Data Storage

SCI-UC provides and support, for all HPC clusters and computing infrastructure, several file systems optimized for different purposes. Each user has access to a home directory, a personal working area, and a shared data lake.

PATH File System Size Quota Purpose
/nfs/home/$GROUP/$USER NFS 120G Personal home directory (configuration files, scripts, source code, ...)
/nfs/software/$GROUP NFS Group-shared software directory (apps and libraries)
/nfs/shared/$GROUP NFS Group-shared directory (data/file sharing)
/lustre/$GROUP/WORK/$USER LUSTRE 246T 10T bytes / 1M files Personal working storage
/lustre/$GROUP/DATA LUSTRE 313T 10T bytes / 1M files Common and supervised data lake
/scratch/$GROUP LOCAL 1T <> Common and supervised data lake

User HOME Storage [NFS]

The home directory is located on an NFS file system and is accessible from all nodes of the cluster.

/nfs/home/$GROUP/$USER

It is designed for:

  • Storing configuration files (.bashrc, .ssh/, etc.)
  • Small scripts and source code
  • Job submission files

Avoid storing large data files or running I/O-intensive operations from your home directory. Use your Personal Working Storage directory instead.

/nfs/software/$GROUP/

/nfs/shared/$GROUP/

HPC Storage [LUSTRE]

Personal Working

Each user has a WORK directory on the Lustre file system:

/lustre/$GROUP/WORK/$USER

This space is:

  • Accessible from all cluster nodes
  • Intended for temporary and working data
  • Optimized for high-performance parallel I/O

Each user has a storage quota and a file count limit. You can check your current usage with:

[user@ui ~]$ lfs quota -hu $USER /lustre/$GROUP/WORK
Disk quotas for usr usuario (uid 15999):
     Filesystem    used   quota   limit   grace   files   quota   limit   grace
/lustre/geocean/WORK
                 1.197T     10T     15T       -   27149  1024000 1228800       -
  • used: Current usage
  • quota: Soft limit (you will be warned when exceeded)
  • limit: Hard limit (cannot be exceeded)
  • files: Number of inodes used

Shared Data Lake

Each research group also has a shared directory:

/lustre/$GROUP/DATA

This area is intended for:

  • Long-term datasets shared among group members
  • Common input/output data used in multiple projects
  • Data supervised by the group’s PI or data manager

Access is group-wide, and usage is typically monitored by the administrators.

The DATA space is not intended for scratch computations or temporary job files — use your Personal Working Storage area for that.

Scratch Storage [LOCAL]

Each compute node provides temporary local (NVMe/SSD) scratch storage, accessible at /scratch once a Slurm job has started on that node. This storage is intended exclusively for temporary data generated by jobs running through the queueing system on the allocated node. Using local scratch space can significantly improve job performance, particularly for I/O-intensive workloads.

Please note: - Data stored in /scratch resides only on the local compute node where the job is running. - This storage is not shared across nodes. - No backups are performed on this space. - Data may be removed at any time due to system or service requirements.

Users are responsible for transferring any required output data to persistent storage before their job completes.

Snapshotting and File Recovery Policy (Lustre and NFS)

To protect user data against accidental deletion or modification, all cluster file systems are protected by a daily snapshot policy.

Snapshot Policy

  • Snapshots are taken once per day on all file systems:

    • /nfs/home/$GROUP/$USER/ *lost data/files recovery not available yet
    • /nfs/software/$GROUP/ *lost data/files recovery not available yet
    • /nfs/shared/$GROUP/ *lost data/files recovery not available yet
    • /lustre/$GROUP/WORK/$USER/
    • /lustre/$GROUP/DATA/

  • Snapshots are retained for 10 days

  • After 10 days, snapshots are permanently deleted

This means that files deleted or modified within the last 10 days may be recoverable.

File Recovery Procedure

  1. Identify:

    • The exact file or directory path
    • The approximate date when the file was last known to be correct**

  2. Send a recovery request by email to:

Important Notes

  • Recovery is only possible within the 10-day retention window
  • Restoration is performed by system administrators
  • Recovery time may vary depending on system load
  • Snapshots are not a substitute for proper data management practices
  • Users are responsible for backing up critical data externally if long-term retention is required