Which compute platform is right for me?

Large and Huge HPC Workloads

• You are an experienced user of HPC systems.
• You want to run large MPI(/OpenMP) parallel simulations on thousands of CPU cores.
• You need several million core-hours to complete your simulations.
• You need very large and fast parallel storage resources for results and data processing.
• You require access to special application software installed on SuperMUC-NG like ANSYS CFX & Fluent, Star-CCM+, Abaqus,... for large-scale simulations with several millions degrees of freedom.

Small and Medium HPC Workloads

• You want to run serial or small to medium sized MPI-parallel applications up to a few hundred cores.
• You need less than a million core-hours of compute time for your project.
• You need access to teramem, a node with 192 CPU cores and 6 TByte of shared memory capacity for data analysis.
• You want to obtain experience in optimizing and executing MPI-parallel applications on HPC systems and need a platform to develop parallel programs.
• You want a long-term and reliable compute resource for your work group.
• You need to execute parallel jobs that run commercial application packages licensed by LRZ for campus use
• You require access to special application software installed on the Linux cluster (Matlab, ANSYS, Star-CCM+, Abaqus, Gaussian, VASP, ...).

Users from Bavarian universities

Access to the Linux-Cluster

Data Analytics, Machine Learning, AI (Model Training/Inference)

• Various compute and storage components for AI (model training, inference), data analytics and big data research and application.
• CPU-only systems with large memory for big data applications.
• Individual single & multi-GPU servers for small- to medium-scale data analytics and AI workloads.
• Clusters of multiple multi-GPU systems for medium- to large-scale AI workloads.
• Dedicated ML & AI storage systems for high-performance, low-latency data input/output.
• Access via traditional SSH-based login nodes as well as a user-friendly web interface, including popular development environments JupyterLab and RStudio Server.
• OS-level virtualization/containerization for highly adaptable, yet consistent and reproducible work environments.

Servers with High Availablity and Reliability

• You need one or more highly available virtual server to run important service(s) and you need a solution which protects you from loosing data.
• Only master users can request virtual machines.

SuperMUC-NG users

Users from Munich or Bavarian universities

Governmental and non-governmental research institutions that are predominantly funded by public funds and other science-related institutions