Getting Started with the Morehouse Supercomputing Facility (MSCF)

Welcome to MSCF! While we build out our own supercomputing infrastructure, we are using resources at the Texas Advanced Computing Center (TACC) to provide HPC access to our researchers. MSCF is not hosted at TACC — TACC is a partner providing compute time while our facility is under development.

This guide walks you through everything you need to get set up on TACC and run your first job.

Step 1: Create Your TACC Account

After you submit the access request form, the allocation manager will add you to the project. You’ll receive an email from TACC with the subject line:

“TACC Project Invitation Action Required: Account Request”

Here’s what to do:

  1. Open the invitation email and click the link to create your account
  2. Fill out the registration form:
    • Use your institutional email (e.g., @morehouse.edu)
    • Choose a username you’ll remember — this is what you’ll use to log in
    • Set a strong password
  3. Complete the account verification if prompted

Important: You must use the link in the invitation email — do not go to the TACC portal separately. The invitation link connects your new account to the project allocation automatically.

Step 2: Set Up Multi-Factor Authentication (MFA)

TACC requires MFA for all logins. Set this up before trying to SSH in.

Do NOT use SMS/text messages for MFA. Use an authenticator app instead.

There are two ways to get to the MFA setup page:

Option A: Go directly to TACC Accounts at accounts.tacc.utexas.edu and click Multi-factor Auth in the sidebar.

TACC Accounts sidebar — Multi-factor Auth

Option B: Log in to the TACC User Portal at portal.tacc.utexas.edu, click Manage Account in the sidebar, then click Manage Multi-factor Authentication.

TACC User Portal — Manage Account

Once you’re on the MFA page, set up an authenticator app — here are our recommended options:

  1. Follow the on-screen instructions to pair your device
  2. Test it by logging out and back in to the portal

Step 3: Log In via SSH

Once your account is active and MFA is set up, connect to the system.

On Mac or Linux (Terminal)

Open your terminal and run:

ssh your_username@system_hostname.tacc.utexas.edu

The part after @ is the hostname of the system you want to access, followed by .tacc.utexas.edu. Replace it with the correct one from the table below. For example, to connect to Lonestar6:

ssh your_username@ls6.tacc.utexas.edu

When prompted:

  1. Enter your TACC password
  2. Enter your MFA token (from your authenticator app)

Note: When you type your password and MFA token, nothing will appear on the screen — no characters, no asterisks, no dots. This is normal. The terminal is still receiving your input. Just type carefully and press Enter.

On Windows

Use one of these SSH clients:

System Hostnames

System Hostname Details
Lonestar6 ls6 System specs
Frontera frontera System specs
Stampede3 stampede3 System specs
Vista vista System specs

Not sure which system to use? Check the system specs linked above to see which one fits your workload.

Step 4: Understand the File System

TACC systems have three main storage areas. Know which to use:

Location Path Purpose Quota Backed Up? Purged?
$HOME /home1/0xxxx/username Config files, scripts ~10 GB Yes — backed up regularly No
$WORK /work2/0xxxx/username Code, libraries, datasets ~1 TB No — not backed up No
$SCRATCH /scratch/0xxxx/username Active job I/O, temp files No limit No — not backed up Yes — files unused for 10 days

Important: $SCRATCH is not permanent storage. Files that have not been accessed in 10 days are automatically purged. Do not store anything there that you can’t afford to lose. $WORK is also not backed up, so keep your own copies of critical data.

Rule of thumb:

Step 5: Load Software Modules

TACC uses the Lmod module system to manage software. Nothing is loaded by default.

# See what's available
module avail

# Search for specific software
module spider python

# Load a module
module load python3/3.11

# See what you have loaded
module list

# Unload a module
module unload python3

# Reset to defaults
module reset

Common modules you might need:

module load python3       # Python
module load gcc           # GNU compilers
module load cuda          # GPU computing (if applicable)
module load intel         # Intel compilers
module load mvapich2      # MPI for parallel computing

Step 6: Submit Your First Job

TACC uses the Slurm workload manager. You do NOT run compute-heavy work on the login node — you submit it as a job.

Create a Job Script

Create a file called my_first_job.sh:

#!/bin/bash
#SBATCH -J my_test_job          # Job name
#SBATCH -o my_test_job.%j.out   # Output file (%j = job ID)
#SBATCH -e my_test_job.%j.err   # Error file
#SBATCH -p normal               # Queue/partition (normal, development, gpu, etc.)
#SBATCH -N 1                    # Number of nodes
#SBATCH -n 1                    # Number of tasks
#SBATCH -t 00:10:00             # Wall clock time (HH:MM:SS)
#SBATCH -A YOUR_ALLOCATION       # Allocation/account name — provided by the allocation manager

echo "Job started at $(date)"
echo "Running on node: $(hostname)"
echo "Hello from TACC!"

# Replace the lines below with your actual work
module load python3
python3 -c "print('Python is working on TACC!')"

echo "Job finished at $(date)"

Submit the Job

sbatch my_first_job.sh

Monitor Your Job

# Check job status
squeue -u $USER

# Get detailed job info
scontrol show job <job_id>

# Cancel a job
scancel <job_id>

Check Output

Once the job completes, check the output and error files:

cat my_test_job.<job_id>.out
cat my_test_job.<job_id>.err

Step 7: Transfer Files

If your files are hosted online (Google Drive, GitHub, Dropbox, a web server, etc.), you can download them directly to TACC using wget:

cd $WORK
wget https://example.com/link-to-your-file.zip

To download from Google Drive, use the shareable link with wget:

wget -O my_file.zip "https://drive.google.com/uc?export=download&id=YOUR_FILE_ID"

Tip: The YOUR_FILE_ID is the long string of characters in your Google Drive sharing link between /d/ and /view.

Downloading Datasets from the Web

You can use wget to pull publicly available datasets directly onto TACC — no need to download to your laptop first and then re-upload.

# Download a file from any public URL
cd $WORK
wget https://data.example.org/dataset.csv

# Download and rename the file
wget -O my_data.csv https://data.example.org/some-long-filename.csv

# Download an entire directory listing (use with caution)
wget -r -np -nH --cut-dirs=1 https://data.example.org/my-dataset/

This works with any publicly accessible URL — research data repositories, government datasets, course materials, etc. As long as the file has a direct download link, wget can grab it.

Common Slurm Queues

Queue Max Nodes Max Time Use Case
development 4 2 hours Testing and debugging
normal 256 48 hours Standard production jobs
gpu-a100 varies 48 hours GPU workloads (check your system)
large 512+ 48 hours Large-scale parallel jobs

Queue names and limits vary by system. Run sinfo to see available queues on your system.

Quick Reference

Task Command
Log in ssh user@system_hostname.tacc.utexas.edu
Check allocation balance allocations or /usr/local/etc/taccinfo
See available modules module avail
Load Python module load python3
Submit a job sbatch script.sh
Check your jobs squeue -u $USER
Cancel a job scancel <job_id>
Check disk usage du -sh $WORK
Download files to TACC wget https://your-link-here

Getting Help

Checklist Before You Start Computing

Once all boxes are checked, you’re ready to go!