Skip to content

dataLAB login screen

Slurm GPU Cluster

Access to the Cluster

Access to the Cluster is available to all Researchers and Professors at the TU Wien. Students writing on a Masters Thesis are also elegible. Exceptions to this limitations are possible but need confirmation from a Supervisor.

To gain access to our Slurm Cluster please complete this Tuwel course, to learn the basics of how to work with our cluster and Slurm in general:

dataLAB Cluster Essentials password: dataLABW2025

Data and user retention

Users that have not logged in within a year will be deactivated and their data transfered into the archival storage. After another 3 Months of inactivity the user and his data will be permanently deleted. Users affected by this policy will be informed of the deactivation and deletion of their account via their stored mail address.

Information about the Cluster

Cluster Nodes

Compute Nodes

Hostname CPU Type CPUs Cores/CPU Threads GPU Type Bus Count GPU Mem IB
a-a100-o-1 AMD 2 32 2 a100 SXM4 8 80GB YES
a-a100-o-2 AMD 2 64 2 a100 SXM4 8 80GB YES
a-a100-os-3 AMD 2 64 2 a100s SXM4 8 40GB YES
a-a100-os-4 AMD 2 64 2 a100s SXM4 8 40GB YES
a-a100-q-5 AMD 2 64 2 a100 SXM4 4 80GB YES
a-a100-q-6 AMD 2 32 2 a100 SXM4 4 80GB YES
a-a100-qs-7 AMD 2 64 2 a100s SXM4 4 40GB YES
a-a100-qs-8 AMD 2 64 2 a100s SXM4 4 40GB YES
a-a40-o-1 AMD 2 24 2 a40 PCIe 4 8 48GB NO
a-l40s-o-1 AMD 2 32 2 L40S PCIe 5 8 48GB NO
a-l40s-o-2 AMD 2 32 2 L40S PCIe 5 8 48GB NO

DGX Systems

Hostname CPU Type CPUs Cores/CPU Threads GPU Type Bus Count GPU Mem IB
dgx-h100-1 Intel 2 56 2 h100 SXM5 8 80GB YES
dgx-h100-2 Intel 2 56 2 h100 SXM5 8 80GB YES
dgx-h100-3 Intel 2 56 2 h100 SXM5 8 80GB YES

Virtual Machines

Hostname CPU Type CPUs Cores/CPU Threads GPU Type Bus Count GPU Mem IB
ivm-a40-q-2 Intel 2 10 2 a40 PCIe 4 4 48GB NO
ivm-a40-q-3 Intel 2 10 2 a40 PCIe 4 4 48GB NO
avm-a40-o-4 AMD 2 40 1 a40 PCIe 4 8 48GB NO
avm-v100-d-5 AMD 2 8 1 v100 PCIe 4 2 32GB NO
avm-a100-qs-9 AMD 2 34 1 a100s PCIe 4 4 40GB NO
avm-a100-d-10 AMD 2 18 1 a100 PCIe 4 2 80GB NO

Storage

The main Storage used is Ceph as a network file system that includes /home and /share on all servers

  • /home for user's home directory
  • Limited to 250GB
  • /share to have groups directory for collaborations (please reach us for special groups and permissions)
  • No limit. Please try to avoid duplicate datasets and instead put them in some of the public folder like "/share/models".
  • /scratch is a local nvme storage RAID 0 it exist only on GPU nodes - they are mounted on head nodes as NFS as well.
  • The scratch storage is the fastest option. Please put any files created during a job there and move them to /home or /share afterwards. Copying datasets to the scratch before execution can speed things up for longer jobs.
  • /archive is a HDD based Ceph storage meant for currently unused files. Please move any currently not required files to this place.
  • No limit. Please try to avoid duplicate datasets and instead put them in some of the public folder like "/share/models".

Partitions (queues)

  • GPU-v100: i-v100-o-1,i-v100-h-2,i-v100-q-[3-4]
  • GPU-a40: a-a40-o-1
  • GPU-a100: a-a100-o-[1-2],a-a100-q-[5-6]
  • GPU-a100s: a-a100-h-9,a-a100-os-[3-4],a-a100-qs-[7-8]

Default parameters

There are default parameters set it is different to each partition for example:

  • Max Nodes per job are 2.
  • Max job time is 7 days.
  • Default time for any job is 24 hours, to overwrite this value you should set in your script the time limit option --time which can be maximum of 168 hours.
  • Default MEM per CPU, default CPU per GPU, default memory per GPU, these can vary between partitions but you can overwrite them in your script.

To facilitate creating your Slurm bash script parameters, you may use this code generator platform and also check the command sbatch man pages for more information.