RE: Are we insane? Unique use of native DB2

LF1 would have an access path that would be shared by LF3. If you did a
DSPFD of LF3, it would say Shared of the owning access path LF1. Think of
it as the KEY of LF3 is the subset of the Key of LF1. LF2 in this example
has its only access path and not involved..

This is easy to implement. If LF1 exists, just delete and recreate LF3 and
it will now share LF1. The size of LF3 will also be very very small like
16K as the LF is just a pointer to the Access Path of LF1.

You will be amazed at performance improvements that you get. To really
simplify this, think about this. First, on writes to the physical file and
then updating the pointers in the access path to the records in the Physical
file, in the above example I am only updating two Access paths (LF1 & LF2) &
one physical file so you could say I reduced my I/O on this one write by
approximately 25%. LF3 is just a pointer to Access Path of LF1.

Pete Massiello

-----Original Message-----
From: midrange-l-bounces@xxxxxxxxxxxx
[mailto:midrange-l-bounces@xxxxxxxxxxxx] On Behalf Of David FOXWELL
Sent: Monday, March 17, 2008 4:37 AM
To: Midrange Systems Technical Discussion
Subject: RE: Are we insane? Unique use of native DB2



-----Message d'origine-----
De : midrange-l-bounces@xxxxxxxxxxxx
[mailto:midrange-l-bounces@xxxxxxxxxxxx] De la part de Alan Campin
Envoyé : jeudi 13 mars 2008 18:29
À : midrange-l@xxxxxxxxxxxx
Objet : RE: Are we insane? Unique use of native DB2

Alan Campin :
<snip>
OS/400 will share all the access paths, even on a partial key.
</snip>

Not sure if I fully understand that. If I have :
LF1 KFLD1, KFLD2
LF2 KFLD2, KFLD3
LF3 KFLD1

I would imagine you mean that LF3 shares with LF1 but LF2 would not share.
Is that right?


<snip>
I did some consulting at a company two years ago. Saved some 150gb of
storage by just recreating indexes and sharing the indexes.
</snip>
In the above example, if I were able to replace LF2 so that it had the same
key as LF1, do I reduce the amount of storage needed?



<snip>
This is really an exercise any shop should do. Check to see if you have
indexes that could be recreated and would then share an existing access
path.
</snip>

How would you start such a check? Go through all the applications to see if
a common access path could be created?


Thanks to all for a really interesting discussion.