>> On Sat, 8 Sep 2007 09:26:40 +1000, "Pahari, Dinesh P" <dinesh.pahari AT EDS 
>> DOT COM> said:


> One of the restore I am doing is taking 3hrs to do just about half
> of GB. I understand that it had to go through about 50 different
> tapes to get that data, but I would still imagine that it should
> have been faster than that.

2-3 minutes mount and seek latency * 50?  I think you're doing about
right.  This is why you want to collocate to whatever extent is
possible.

You might be able to make more efficient use of multiple tape drives
by doing, for instance:

move nodedata [node] [fs] fromstg=tapes-place tostg=disk-place 
maxproc=number-of-drives.

that could take some time off your restore, if you've got enough disk,
and enough drives to counteract the fact that you can't start the
restore until the move completes.

- Allen S. Rout