An update for you....

I was trying to calculate how long a Disaster Recovery of our clients might
take.
I had some VERY APPROXIMATE rules for how long it would take to restore a
node, to which I added a % "fiddle factor" to allow for tape handling and
DLT load / unload issues. These times were then cross checked to actual
times from a Disaster Recovery week we arranged.


Server is NT (2 x 400MHz, 512 MB memory, lots of disk) with DLT 7000 tape
library. Clients are connected over 100 MBit network links.

Some clients have many small files, others have a few large files - hence
the following rules:
 - Clients with "small" objects are those where "Number of Objects" divided
by MBytes is more than 50.
 - Clients with "large" objects are those where Objects / MBytes is less
than 5.
 - Clients with "medium" objects are the others.

"Small" object clients can be restored at 25,000 objects per hour.
"Medium" objects clients can be restored at either 15,000 objects per hour,
or 1.5 GByte per hour (which ever is the largest).
"Large" objects clients can be restored at 4 GByte per hour.


I added 30% to the expected restore time from these figures. This final
number was then close enough to the actual restores so we are not going to
re-calculate the rules.


Cheers, Chris