ADSM-L

Re: Squishing the unused space out of my ADSM database

1999-05-24 05:59:38
Subject: Re: Squishing the unused space out of my ADSM database
From: Brian D Chase <bdc AT WORLD.STD DOT COM>
Date: Mon, 24 May 1999 02:59:38 -0700
On Mon, 24 May 1999, Sheelagh Treweek wrote:

> I would be a bit cautious on the expected outcome of a smaller database
> when using the dump/load DB.  There is a very strong possibility that
> the database will end up bigger than you started with!  I can't explain
> this in full detail but it is to do with the way records are packed into
> DB pages when the load occurs, and space is left in pages for later
> expansion.

Hmm... I saw a note for loaddb stating something similar to that.  I'd
hoped given excessively Swiss-cheesed nature of my current database,
that there'd likely be a decrease in the database size.  I'd be really,
well sad I guess, if my 40GB database grew larger.

> I think Trevor is right with the time factor ... it would take
> significantly longer (and then you might have to do the restore DB as
> well to get back to what you are now).

So I'm looking at somewhere between 24-48 hrs just for the dumpdb/loaddb,
precede that with another 2 hrs to make a clean full database backup, and
then add 6 hrs for a potential database restore. I can tell that I'm in
for an upcoming weekend filled with merriment and joy.  Thank goodness for
detached screen sessions and remote dialup access.

> I think there is a need for a tool for optimising and reclaiming space
> in the ADSM DB and IBM should be encouraged to produce such a tool.

... should be encouraged with baseball (and/or cricket) bats ...

Even if it meant that I had to create new empty db volumes and move the
active pages into them, that would be fine. How difficult is it to
identify the database pages which have "stuff" in them vs the ones that
don't, and then how difficult is it to move those good pages in a nice
orderly fashion into brand new shiney db vols?  Or for that matter, how
difficult is it to reorganize those pages internal to the exisiting
database volumes?  Just start at the far end of the database and
reintegrate any active pages back into the empty holes at the front end of
the database.

I'm not a database software system designer, so I can't really speak with
any authority on how difficult it would be to add this to ADSM.  But
obviously ADSM has some sort of concept of active and unused pages -- I
mean given that it can report on database utilization. This would seem
like one of the primary components necessary to make an optimization tool.
Other people have figured this out as well with their database systems,
yes?

Wearily,
-brian.
---
---
Brian Chase | http://world.std.com/~bdc/ | VAXZilla LIVES!!!
Brian Chase | http://world.std.com/~bdc/ | VAXZilla LIVES!!!