Data Transfer Best Practices

From HPCwiki
Jump to navigation Jump to search

Guidance for moving data to and from Anunna efficiently and reliably. For the tools themselves, see Data Transfer Methods.

Choose the right tool

  • For command-line transfers, prefer rsync over scp for anything large or repeated — it resumes and only copies what has changed.
  • The browser-based File Browser is fine for small files, but not for large or many-file transfers; use rsync, scp, or an SFTP client (WinSCP, FileZilla) instead.

Transfer to the right place

  • Transfer compute data directly to Lustre, not to your home directory.
  • Use the correct group path under /lustre — see Compute Storage.

Make transfers reliable

  • For long transfers, run them inside a persistent session (screen or tmux) so they survive a dropped connection.
  • Verify that large transfers completed before deleting the source — rsync reports this, or compare checksums.
  • Resume an interrupted transfer with rsync rather than starting over.

Be considerate

  • Transfers run on the login nodes, which are shared — very large transfers can load them, so be mindful of other users.

See also