2021-03-02-Election for the steering committee

lrz

"Computing time should be used for impactful science"

Experience counts: For the third time, Prof. Dr. Peter Bastian from Heidelberg University has been elected chairman of the steering committee at the Leibniz Supercomputing Centre (LRZ). Bastian heads the "Parallel Computing" working group at the Interdisciplinary Center for Scientific Computing (IWR): "We primarily develop efficient and robust methods for solving certain systems of equations," is how he describes their task. Bastian's deputy is Prof. Dr. Gerhard Wellein from Friedrich-Alexander Universität in Erlangen-Nuremberg, who is also the spokesman for the Competence Network for Technical Scientific High-Performance Computing in Bavaria (KONWIHR). The steering committee is elected every two years and consists of a total of 15 scientists representing the LRZ and its parent organisation, the Bavarian Academy of Sciences and Humanities, as well as the German Research Foundation (DFG). The board is completed by two advisory members. It is true that the major projects for the SuperMUC-NG the large-scale tasks with more than 45 Million core hours per year, are awarded via the network of the Gauss Centre for Supercomputing (GCS). But the LRZ Steering Committee extends computing time and evaluates the scientific merit of research below this limit. Last year, that numbered about 70 applications.

The steering committee assesses which scientific work is computed on SuperMUC-NG: Are there often discussions about this in the committee? Prof. Dr. Peter Bastian: No, not as a rule. Each member of the steering committee is responsible for his or her own area. The chairmen commission reports from external experts and use these to draw up a funding proposal. The chairman of the steering committee and his deputy are also members of the computing time commission of the Gauss Centre for Supercomputing (GCS), where the large-scale applications are reviewed. There, however, we often see discussions, and that is a good thing.

How many papers per year does the LRZ Steering Committee evaluate? Bastian: In 2020, there were a good 70 proposals, 20 of which were large-scale projects that ran through the GCS. In addition, there are a good 30 cost-neutral computing time extensions that can be decided by the umpires without an expert opinion.

At the moment, the handling of idle times at the SuperMUC-NG is being discussed. Without an expert opinion from the steering committee, however, nothing is possible - how does the committee intend to fill the gaps? Bastian: The LRZ has to buy a certain amount of electricity (well, physically correct it is energy. The unit is kWh.) per year from its supplier to get a favorable price. Otherwise, a large extra payment is due. So, filling idle times actually saves money. They occur at the SuperMUC-NG when the machine is cleared for large jobs. These voids can only be claimed by certain work loads that are not yet available on SuperMUC-NG. The idea is to test the technical side first with a designated customer. If this works, there could be a (limited) supply here, nevertheless the computing time should of course be used for impactful science.

Do you actually compute on our machines yourself? Bastian: No, my research group mainly develops efficient and robust methods for solving certain systems of equations. At present, the challenge is not so much the use of a large number of processors as achieving high computing power on a CPU or GPU. Therefore, the computers on site in Heidelberg are sufficient for us.

As a professor, you assess research work, doctoral theses at your chair anyway - why do you engage in the LRZ steering committee on top of all that? Bastian: Well, the mutual review of research papers is of central importance. Working as an umpire is very exciting because you get to see great research.

You know the SuperMUC, but now also the SuperMUC-NG: Have the research projects on the two computers changed - in which direction? Bastian: The projects have shifted somewhat toward solid-state and quantum physics. However, this is not so much due to the change from SuperMUC to SuperMUC-NG - the architectures are very similar - but to a new acquisition in Jülich. Temporarily, less computing power was available within the GCS infrastructure.

In light of new quantum computing technology, HPC is is considered in some circles as oldfashioned. Will quantum computing soon replace supercomputing? Bastian: Quantum computers are certainly an exciting development, but this approach will first be beneficial for certain applications, such as tasks in discrete optimization, quantum chemistry or cryptography. It will take quite a long time before you can use it to simulate the flow around an airplane, the combustion processes in an engine, or the groundwater balance in southern Bavaria. So the SuperMUC-NG won't become obsolete any time soon.

Who would you like to discuss HPC with sometime and why? Bastian: I can discuss with most colleagues at conferences, but if I could talk to the mathematician John von Neumann, one of the fathers of scientific computing, up in heaven, that would be very exciting – how he had the idea for the von Neumann computer, what possibilities he saw for computers and how the technology was actually. (Interview: vs)

bastian  

Prof. Dr. Peter Bastian, Universität Heidelberg

wellein

Prof. Dr. Gehard Wellein, Friedrich-Alexander-Universität Erlangen