Glenn Cole <[log in to unmask]> wrote:
 
>Uh, oh! Just when you think you understand...
 
>Michael Hornsby <[log in to unmask]> wrote:
><<
>Use Adager to reorganize the master data set. Secondary entries are placed
>in a way that helps to avoid clusters.
>>>
 
>Is this true? I mean, I believe (and I certainly could be wrong here) that
>Adager makes two passes when re-adding Master entries -- ONLY primaries in
>the first pass, and secondaries in the second. However, if the data is
>going to cluster WITHOUT Adager, then it will cluster WITH Adager, right?
>Or am I missing a key detail here?
 
Adager's repacking of a master dataset can make a difference with clusters
if the clusters are made up of primaries and secondaries.
 
If a cluster is made up of all primaries then, yes, that cluster will still
exist after
repacking because Adager has no control over where the primaries should be
placed.
 
Adager, however, does have control over where secondaries should be placed
and
uses a different algorithm then IMAGE.  IMAGE starts looking in the block
that the
record should have been put in and searches forward looking for the first
available
opening.  Adager makes sure that it does not fill up a block when placing the
secondaries.  (Check your MAXBLOCKS before and after a master repack with a
tool like Robelle's HOWMESSY)
 
This should have a positive impact when adding new records to the dataset
because
IMAGE should find an available spot sooner.  This may, however, have a
negative impact
on retrieval of the secondaries if they are placed too far away and force an
additional I/O.
(Check your INEFF PTRS before and after with HOWMESSY)
 
Just remember that there is no FREE lunch.
 
This is again why I believe it is very important to find out if your
performance problems
are happening while you are adding data or while you are retrieving data.
 This one answer
can save you a lot of time in finding the solution.
 
Hope this helps.
 
Ken Paul (a.k.a. [log in to unmask])