The ADSM design is very different from other backup products for any
platform.  Other products need a cycle of full,  some incrementals, full,
some incrementals, etc.  because the backup system keeps track of the backu
p events, not the individual files.  Thus a full volume restore involves
restoring the backup event entities in order.  The management policy
necessitated by this design is that you can't have to many incrementals bet
ween fulls or else the restore will become to long, or even impossible if
bad media breaks the chain of entities.

ADSM tracks the files that are backed up, not the backup events.  It
implements the concepts of 'active set' and 'inactive set'.  The active set
is made up of the lastest versions of each file in the filespace.  The
inactive set is all prior versions of files in the filespace.  When a new
version of any file is sent to the server, the new version knocks the prior
version out of the active set and into the inactive set, and the new
version becomes part of the active set.

To answer your question about how many versions to keep to guarantee a
restore of a file which is backed up in the first incremental backup and
then never changed, the answer is it doesn't matter!  All the parameters in
the copygroups about number of versions and time to live apply only to the
inactive set and the file in question will stay in the active set.

Read  'define copygroup' in the command reference manual carefully,
especially the 'retextra' parameter.

One feature of other backup systems is that they have a good point in time
restore capability because restore is done by entities.  ADSM needs to
provide a pit feature.  The new client parameters '-(from/to)(date/time)' c
ome close to it, but they allow deleted files to be restored which
shouldn't be.
_________________________Reply Header_________________________
Author: ADSM-L AT vm.marist DOT edu
Subject: Incremental vs Full Backups
05-23-96 05:10 PM

We are in the process of implementing ADSM for the first time in our shop
and I'm trying to clear up some confusion on our part re: backups.  We
come from a mainframe environment where we do weekly 'full' backups of
datasets (backup all files in the shop) and daily 'incrementals' (backup
only files changed since the last backup).  If we need to recover our
system we restore the latest backup of a dataset which is, at most, only a
week old.

In ADSM we see only incremental backups.  Granted the first backup is truly
a 'Full' since nothing has been backed up prior.  If a file does not get
updated periodically it won't be backed up again.  The concern is how long
we must keep backups to guarantee there being an available means to recover
a file that, e.g., may be read only or updated very infrequently.

Any feedback would be helpful.  Do we need to look at this from a totally
different perspective than the mainframe world?