I am looking to do some data size forecasting for budgeting purposes. 
So that I know when my Master server is about to run out of horsepower
and when I need a new tape drive.  The horsepower thing is very much a
vendor spec but ties into data size in that, when I am backing up X GB's
to Y tape drives, my server can no longer stream them effectively.  Now
in this master server example it is primarily a function of Y tape
drives but Y is directly related to X gb's.

I know all the specs, 22gb's/hour/DLT7k if streamed nicely, much less if
shoeshining.  Gigabit on Sun requires 100MHz of cpu as do every DLT7k
drive so a 2x400MHz machine should be able to handle 7 DLT7000 drives
with gigabit, so with 7x22gb's/hour I should get 154gb's/hour. 
Regretably nothing is ever so effecient as to increase linearly with
specs.  There's always a catch; too many clients started on this drive
first, another client is slower and hogging a drive, etc....

So basically, unless someone already has a solution to the above, I want
to know if anyone out there already has in place a historical datasize
solution.  Something along the lines of every monthly backup is broken
out into datasize by dataset.

Example output of some script:

Offsite backups performed on 10/31/01 had the following dataset
sizes....

NDMP - 2,450,900 GB's
Unix - 250 GB's
NT/2k - 5 GB's
Novell - 873 GB's

These datasets were backed up in....

NDMP - 23 hours
Unix - 1 hour 30 minutes
NT/2k - 15 minutes
Novell - 14 hours 27 minutes

Given the type and quality of scripts that already exist I'm sure it's
possible to put this together, but if someone has already done it (or
done it better) please let me know.  It would take me months to pull
this together on my own, not much of a programmer...

Thanks.
-- 
=====================
Jeff Kennedy
Unix Administrator
AMCC
jlkennedy AT amcc DOT com