All the environments I've been in typically use a two hop approach, backups are written to disk and the backups are then swept to tape which are taken offsite. If you have a LOT of servers this can be problematic as the network becomes a bottleneck.

And as always you have to periodically TEST that the tapes are readable, at a former employer we had to spend like $15K to send a couple tapes out for recovery when there was a disaster and a server died. They hadn't been testing and just assumed that if it was on tape it was OK. Altogether I think for the better. But 3 days of downtime, a bunch of idle people, and $15K for recovery, I would guess the costs were over $100K..

This was a HARD lesson and there was real change after to make sure the backups were good.

CEWII