Massive Parallel CPU/GPU Compute System

• Term: Funded in 2021
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 456666331

Since 2012, a computer cluster with 512 CPU cores, 2 TB of global memory and 4 GPUs has been successfully operated at the University of Osnabrück's computer center. However, in terms of the number of available CPU cores, GPUs and working memory, this cluster no longer meets the current requirements for advancing research in the applicant research groups. Since this device is the only one of this performance class at the University of Osnabrück, this application is intended to finance a replacement. The planned projects focus on computer science, physics, biology and cognitive sciences. The main applications will be computer simulation of condensed matter, machine learning on image data, simulation of computer networks and the processing of large amounts of sensor data. For this proposal, the need for massively parallel computing power of the applicant working groups was estimated in advance, based on the work already done and future planned work. This estimation justifies the configuration of the proposed device. Due to the heterogeneous research topics and respective applications, a computing cluster is needed, that both covers the need for many computing cores with sufficient working memory (simulations in physics, sensor data processing and network simulation in computer science) and supports the planned research in the field of machine learning (computer science, biology, cognitive sciences) by providing powerful GPUs. The selected machine provides a total of 5,120 cores, 8 GB RAM per core, 8 high performance GPUs and 500 TB redundant disk space. This configuration represents a significant improvement over the outdated existing cluster computer and will allow the applicant research groups to continue to conduct nationally and internationally competitive research. In addition, a new computer center was put into operation at the University of Osnabrück in 2020. It complies with Bitkom Category C and thus also fulfills all requirements regarding accommodation, air conditioning and access authorization for the new equipment.

• DFG Programme: Major Research Instrumentation
• Major Instrumentation: Massiv-paralleles CPU/GPU-Rechnersystem
• Instrumentation Group: 7000 Datenverarbeitungsanlagen, zentrale Rechenanlagen
• Applicant Institution: Universität Osnabrück
• Leader: Professor Dr. Joachim Hertzberg