Michael Resch, Director of HLRS — High Performance Computing Center Stuttgart, shares with us his visions about the next challenges in HPC and on the development of the merge of HPC and GSS.
The major challenges in HPC for the coming two years are two: one is the end of Moore’s law; the second challenge is that we would like to reach the exaflop, which we can only do if we use millions or tens of millions or even hundreds of millions of cores, which will make it extremely difficult to program these systems.
But which is the vision behind the merge of HPC and GSS? In Resch’s words it looks like an idyllic encounter: «On one hand we have a field that has a tremendous need for compute power and it’s only in its beginning when it comes to simulations, while HPC will reach the exaflop in maybe three to four years from now, which will provide this simulation performance that is required to solve some of the problems of global systems science».
The difficulty is that HPC and GSS are two different fields that talk two different languages.
«We have to find the right models for describing Global Systems and we need to turn these models into programs, that are scalable on millions of cores», Resch said.
With the merge of HPC and GSS we target a new challenge.
«Most of the Centres of Excellence focus on field in which HPC is active for decades — explain Resch — . GSS is a field that has to be developed, so the purpose of the CoE is not to get another “epsilon” on the development of the field, but rather to provide or prepare for a quantum leap, for the first time what we offer is that people working in GSS actually have access to the fastest systems in the world».