My name is Gerhard Hellwellein. I also would like to welcome you on behalf of two institutions,

first on behalf of our recently established new National Center for High Performance Computing,

NHR at FAU, but also on behalf of the National NHR Alliance, where I am a member of the Board of Directors.

So I have the honor to give the first talk. Basically, this is more kind of a formal talk and information talk.

This is the boring side of your workshop symposium, but I'm pretty confident that the more we get into science in the later talks,

the more interesting, the more exciting this workshop will be. I will report on a newly established infrastructure at a national level

and at FAU, which will be open to you and researchers from all German universities.

So let's start with high performance computing in general. So I assume as you're doing simulation in the context of atomistic structures

that at some point in time, you may have already used high performance computers, large scale computers, computational resources

to solve problems, scientific problems through simulation. So nowadays HPC is central to simulation in various fields of science and research.

So as you may know, HPC systems are getting more and more complicated in various dimensions. So the first one, this is more our side of the game.

So we need to do larger and larger investments in these systems. We have to cover increasing operational costs.

You all know about increasing electricity prices, the systems themselves as we approach the end of Moore's law, consume more and more power.

And at the end, this also means that we have to build better or even completely new infrastructures to operate and run these systems efficiently.

Using the systems, we have to face the same challenges. We see new programming models, we see new architectures, we see heterogeneous architectures.

We see massive parallelism at various levels, starting at the chip level up to the large scale systems.

We have to consider new algorithms to match the requirements and also to match the capabilities of these new hardware structures.

And last but not least, even using those systems, just running a code is also becoming a more and more complicated task,

because I have to consider which of those many machines is best for my application at hand, which solves my problem fastest, maybe also most efficient, and how to apply for those systems.

So this is a complex environment, and I would like to report on developments in the context of this new infrastructure, which should help you to make best use of these infrastructures.

So let's start simple. For a long time now, the so-called HPC supply pyramid is established in Germany, which defines the strategy as outlined by the German Wissenschaftsrat

to provide HPC capabilities and capacities to German researchers. So basically in this pyramid, the capability of the system increases from bottom to top,

but also the number of the systems available in Germany increases from top to bottom. So you may have seen that before.

And the statement is that at all levels of this pyramid, we need sufficient systems matching the demands of science and research.

So you may know that at the top, tier 0, tier 1, we have the so-called Gauss Center for Supercomputing.

So these are three large computer HPC centers located in Garching, Jülich, and Stuttgart, which run basically top-of-the-edge class systems,

systems in the range nowadays of multiple of dozens of petaflops, consisting of hundreds of thousands, maybe up to a million of cores.

It has been early recognized that operating these systems and doing these investments, and the machines are offered nationwide.

So covering these costs cannot be done by a single state or institution. For that reason, the Gauss Center has already an established funding structure

where all the costs, including operational costs, investment, and personnel, are covered by the state governments and the federal governments.

Below that, for some time, for a long time, we had so-called tier 2 systems and the tier 3 systems,

where only half of the investments were covered by DFG and all the other costs had to be covered by the local institutions.

And with NHR, this funding structure now changed. So this Bundländerfinanzierung, GWK-finanzierung, has been extended also to the tier 2 systems,

mostly located, or all of them located, at German universities or directly integrated into academic research and into academic applications.

So with this joint funding, also the resources provided by those centers will be available for all researchers in Germany, independent of the location of these NHR centers.

So the idea, as requested by Wissenschaftsrat in 2015, is that we establish a coordinated network of national high-performance computing, NHR, at German universities,

which is jointly funded by federal and state governments, so no costs in operating for the local institutions, but offering all services to German institutions, to German universities.

So basically, this is maybe not the surprising, the exciting part of the story, because you just get another layer of hardware, which you can use for your simulation.

So the important point in this funding program is that not only infrastructure costs, hardware costs, and operating costs are covered.

There's also a strong emphasis that this new NHR network, the NHR alliance, shall strengthen the methodological competences at the centers and their users through coordinated training and continuing education of the users.

And a strong emphasis is here to promote young scientists, to make them HPC experts, to make them educated in this complex environment.

So what is the financial view of that picture? Basically, this is a long-term investment of Bund und Länder.

The total program has a volume, you may have heard that in the welcome address of our vice president.

So there's a total funding available for 10 years of 625 million euro, and currently eight NHR centers form this NHR alliance, and this NHR alliance, these centers officially started January 1st this year.

So this is the background, that means we have a completely new infrastructure, which should foster HPC in general and numerical simulation in general, in Germany.

So what are these eight, currently eight centers of the NHR alliance? So I have shown here an overview over the eight centers.

So I expect that most of you at least know one of these centers, all the universities which run or operate these centers.

So beside FAU, beside Paderborn and the Zuse Institute Berlin, the three centers which also have proposed or established this atomistic simulation lab.

We have RWTH Aachen, we have Technical University Darmstadt, KIT, and we have Technical University Dresden as current members of the NHR alliance.

So let's again dive a little bit deeper into the mission, the ideas and the goals of this new NHR alliance.