HPC | High Performance Computing
For support or further information about the HPC Systems at FHI, please contact:
Simeon Beinlich (Phone: +49 30 8413 5270, Email: beinlich@fhi-berlin.mpg.de)
General Computing Services at the FHI
We provide compute servers for various use cases: Compute servers with virtual desktop access (for GUI applications), python servers (jupyterhub) and compute servers with GPU acceleration (in progress). For specific use cases that require dedicated servers, e.g. for data postprocessing, real time processing, etc. , we are open to discuss possible pathways and offer suitable solutions. Due to the effort of managing multi-user Microsoft Windows servers, we only provide Linux-based (Ubuntu, Debian, etc.) compute servers.
Any member of the FHI can use these services, however registration might be required and access is only possible from within the FHI / eduroam network.
JupyterHub - Python Server
A powerful JupyterHub Server is provided at jupyterhub-01.compute.fhi.mpg.de. It can be used to execute Python scripts, notebooks and other code.
For users that are new to scripted analyses, we offer to support them in setting up the Python setups (packages, modules, scripts etc.) and also to provide starting points for their first python scripts or notebooks.
Please register by Email (Email to beinlich@fhi-berlin.mpg.de), and have a look at JupyterHub - Python Server!
Analysis - Virtual Desktop Server
A Virtual Desktop Server with full Linux Desktop Environments (e.g. for applications that require desktop access) is provided at analysis-01.compute.fhi.mpg.de.
This system can be used for compute-intensive tasks and resource-intensive software - while still having the 'usual' desktop experience.
This service requires you to register by Email (beinlich@fhi-berlin.mpg.de) so that we can send a personalized link for your (persistent) desktop session.
Please have a look at Analysis - Virtual Desktop Server!
GPU - GPU Computing Server
We are currently setting up a compute server for GPU accelerated computing (with NVIDIA Tesla A100 GPUs). It will be available on request and use a similar setup as the Virtual Desktop Server.
Storage - NFS / SMB Server
The computing servers above share a common storage space based on NFS and are accessible via Windows Network Shares (SMB) from FHI Desktop systems. For more details please visit: Storage - NFS / SMB Server !
Status Monitor
The status and current usage of the compute systems can be monitored here: Monitor - Compute Status Monitoring !
Former Computing Services at the FHI
Some of the following former services will be replaced by the ones above.
The Littlest Jupyter Hub TLJH
For new and current users: please switch to the new Jupyterhub above. We run a small JupyterHub Server at tljh.fhi.mpg.de . Please have a look at Jupyter Hub TLJH!
HPC Systems (Supercomputer) at the FHI
Please subscribe to the FHI-HPC-Users Mailing list for important information on the FHI HPC systems and maintenance notes. This is a newsletter mailing list, not a public one, so you will only be notified for important messages!
The former Q-cluster is decomissioned end of 2024. FHI members can use the compute services above or can request access to the MPCDF VIPER supercomputer
HPC Systems at the Max Planck Computing and Data Facility (MPCDF)
The central computing resources at FHI will be supplemented in future by various computing services provided by the Max Planck Computing and Data Facility (MPCDF). These resources are independent of any existing dedicated MPCDF resources of the departments.
Members of the FHI can use the Viper Supercomputer of the MPCDF. For getting access, please contact us, and we will support you in deciding how to run your jobs there and get an account. The user documentation for Viper can be found here: Viper User Documentation. Please get in touch if you need support using these systems.
Access to the last generation MPCDF supercomputers (Cobra, Raven currently is only provided at department level.
Former HPC Systems at the FHI
Q Cluster
The Fritz Haber Institute has its own compute cluster 'Q' which is operated by the PP&B and accessible to all members of the FHI. It will be replaced soon by a new system.
xFHIX and yFHIX Cluster
The previous generations of computer clusters ('yFHIX' and 'xFHIX'), are no longer available. However, some of the hardware is still accessible via the 'Q' cluster in a separate scheduler queue 'yfhix.q' (see compute cluster 'Q').