SuperMUC-NG: Master of scientific Diversity
Every two years, the Status & Results Workshop takes place at the Leibniz Supercomputing Centre. Here, the international HPC community learns about projects which were computed on SuperMUC-NG.
The numbers speak for themselves: since 2019, when SuperMUC-NG went into operation, its 311,040 computing cores have worked almost 7.5 billion core hours. The leadership class of the Leibniz Supercomputing Centre (LRZ) in Garching near Munich has completed around 2 million jobs for 471 research projects. In May 2023, users presented selected work at the virtual three-day Status & Results Workshop. "We are thrilled by the international interest in this workshop," said Prof. Dieter Kranzlmüller, head of the LRZ. "We learn from the researchers how we can further optimise our HPC and IT offerings." Summaries of the lectures can be found in the report "High Performance Computing in Science and Engineering" or as "Lectures on Demand" on the LRZ YouTube channel.
The 20 lectures illustrate the diversity of scientific disciplines that SuperMUC-NG is used for. For example, Prof. Andreas Hinz from the Ludwig-Maximilians-Universität in Munich showed how supercomputing helps to solve old mathematical puzzles. Hinz and his team are investigating the metric structure of Hanoi graphs, on which the strategy game Tower of Hanoi is based. They also provide mathematical models for disciplines ranging from neuropsychology to computer science.
From biomechanics to seismology, math and transport
Prof. Frauke Gräter at the University of Heidelberg is working on molecular biomechanics and tracking down signs of tissue fatigue. Her research group has used SuperMUC-NG to model the mechanical stresses under which collagen - the protein that gives connective tissue its stability - can rupture. Fluid mechanics and turbulence is the field of research of Prof. Jörg Schumacher at the Technical University of Ilmenau: he uses HPC to analyse thermally driven flows, as they occur in technical devices or in astrophysics. SuperMUC-NG has also simulated earthquakes, emissions from road and air traffic, and blood flow in veins and arteries. The model of the CompBioMed2 research excellence cluster was visualised at the LRZ. "One development we have observed in recent years is a steady increase in projects in the field of high-energy physics, especially quantum chromo dynamics (QCD), but the need for supercomputing is also growing in the natural, engineering, biological and environmental sciences, as well as in pharmaceuticals and medicine," summarises Dr. Gerald Mathias, head of the Computational X Support Team at the LRZ. "More and more projects are using techniques such as deep learning and machine learning to analyse the huge amounts of data they generate".
Phase 2 contains CPU and GPU
For this reason, the 240 compute nodes of SuperMUC-NG Phase 2, which will soon be operational, will not only integrate CPU, but also the latest GPU from Intel (Ponte Vecchio), specialised in artificial intelligence methods. The new system, like other LRZ services, was the subject of the Extended User Forums on the workshop afternoons: in addition to storage options, cloud services and course offerings, the technical resources for deep and machine learning were presented, in particular the LRZ's systems for artificial intelligence as well as the CS-2 system from Cerebras Systems with an extra-large chip and flexible, large data storage. The first services in the field of quantum computing and practical tools for structured, professional management of research data also attracted the interest of the 100 spectators.
Good to know
Some lectures of the status & result workshop are available at YouTube. Descriptions of more projects for SuperMUC-NG and some experiences with the system you will find in "High Performance Computing in Science and Engineering". Get a copy here.