Interesting changes you plan, my son,
the upgrade to an expensive tape technology with 
maybe the best reliability available at all
and downgrade to an cheap OS with arguable lower
availability (at least in the terms of up-time)
than your current OS has.
Anyway, I still am sure this combination 
will work good enough, if set-up correctly,
I am just a bit surprised
an curios about the reason for such a tandem.

I see no or little problem with the migration you 
are planning.
Your 2 servers do share a common media, 
the one of the type "shared disk". As long this
ressource is large enough to hold either
backup or archive data of your largest node,
the usage of device class=file
is a *very* comfortable way to export/import data.

Maybe you could check how problematic would it be to 
attach your old library to the NT server for the time
of migration? In this case you only had to export/import 
server (dump/load), connect old library to the NT,
update drives definition so that they match
new scsi connection,
define additional library with 3590,
define storage pools within 3590, 
and issue "move data" from 
one library to another one.

I strongly suggest migration of one client at time.
This processes are not restartable, so any error leads
to a new start, which can cost a plenty of time lost.

By the way, you can see the task as a part of disaster 
recovery practise as well - and - it will work, do not worry.
At least if you will have read 
parts of administrators reference before the start.
In this sense you may want to use the 8mm drive,
temporarily attached to NT, to "restore stg" to
your NT server, instead of using export/import.
It would be slower, but it would test
your disaster recovery ability.


Good Luck!
Salak Juraj

KEBA AG
Softwareentwicklung Bankautomation
Gewerbepark Urfahr 14 - 16
Postfach 111
A-4041 Linz
Österreich

Tel. ++43/732/7090-7461
Fax ++43/732/730910
e-mail: sal AT keba.co DOT at
www.keba.co.at