sofia @ VUB-HPC#

General information#

Important

sofia is currently in pilot phase. No starting grants or full project grants will be issued. The VUB-HPC team is working hard to make it production-ready. If all goes well, the cluster will be open for starting grant applications by May 2026, and for full project applications by June 2026.

sofia is the 4th VSC Tier-1 cluster, following muk (hosted by HPC-UGent, 2012-2016), BrENIAC (hosted by HPC-Leuven, 2016-2022) and Hortense (hosted by HPC-Ugent, 2021-2027).

It will be available in the second half of 2026, is hosted by Vrije Universiteit Brussel, and maintained and supported by the VUB-HPC team.

Hardware details#

The sofia cluster consists of the following partitions:

zen5_dense: CPU partition:
- 56 compute nodes, each with:
  
  2x AMD EPYC 9965 192-Core CPU (Turin Dense) 2.25 GHz (“Zen 5c” microarchitecture, 384 cores per node)
  
  770 GiB usable RAM (~2 GiB/core), 7.6 GiB swap (local SSD)
  
  1.4 TiB SSD local disk
zen5_himem: large-memory partition:
- 16 compute nodes, each with:
  
  2x AMD EPYC 9655 96-Core CPU (Turin) 2.6 GHz (“Zen 5” microarchitecture, 192 cores per node)
  
  1.5 TiB usable RAM (~8 GiB/core), 7.6 GiB swap (local SSD)
  
  1.4 TiB SSD local disk
zen4_h200: GPU partition:
- 22 compute nodes, each with:
  
  2x AMD EPYC 9654 96-Core CPU (Genoa) 2.4 GHz (192 cores per node)
  
  8x NVIDIA H200-SXM5 (141 GiB GPU memory), Nvlink4
  
  1.5 TiB usable RAM (~8 GiB/CPU core), 7.6 GiB swap (local SSD)
  
  1.4 TiB SSD local disk
zen5_vis: interactive and debug partition:
- 4 compute nodes, each with:
  
  2x AMD EPYC 9655 96-Core CPU (Turin) 2.6 GHz (“Zen 5” microarchitecture, 192 cores per node)
  
  2x NVIDIA RTX 5000 Ada Generation (32 GiB GPU memory)
  
  1.5 TiB usable RAM (~8 GiB/CPU core), 7.6 GiB swap (local SSD)
  
  1.4 TiB SSD local disk

Shared infrastructure:

storage: 4.3 PiB shared scratch storage, IBM Storage Scale System
interconnect: NVIDIA Quantum-2 InfiniBand, fat tree network with 2:1 blocking
- ~400 Gb/s for GPU and large-memory partition
- ~200 Gb/s for CPU partition

Getting access#

Access policy#

The sofia VSC Tier-1 cluster can only be accessed by people with an active Tier-1 compute project.

Tier-1 compute project resources are always allocated for a given period. After this period expires, the users of this project lose access to all resources granted within that project. This includes storage in addition to compute resources (CPU and GPU).

As soon as a project expires, project members:

will no longer be able to use any remaining CPU cycles, GPU cycles or credits
will no longer have access to the dedicated project folders
will lose access to the Tier-1 cluster, unless they have another active project

See https://www.vscentrum.be/compute for more information on requesting access, rules and regulations.

Login nodes (SSH)#

You can use SSH to connect to the login nodes of the Tier-1 sofia cluster with your VSC account. From within the Flemish university network, use your VSC-id to connect to sofia.hpc.vub.be.

Note

If you are not inside the Flemish university network you can ssh into sofia from the login nodes of any Tier-2 cluster. Do not forget to set up your SSH keys there or to forward your SSH agent.

More general information about SSH login is available in the Access: Terminal Interface section. Note that for access from abroad an extra step is needed. See the VUB tab at Location Access Restrictions.

There are 2 login nodes for sofia: login01 and login02, both running the Rocky Linux release 9.7 operating system. When logging in using SSH, you will be assigned to either of these login nodes, based on the IP address of the host you are connecting from.

If you need to access a specific login node (for example because you have a screen or tmux session running there), just run “ssh login01” to jump to login01 if you were logged in to login02, or vice versa. Make sure your SSH keys are properly configured or forwarded.

Note

The available resources on the sofia login nodes are shared by everyone who is logged in there.

Please only use the sofia login nodes as an access portal!

For resource-intensive interactive tasks, like software compilation, testing software or job scripts, etc., please use an interactive job, preferentially on the zen5_vis partition, either via salloc -p zen5_vis or in the future through the Web portal.

Host keys#

The first time you log in to the sofia login nodes, a fingerprint of the host key will be shown. Before confirming the connection, verify the correctness of the host key, to ensure you are connecting to the correct system.

The type of fingerprint that will be shown depends on the version and configuration of your SSH client.

Please verify that the fingerprint of the host key is *one* of the following:

for ECDSA host key:
- 78:2f:bd:30:95:7f:ba:eb:c4:bf:29:71:fd:1f:2b:b7 (MD5)
- Trw7Gwirs3k9QdH2n2fB9NGrNKPwG8mVK3Yfy8VPc2Y (SHA256)
for ED25519 host key:
- bf:4d:d3:55:ce:72:cd:dd:a7:bf:44:62:97:5f:22:2d (MD5)
- IPZeanj7AVbBYFyc1KBFfd+hMxyDbIGyOghn2uQkuww (SHA256)
for RSA host key:
- 5e:2e:21:11:c6:8e:1e:1f:c9:7c:d2:38:c7:f6:f9:35 (MD5)
- vS9fAoDO52dWCXS/obPS5u95irhJiKo1PfV4dyJl2Mg (SHA256)

Web portal#

The sofia web portal is not yet be available during the pilot phase.

More information about the usage of the web portal is available at Access: Web Portal.

Storage#

Similar to Hortense, the user’s $HOME directory is located on the scratch file system and is distinct from the user’s $VSC_HOME. The advantage of this setup is that sofia remains accessible even if the user’s home institution cluster is down.

The Tier-2 Storage Infrastructure, specifically $VSC_HOME and $VSC_DATA, is accessible, but only on the login nodes, and only during the pilot phase. These will be removed when the pilot phase concludes. We highly recommend using Globus for file transfer between sofia and Tier-2 storage.

sofia scratch via Globus#

Accessing the sofia scratch via Globus is not yet possible during the pilot phase.

More general information about Globus is available at Globus data sharing platform.

Getting help#

For questions and problems related to Tier-1 sofia, please contact the central support address for Tier-1 support: vsc-support@hpc.vub.be.

Questions regarding Tier-1 Hortense will need to go to: compute@vscentrum.be.