuroHPC is the potential for a true win-win situation. We all get access to better and bigger equipment by co-funding, by putting our eggs together, rather than everyone buying their own equipment. What we have seen - and I think that is something that has been very positive and may not have been originally envisaged - is that even on the national funding side, the participating states have joined in consortia in order to pull their money along with the money coming from the Commission. If we have really managed to get access to some machines, this may not have been possible at least for some participating states on their own.
For the systems it worked out fine but for the projects we think it worked out not as good as was hoped.
Anders Jensen: For the projects, this is one of the discussions I am currently having with the Governance Board members. Being new, it is important to take stock of where they wish and would like to see things differently. I think there are various views on what could be tuned on the project side and I am very much listening to those so that we can try and adjust course if it is required. If you look at the pipeline there have been some really interesting projects and project consortia as well. I think there are successes amongst them. If we can broaden the participation, I think that would be very positive. Whether we can further broaden the participation in some of the calls is something where I am going to have to get my teeth in.
In some smaller countries, like for instance the Netherlands, some people are not that happy, that is not with the joint undertaking but with the fact that the co-financing was so very difficult and basically non-existing. This means that you have to do the co-financing yourself which is kind of complicated but we will see what happens in the next phase. This brings us to another topic. The Commission proposed a new council regulation and the council regulation is in fact the legal basis for the whole joint undertaking and all the co-financing. The reason for bringing out the new regulation is that the previous one only ran until the end of this year so you need a new one that will cover the period until 2026 or 2027. What will be the process? The Commission proposed the regulation so what will happen next?
Anders Jensen: As you correctly said, this is a proposal of the Commission and right now this is something between the Commission and the participating Member States to negotiate. I cannot prejudge the outcome of that. All I can say is that the proposal that is on the table is extremely exciting and includes some very interesting elements. From my point of view, being new in the seat, it is fantastic to see such a backing. It is clear that it is believed that what we are doing in EuroHPC and what we will be doing, going forward, is extremely important to Europe. You cannot ask for a better backing sitting in my seat than that. It includes some fantastic new elements. Some are natural extensions of what we have already done. Of course, we need to continue into the exascale and post-exascale but there is also the notion of what happens to quantum computing, what happens to neurologic processing, which is just extremely fascinating. However, I am here to administrate the current regulation as it is until I am given a new regulation. I will be monitoring it closely but this is on the operational arm you can say or the executing arm and the Commission people will be the ones negotiating the new regulation.
Yes, of course, and we have to wait until the Council finally approves it. One of the questions is when will we have exascale in Europe. Until now, the plans say it will be in 2023. Are we still on track for that?
Anders Jensen: This is one of the things that the new regulation sets out. I think it is a very ambitious target and we should work towards it. We all need to recognize that there is some research and some work that need to go into that. I must admit that this is one of the things that is extremely exciting: to be able to plot out that agenda so that we can reach this ambitious target. From my point of view, this is what is being laid out in the new regulation as the ambition so that is what we are going to have to work on.
What we understand is that the plans are still to have a machine which is as fast as possible and another machine which is also as fast as possible but is using as much European technology as possible too.
Anders Jensen: The aim is definitely - and I think this is a very important goal of EuroHPC - to get European technology into the game. Whether it is one machine or whether it is both of them, at the end of the day, the aim must be that we go for as much European technology as we can. We set the research agenda to make sure that we develop that knowledge in the areas where we may be in order to get to that goal.
When we see the plans, there is also talk of post-exascale systems but that will be in 2026 so the proposal from the Commission reaches to the end of 2027.
Anders Jensen: The proposal is basically that we wish to continue to lead in the field. Of course we need to continue to have bigger and bigger machines in order to be able to maintain that lead. We are now investing heavily in the eight machines but if we just lay back afterwards and watch those, in a couple of years we will be back to square one. That is also why I find the proposal of the Commission very exciting because it demonstrates that we need to continue to invest in the field. At the end of the day, once we have managed to climb the hill that is called exascale, there will be another hill afterwards. If we stop, we are back to square one.
Yes, since there is also a plan to have accelerators like quantum and neuromorphic computing.
Anders Jensen: Yes, and that is an exciting parallel track to start looking at and find out exactly because we still need to find the good applications for these technologies. If we do not look at them we risk being in a situation where the world outside Europe may have found those applications ahead of us. I think it is important that we make sure that we build up the knowledge and that we are able to deliver on those aspects.
Concerning the financing for the next period, it is not clear to everyone how that works. What we understand is that the financing will come from the next Framework Programme that runs until 2027, but the cost that will be spent will be spent until 2033. If you acquire a machine in 2027 you still have to maintain it. Is this correct?
Anders Jensen: I have to be careful with saying this is correct because this is the proposal of the Commission. However, it is clear that if we were to put in the last machine in 2027, it would be really silly not to continue to operate it right. Just like the machines that are being delivered now as part of the 2020 plans will continue to operate into the next era. So, the answer is yes. That is the logic answer.
Yes, we were already thinking about the situation after that period.
Anders Jensen: Of course, my ambition will be that this is so successful that there will be another round of funding for the next era. That has to be the goal.
Great, we will never have a dull moment in the coming years.
Anders Jensen: No, give me another couple of months to get properly into the seat and then ... but so far, it has not been done.
Thank you very much for this interview.