Tim> Yepper. On volumes with lots of files, that ndmp history/catalog
Tim> pass can be crazy!

Hah!  We've got a volume with 50 million files and it sucks for NDMP
backups.  Fulls are actually not terrible... but the index pass just
takes forever.

Tim> I once tried to use ndmp backup on a volume with 500,000+
Tim> files. After 8 hours and a huge load (FAS6080) the ndmp gave out
Tim> and quit. Eight hours of building the index and it didn't finish.

This is the probably with *any* file level backup.  Unfortunately,
there's no good solution in terms of price for backing up Netapps that
I'm aware of.

You could buy a cheaper pair of heads with lots and lots of cheap(ish)
SATA storage, but it's still god-awful expensive.  And the snapmirror
licenses aren't cheap either.

In my engineering environments, I've been trying to encourage them to
only backup what they need, and to work in scratch areas instead, but
it's hard to get people to change.

The Netapp is just so reliable that is really does keep data for years
and years without problems.  In more cases.

Heh.

But some $WORK orgs have off-site requirements, and they're not
willing to pay the price for a remote co-lo Netapp and storage and the
bandwidth to make backups work reliably and quickly enough.  So off to
tape it is.  Make's JSOX happy too...