- more details and access regulations
(internal use only)
The Experimental Compute Cluster: ehlers provides a basic infrastructure which enables researchers in SimTech to carry out large-scale simulations, such as compressible flows that occur on commercial aircraft or wind turbines, molecular dynamics, or porous media flows. These simulations require immense computational power and benefit from the parallel processing capabilities of experimental compute clusters.
Experimental compute clusters enable researchers to conduct extensive parameter exploration and sensitivity analysis, helping to identify optimal configurations and uncover hidden dependencies within complex systems. This iterative process accelerates the refinement of simulation models and enhances the reliability of scientific results. They facilitate the integration of machine learning algorithms into simulation workflows, optimize simulations in real-time and enhance the predictive capabilities of scientific models.
With a compute cluster comprising twelve CPU nodes, four GPU nodes, a single NVIDIA DGX, and miscellaneous administrative and file servers with a total of around 700TB available storage space, the Experimental Compute Cluster: ehlers thus builds a bridge between the institutes at the University of Stuttgart and larger national and international supercomputing centers, such as the HLRS.
As an experimental platform, the Experimental Compute Cluster: ehlers is maintained and extended continuously. Researchers get technical and scientific support to scale up their simulations gradually, test novel architectures, and optimizing code and workflows for larger systems.
Experimental compute clusters play a pivotal role in advancing the field of simulation, providing researchers with the tools to explore new frontiers in hardware architecture, simulate large-scale systems, conduct parameter studies, integrate machine learning, and optimize performance.
The Experimental Compute Cluster: ehlers consists of a compute cluster comprising twelve CPU nodes (each with 128 AMD EPYC 7702 cores), four GPU nodes (each with 64 AMD EPYC 7543 CPU cores and 4 NVIDIA A100 GPUs), a single NVIDIA DGX, and miscellaneous administrative and file servers with a total of around 700TB available storage space.
To enable fast simulations the cluster is built with large compute nodes each with multiple data-center GPUs and additional high-speed storage.
All compute nodes of the Experimental Compute Cluster: ehlers are connected with a high-speed Infiniband network to enable the fast and efficient computation of large-scale simulations.
The compute lab is named in honor of Prof. Dr.-Ing. Wolfgang Ehlers, the first director of SC SimTech and the spokesperson of the former Cluster of Excellence EXC 310.
Contact:
Malte Brunn
Cluster Computing Systems Expert