Dar used within the Max Planck Institute

Who: Max Planck Institute for Gravitational Physics
Context: The Atlas computing cluster (HPC)
Start date: Year 2021
Backup/archive size: from a few 100GB up to 45TByte (compressed) so far - all in 1 TByte chunks with isolated catalogue
Media used: locally HDD (NAS/ZFS), remotely tape LTO8/9 I think at the moment
Why dar: before using regular old tar/star but seqential only access is prohibitively expensive when only need some files from very few chunks in the middle
other remarks: only downside right now is that it seems to be very slow when generating archive/compressing very large directory structures with a LOT of files. But so far, no real time to investigate on my part, just running something like: dar --create PROJECT --verbose=messages \ --fs-root /srv/00transfer/PROJECT --alter=binary\ --slice=1T --min-digits=4 \ --compression=zstd:20:1M --multi-thread=12 \ --slice-mode=0600 --hash=sha512
Source: Carsten Aulbert in December 2025