High-Performance Computing Cluster

• Term: Funded in 2022
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 502227537

The research at the University of Potsdam is performed by seven faculties and four Research Focus areas including Data-centric Sciences. Various groups require high-performance computing (HPC) facilities, and ongoing studies range from the simulation of high-energy astrophysical phenomena in our Cosmos to the sequencing of genomes and the study of molecular dynamics. Over the last years, the university-wide HPC infrastructure has undergone several improvements, e.g., through upgrades of the batch system and the inclusion of the HPC network into the ESXi host system. Despite these upgrades, the main HPC system consists of only a dozen nodes. Planning for the future, the University of Potsdam has recently hired an HPC System Administrator, such that personnel resources are available for an extension. In addition to this administrative support, the ZIM (Zentrum für Informationstechnologie und Medienmanagement) provides infrastructure and support for the long-term storage of scientific data (44PB), which ensures that research data are findable, accessible, interoperable, and reusable. The continuous investment in computational sciences is also evident in the extension of the Science Campus Golm, where many buildings are under construction, in particular, the new building of the Institute of Computer Science hosting a dedicated machine hall with an uninterruptible power supply, a cooling infrastructure, and a fast connection to the intranet. The environmental impact is kept minimal through re-using the produced waste heat of the servers for heating the building. The importance of computational science at the University of Potsdam is reflected in the curriculum by introducing the Master of Science Program `Computational Science'. Besides this dedicated program, several individual lectures focusing on computational sciences complement numerous B.Sc. and M.Sc. Programs. Despite these investments, the current computational demand is not covered by the existing infrastructure, and a significant upgrade is required to support state-of-the-art scientific studies. An 'in-house’ HPC system located at the Science Campus Golm, with capacities between large HPC centers and small servers, is essential for a productive working environment. Such a system allows medium-size tests and optimization studies, as well as the development of computational tools that require too many resources for single workstations. Importantly, this medium-sized system prevents delays in development work through additional administrative procedures and long queuing times. Furthermore, an HPC infrastructure at Golm with a fast connection to individual institutes on campus would expedite the data transfer, a bottleneck for some studies. Given these benefits, we are applying for financial support to install a new HPC system. The system will consist of three main components: ~100 compute nodes; a few high-memory nodes equipped with large DRAM/DCPMM memory; a few GPU-ready-nodes.

• DFG Programme: Major Research Instrumentation
• Major Instrumentation: Hochleistungsrechner
• Instrumentation Group: 7040 Vektorrechner
• Applicant Institution: Universität Potsdam
• Leader: Professor Dr. Tim Dietrich