Getting Started
This document covers how to connect to the HPC cluster and get oriented with the available resources.
Prerequisites
- You must be connected to the internal network before accessing the cluster
- Your account must be created by the system administrator before you can log in
- Contact the system administrator to request an account
Cluster Overview
The cluster consists of two workstations and a shared NAS storage:
| Node | IP Address | Role | Use For |
|---|---|---|---|
| Control Node (node01) | 192.168.220.75 | SLURM controller + compute | Batch jobs, CPU compute, SLURM submissions |
| Compute Node (node02) | 192.168.220.76 | SLURM compute + Docker | Docker containers, GPU workloads |
| NAS | 192.168.220.80 | Shared storage | Home directories (/home) |
node02 is the dedicated Docker machine. You can SSH directly into node02 and run Docker containers freely without SLURM. See the Docker guide for details.
node01 enforces SLURM for compute jobs. Any non-SLURM process running longer than 5 minutes will be automatically terminated.
Connecting via SSH
Linux / macOS
ssh USERNAME@192.168.220.75
Windows
Use MobaXterm or PuTTY for SSH connections.
Running GUI Applications (X11 Forwarding)
If you need to run graphical applications (e.g., rviz2) remotely, use X11 forwarding:
# Enable server access control on your local machine (first time only)
xhost +
# Connect with X11 forwarding
ssh -X USERNAME@192.168.220.75
# Test X11 forwarding
rviz2
If configured correctly, the GUI application will appear on your local machine while running on the remote server.
For better performance with OpenGL applications like rviz2 or Gazebo, use VirtualGL instead of plain X11 forwarding. See the VirtualGL guide for setup details.
What to Do Next
Once logged in:
- Submit compute jobs through SLURM — see the SLURM Tutorial
- Run Docker containers on node02 — see the Docker guide
- Transfer files to/from the cluster — see the File Transfer guide
SLURM usage is mandatory for compute jobs on node01. Any non-SLURM compute process running longer than 5 minutes will be automatically terminated. Always use srun or sbatch to submit jobs.