[gmx-users] Backing up trajectories

[Esztermann, Ansgar]

Hi everyone,

I am just wondering how backups of Gromacs trajectories are handled "out there". 
We are using TSM. The system is very reliable, but it does not scale well in conjunction with Gromacs: trajectories can get very large when -append is used (dozens of Gigabytes), but TSM cannot do block-incremental backups (at least on Linux).
So every time a trajectory is backed up, the whole xx GB are copied to the backup server. There are mechanisms in place to limit .trr/.xtc backups to once a week or so, but for runs lasting several months, this still means a huge amount of wasted network bandwidth and server disk space.
Since computational power increases faster than network bandwidth does, this is bound to become a bottleneck sooner or later.

Is anyone willing to share how backups are handled at your site?


Thanks,

A.

-- 
Ansgar Esztermann
DV-Systemadministration
Max-Planck-Institut für biophysikalische Chemie, Abteilung 105