Magus02/Getting Started

📍 Magus02 Knowledge Base → Basics → Getting Started

This guide walks you through the essential steps to start using the Magus02 HPC cluster — from getting access all the way to running your first batch job.

Access — How to get a Magus02 account

To use Magus02, you need an account provisioned by the HPC Administration Team. Depending on your role:

• Students / Research Scholars: Request an account through your supervisor or PI. The PI submits an access request to the HPC admin team.
• Faculty: Contact the HPC admin team directly at [email protected].
• External collaborators: Require a formal request letter from a SNU faculty sponsor.

Once your account is created, you will receive a welcome email with your username and initial password. You should reset your password on first login.

⚠️ VPN Required for Off-Campus Access: Magus02 login nodes are only directly reachable from within the SNU campus network. If you are working remotely, you must first connect to the SNU VPN. Contact the HPC admin to obtain VPN credentials.

Login — How to connect to Magus02

Magus02 uses SSH (Secure Shell) for all user logins. Connect to the login nodes — not the compute nodes directly.

From Linux / macOS

Open a terminal and run:

$ ssh [email protected]

On first login, you will be prompted to verify the host key fingerprint. Type yes to accept.

From Windows

Use one of the following free SSH clients:

Client	Description	Download
PuTTY	Classic SSH client, widely used	putty.org
MobaXterm	SSH + X11 + file browser in one	mobaxterm.mobatek.net
Windows Terminal / cmd	Built-in since Windows 10 1709 — just run ssh	Built-in

Enter magus02.snu.edu.in as the hostname and port 22.

For graphical applications (X11)

If you need to run GUI tools remotely, enable X11 forwarding:

$ ssh -X [email protected]

See the full SSH guide for more details including key-based authentication.

File Transfer — Moving data onto / off the cluster

You cannot run jobs from your local machine — files must be on the cluster's filesystem. Use one of these methods:

• scp — for individual files (built into Linux/macOS/Windows 10+)
• rsync — for syncing directories efficiently
• WinSCP or FileZilla — graphical clients for Windows users

Quick example with scp:

# Upload a file to your home directory on Magus02:
$ scp myinput.dat [email protected]:~/

# Download results back to your local machine:
$ scp [email protected]:~/results/output.log ./

See the full File Transfer guide for rsync usage and graphical client setup.

Scheduler — How to run programs on the cluster

On Magus02, you cannot run compute-intensive programs directly on the login nodes. All production work must be submitted as batch jobs to the Slurm scheduler, which manages the queue and allocates resources on compute nodes.

The basic workflow is:

1. Write a jobscript — a shell script that specifies resource requirements and the commands to run
2. Submit it with sbatch
3. Monitor with squeue
4. Retrieve results from your output files

A minimal Slurm jobscript looks like:

#!/bin/bash
#SBATCH --job-name=my_first_job
#SBATCH --partition=small
#SBATCH --nodes=1
#SBATCH --ntasks=8
#SBATCH --mem=32G
#SBATCH --time=01:00:00
#SBATCH --output=job_%j.out
#SBATCH --error=job_%j.err

# Load required modules
module purge
module load intel/oneapi mpi/intelmpi

# Run your program
./my_program input.dat

Submit it with:

$ sbatch my_jobscript.sh
Submitted batch job 12345

→ See Scheduling Basics and Batch Scheduler for full details on queues, job types, and resource requests.

Modules — Using pre-installed software

Magus02 uses the Environment Modules system to manage software. Rather than installing everything yourself, you load the version you need:

$ module avail               # list all available software
$ module load intel/oneapi   # load Intel compilers
$ module list                # check what is currently loaded
$ module unload intel/oneapi # unload a module
$ module purge               # unload everything

→ See Environment Modules for full documentation.

Parallel Programming — Using multiple cores

Most HPC workloads are designed to run in parallel across many CPU cores or nodes. Magus02 supports two primary paradigms:

• OpenMP — shared-memory parallelism within a single node (multithreading)
• MPI — distributed-memory parallelism across multiple nodes (message passing)

Hybrid jobs using both OpenMP and MPI simultaneously are also supported.

Useful Starting Points

• Linux Shell Quick Reference — basic commands you'll use every day
• FAQ: Batch Jobs — troubleshooting common Slurm issues
• Magus02 Main Page — full cluster specs, queue structure, and applications

💡 Tip: Always use module purge at the start of your jobscripts to ensure a clean environment. Jobs that rely on modules loaded interactively can fail unpredictably. See FAQ: Batch Jobs for more.

---

↑ Back to Knowledge Base · Maintained by HPC Administration · wiki.snu.edu.in