The "incremental forever" paradigm, Was: How would you do this?

Curtis J. Pogue wrote:
>
> My company wants to do a full backup every month and incrementals in
> between.  I've argued with them about not needing full backups but
> they just say "That's the way we've always done it!".  Any way, I
> going to either use the absolute value in the backup copy group or
> specify selective in the backup schedule.  My question is should I
> create a separate policy domain and every thing that goes with it for
> the monthly backups?
>
> TIA
Curtis,
        I waited until replies to your post stopped trickling in before
I posted.
        My problem with the "incremental forever" paradigm is that
it assumes that once you write a file to a disk or tape you can just
forget about it, the backup has been done.  There is no need to ever
back up that same file twice, even years later.
        That just seems completely ridiculous.
        What about a bad tape?  Have none of you ever had that happen?
What about a operator loosing a tape, or the tape being sent offsite
and being lost by the offsite provider?  Or a tape being dropped by
the operator and the case broken.  This is reality.  No media is
impervious to problems, that's why we back up to begin with.
        And the problem doesn't go away if you use disk instead of tape
either.  Disks fail, too, and at the most inopportune moments.  And we
do not have the disk space available to support disk pools, anyway.
        It just seems too optimistic to presume that one copy of all your
critical files would be enough, no matter what.
        Also, the idea of backing up a storage pool to a copy storage
pool sounds good to me, but seems to be exceedingly slow.  We have
about 51GB worth of active data, and after I did some measurements it
looks like it would take over 10 hours of dedicated time to do the
tape-to-tape copy.  It would be easier if we could do that in 1 or 2
hour chunks spread out over a week.  I have looked in to using "export
node" to do this, but that is not really a good substitute, since you
would have to reimport the entire node again to restore anything.
        Please explain to this novice why, on the one hand, we
implement a backup strategy to cover the possibility of loosing disks,
be we ignore the possibility of the backup media itself going bad?

Best Regards,
John Schneider

***********************************************************************
* John D. Schneider     * Email: jdschn AT ibm DOT net * Phone: 314-349-4556 *
* Lowery Systems, Inc.  ***********************************************
* 275 Axminister        * Disclaimer: Opinions expressed here are mine*
* Fenton, MO 63026      * and mine alone.  My company is off the hook.*
***********************************************************************