The cluster is a network of machines with distinct roles, not a single computer.

Head Node (mibgard)

Purpose: This is the entry point of the cluster. Use it to log in, edit code/scripts, manage and view your files, submit jobs, and perform other light interactive activities.

Restriction: Do not run analysis or heavy computations here. Jobs running on the head node may be terminated immediately to preserve resources for other users.

Hardware: 128GB RAM, 8 cores 16 threads (Intel Xeon D-2146NT 3.0GHz)

Disks: local scratch 2TB

Interactive Node

Purpose: This node is used for interactive jobs and serves as the web server for tools such as Jupyter Notebook, RStudio, and other web-based applications.

Restriction: Direct login is not permitted. Jobs are submitted through the web interface only. This node is intended for testing and short interactive sessions, not for production workloads.

Hardware: 128GB RAM, 8 cores 16 threads (Intel Xeon D-2146NT 3.0GHz)

Disks: local scratch 2TB

Compute Nodes

Purpose: These machines perform the actual data analysis. Users cannot log in directly; SLURM jobs are submitted via the head node. It is also possible to request an interactive session for testing via slurm.

Restriction: Dedicated to slurm jobs. Direct logins are not permitted.

Hardware: 1x Standard Nodes: 380GB RAM each, 96 cores 192threads (AMD EPYC 9654P @ 3.71GHz) 1x High-Memory Node: 1 TB (1024 GB) RAM, 96 cores 192threads (AMD EPYC 9654P @ 3.71GHz)

Disks: local scratch 2TB