The largest restore I have done was 144GB and since Oracle databases
compress real tight there was not that much data actually transferred (Less
than 60 Gig).

In our experience, it all depends on the speed of the client, what we do is
run a script that will submit "N" sessions (1 For each Mount point), it
monitors them and keeps them going as they end. We have about 39 different
Oracle mount points 4GB each, of which, 36 are in use.

The restore of 144GB on our HP V2500 8 CPU ran in just 1.25 hours, there is
a Fiber EMC 3830 unit attached.

The same data restored to a HP T520 12 CPU in about 8 hours. HP T600 12 CPU
6 hours, these have a F/W EMC 3500.

All of these times are from Disk Storage pools. Tape will take longer
depending on your collocation method, I have found that collocation at the
filespace uses more tapes but works better for concurrent restores. Our Tape
Library STK 9710 has 8 DLT 7000 drives.

The Tape Restore (No Collocation) on the T520 actually ran over 24 hours but
most of the data was on a few DLT tapes.

I have learned that when the Developers want to copy the production database
to tape, we do some planning and leave the data on disk longer than we would
normally by turning on storage pool CACHE=YES.

I have even spent some time reloading "MOVE DATA" the tapes back to disk
storage pools so that I could run more of them concurrently.


Bill