In <bitnet.adsm-l%ADSM-L%96022014513515 AT VM.MARIST DOT EDU>, crockett AT 
vnet.ibm DOT com (Dave Crockett) writes:
>Bill writes:
>> I don't do it anymore on the primary becasue I switched over to
>> collocation. I do it on the copypool tapes because they are not
>> collocated and are offsite.  When that pool is 95% full, I will start
>> rolling it over with 'move data' commands like I used to on the
>> primary.
>
>I just wanted to point out the ramifications of running with your
>primary storage pool with collocation and your copy storage pools
>without collocation (which you have probably already considered).
>The trade off is reducing the number of tapes and  time during backup
>stgpool versus a potentially very lengthy restore in  the event of a
>disaster.
>
>As you are doing incremental backups of your storage pool data using the
>BACKUP STG command, overtime the data for a given client will be spread
>over many volumes in the copy storage pool.  After a disaster, the
>offsite copy storage pool volumes can be brought back onsite and clients
>can immediately start restoring their data.  This of course can involve
>many tape mounts for each client since the copy storage pool isn't
>collocated.
>
>You could instead restore your primary storage pools first, but this
>could involve even more tape mounts if you are restoring directly back
>into a collocated pool.  It will grab a given copy storage pool volume
>and then collocate the data for that volume, involving many mounts.
>Then it will do the same with the next volume.
>
>Probably the best method to use in this mixed collocation environment is
>to restore the storage pool data into a very large disk pool and then
>let migration move the data off to the collocated tape pool in a more
>orderly way.
>
>Dave Crockett - ADSM development
>
>QUIT

I am aware of the issues in not collocating the copypool.  But I wonder
if IBM really has a good solution for me?  Please try to work out how
to manage tapes, both onsite and offsite for a hypothetical 5,000 client
system (is it hypothetical, who out there has the most clients?)

The first thing to say is that IBM make lots more money from an account
that backs up desktop machines instead of just servers.

2nd, if every user backs up at least once a week, at the end of the week
there will be 5,000 tapes in filling status if the copypool is collocating.

Now you can mark these readonly and send them offsite.  For the next week
there will be 5,000 additional tapes to send offsite, and so on and so
forth.  After a year there are 260,000 tapes offsite.  Now how does an
administrator consolidate these tapes?  If they aren't consolidated, you will
have 52 mounts to restore a node, which might be about the same as if the
copypool wasn't collocated.

Bill Colwell
C. S. Draper Lab
Email: BColwell AT draper DOT com
Voice: 617-258-1550