Storage Layout

Directory Structure

The following paths are the same on both pasteur-login and the compute node (pasteur):

Quotas

Each regular lab user has a combined quota of 70 GB across their home, scratch, and apps directories. This is enforced at the filesystem level.

Check your current usage:

$ quota -s
Disk quotas for user jsmith (uid 1002):
  Filesystem   space   quota   limit   grace   files   quota   limit   grace
pasteur:/data   12G    70G    75Gpasteur-login

Home Directory — /data/home/<username>

Your personal home directory. This is where your shell config files, scripts, and important outputs should live. It is accessible from both pasteur-login and the compute node.

• Quota: Counts toward your 70 GB combined quota (shared with scratch and apps/<user>)
• Backed up: No — store important data here but be aware there is no automated backup
• Accessible from: pasteur-login and pasteur (compute)
• Good for: Shell configs, job scripts, small outputs, code repositories

Scratch — /data/scratch/<group>/<username>

High-throughput temporary workspace intended for active job I/O. Use scratch for data that a running job reads and writes — it counts against your 70 GB combined quota.

• Quota: Counts toward your 70 GB combined limit
• Backed up: No — not backed up at all
• Accessible from: pasteur-login and pasteur (compute)
• Good for: Intermediate job outputs, large input datasets during an active experiment, model training checkpoints in progress

#!/bin/bash
#SBATCH --job-name=my_experiment

GROUP=$(id -gn)
SCRATCH=/data/scratch/$GROUP/$USER
mkdir -p $SCRATCH/run_${SLURM_JOB_ID}

# Copy input data to scratch
cp /data/shared/projects/myproject/input.tar.gz $SCRATCH/run_${SLURM_JOB_ID}/

# ... run your job ...

# Copy important results back to home when done
cp -r $SCRATCH/run_${SLURM_JOB_ID}/results/ $HOME/completed/job.sh

Apps — /data/apps/<group>/

Software installations — primarily conda environments. Your personal area is at /data/apps/<group>/<username>/.

# Create a personal conda environment
$ mamba create --prefix /data/apps/$(id -gn)/$USER/my-env python=3.11

# Activate it
$ conda activate /data/apps/$(id -gn)/$USER/my-env

# In a Slurm job script — always source conda first
source /opt/miniforge3/etc/profile.d/conda.sh
conda activate /data/apps/$(id -gn)/$USER/my-envbash

Shared — /data/shared/

Shared storage for the lab. Access is granted per project by an admin.

• /data/shared/containers/ — shared Apptainer .sif images (admin-managed, read by all lab users)
• /data/shared/projects/ — collaborative research datasets (readable and writable by all lab users)
• /data/shared/archives/ — archived user data (admin access only)

Best Practices