I'm considering changing the way I do my TSM DB backups; I've heard some user
experiences about backing up to remote server volumes, and it appears that
this would simplify a lot of work, and save oodles of space.


Here's how I've got it sketched in my mind right now:


A server instance whose sole purpose is to store backup volumes.

Disk sufficient to hold a day's fulls from all my other servers.

Tapes.  (perhaps two or three copy pools?)

Each other server runs a full every day, I make N copies of them, and shunt
them off to tapes.

The "TSM DB Backup Server" ends up hosting a few thousand files at most.  A
full backup of that database is microscopic; megabytes.  I make one every day
to a FILE device class.

The directory in which that FILE devclass is stored gets rsynced all the *#&@
over the place on an hourly basis (overkill: it gets written to once a day.
But who cares?  it's a directory with a few dozen 10MB files in it.)



This way I avoid using N 3590 tapes (where N is my number of servers, and
certainly > 4 ) every day when I'm writing a -total- of well under 20GB of
compressed data.


Opinions?  Gaping holes I've missed?


- Allen S. Rout