Compute Cluster

• Subject Area: Computer Science
• Term: Funded in 2025
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 569233494

The newly established Faculty of Computer Science and Data Science at the University of Regensburg plans to acquire a central CPU/GPU cluster to cover the basic computing time requirements of the various chairs. Modern computing resources play a key role in gaining knowledge, particularly in the research areas of Computational Methods in the Natural Sciences and Human-Centered AI. This applies in particular to research activities in the areas of machine learning and artificial intelligence, image processing, computational statistics, theoretical computer science, bioinformatics, and computational quantum science, which use various approaches to develop and analyze efficient methods for digital data processing. The working groups involved cover a broad spectrum of applications, ranging from the natural sciences, including biomedicine, chemistry, and quantum physics, to clinical research and business informatics. A variety of methods and algorithms are used for these purposes, the optimal use of which requires different resources. Accordingly, we are planning a heterogeneous CPU/GPU cluster with three different partitions, the configuration of which is tailored to specific application purposes. A GPU partition is planned for the numerous algorithms that are fundamentally based on deep learning. Other methods in machine learning, bioinformatics, and computational physics do not benefit from GPU accelerators but require particularly large amounts of RAM. A partition with particularly powerful CPU nodes is planned for these applications. Another CPU partition will cover the needs without special requirements. To enable efficient parallelization across multiple nodes and fast data transfer to the mass storage systems, a connection via InfiniBand is planned. Furthermore, in addition to an HDD mass storage, a global SSD storage is planned, in particular to make the constantly growing data sets available for training deep learning models. The cluster will be located at the Leibniz Supercomputing Center in Munich and will therefore benefit from the there experience and expertise in the professional operation of high-performance and supercomputers as well as AI systems for research purposes.

• DFG Programme: Major Research Instrumentation
• Major Instrumentation: Compute Cluster
• Instrumentation Group: 7000 Datenverarbeitungsanlagen, zentrale Rechenanlagen
• Applicant Institution: Universität Regensburg
• Leader: Dr. Markus Schmitt