Zlatko, I am so tired of hearing about this.  You "hit the nail on the head"
as we say in the USA.  Let me see if I can put some sense of reality into
this once and for all.  In the mainframe days we designed applications to
have cycles of processing where you could roll all the data up and recover
to any point in a financial processing cycle.  Typically, these were monthly
and yearly, but sometimes quarterly and weekly were important for special
sales information.  Somehow, I do not know how, the open systems world has
extrapolated that taking backups on these schedules is the same thing and it
simply is not.  In the mainframe model the application had the data files on
tape named weekly, monthly, quarterly, yearly, and had all the intermediate
files that could recover each of the tapes or created 2 copies of them.  In
any case all of the intermediate transactional data was there because
mainframe integrity was paramount and tapes did fail, but the data was
always recoverable because the application design, not the backup system.
The backup system was for hard drive failures, disaster recovery, and TSO
data files of end users.  End users, we just kept 1 to 13 backup copies,
period, that is all they had.  In the open systems world everything looks
like TSO data so the parameters in the backup product (TSM) had to be
changed to allow for more copies and retention by time in place of or in
extension to the number of copies.

It took me a long time to convince my management they could not recover
anything with any assurance with NetBackup.  They could not tell the
customer what the recovery would be except that it would be a long ways back
if at all.  Now, TSM arrives, and we are well again.  I keep the data
backups for the retention the customer dictates not as the backup product
dictates.  It is possible to achieve with 4 nodenames per node what they are
all talking about here.  But, as you say, "ask them what they want to
restore".  Then, you find out what the business requirement is.

We will eventually pound this into the folks that think
weekly/monthly/quarterly/yearly archives are dumb and provide no sense of
recovery.  The reason it was done is so you had some kind of tape to fall
back to if one was bad, but today, that is so much lost data it is a
disaster.  You have to do your normal cycle and copy it, period.  And, in
our case, copy it for an onsite copy and an offsite copy, because if you
have a disaster while you are rebuilding a failed primary tape from your
offsite copy, you have just lost your disaster recovery set.

Now, the issue of affordability eventually kicks in.  That is where you have
to decide how think the pad is you where on your butt.  Is it 1, 2, 3, etc.
copies.  TSM can do that, but there is a price.  But, as you say there is no
"limitation" with TSM.  You can do it.  It is just not tape retention, it is
object retention.

If we could just get people to think in terms of backup object retention
versus tape retention and understand that paradigm, I think we would be
there.

Take me back to my good old mainframe.

Paul D. Seay, Jr.
Technical Specialist
Naptheon, INC
757-688-8180