At the faculty of science, C&CZ manages a number of compute clusters and we have our own computer room with a few dozen racks for server hardware and also some racks in the Forum computer room.
Managed clusters
The largest cn-cluster is heterogeneous: it holds nodes purchased by individual
departments as well as a number of nodes purchased by C&CZ, which FNWI departments
may use after approval. The cn-cluster includes nodes with a GPU with CUDA cores.
All nodes use SLURM as cluster software; regular maintenance is
performed according to the Compute Cluster Maintenance
policy.
In addition to the cn-cluster, we manage:
- the Astrophysics coma cluster (1048 cores, 8780 GB RAM, 600 TB fileserver storage, GPUs), a separate cluster;
- a separate cluster for the RU department Centre for Language Studies.
Sometimes outside of opening hours the PCs in the computer labs are also used as a compute cluster.
Financial model
The faculty in general doesn’t own (m)any compute nodes, departments purchase their own servers (we can help configure these), which we install and manage during their lifetime. All clusternodes can be integrated in our SLURM cluster management. Maintenance for these clusters is described in the Compute Cluster Maintenance guide.
For departments of FNWI, housing and power is “free”, because we have currently no way to determine what to charge for it. This may change in the future as our monitoring equipment improves and policies change to include charging for power.
External customers have custom financial agreements.
A primitive overview of clusternodes is available, until we can fix something better to replace the old cricket overview.
Other options
…