Update statistics ( not recommended!) On mars
Hi,
*Lawsons* does not use any particular database as such and will
run on Oracle PSybeast or Igrist.
The people who wrote lawsons wrote it in Cobol.
They issue bug fixes by faxing out the new code FOR YOU TO TYPE
in and re-compile yourself. !
Lawsons have absolutely no idia at all what update statistics even does
let alone how it affects performance.
A couple of things indicate this.
1) You cannot 'get rid of' an update stats as their help desk suggests.
2) The whole point of updating stats IS that it enables Informix to
USE ITS INDEXES BETTER.
> evidence of their users, is that all their SQL is written with where
> clauses, and the indexes that exist will be used, which is what is
So ..
Update statistics on one table and see if the system comes
flying apart at the seams ( Which it won't.)
Update stats low just updates the system catalogues by the way.
Do a low then a medium then a high as Informix say you
should.
( If this is the client I think it is, they have so much cache on the
RAID array that the whole database is cached anyway,and
they have not run update stats since it was installed )
Update stats !!!
( I would, but then I don't work for them anymore )
I don't want to slag off lawsons. Its good financials package by the
way but they would not know an informix database if it came up
and poked them in the eye with a plastic daffodil
----------
From: owner-informix-list@iiug.org on behalf of Art S. Kagel
Sent: 14 August 1998 10:30
To: informix-list@iiug.org
Subject: Re: update statistics not recommended!
Neil Truby wrote:
>
> My new site has a third-party financial package (Lawsons), running on
> v7.13. The underlying DBMS is Informix. The Lawsons help desk has told
> me that they recommend that we never run UPDATE STATISTICS, as it will
> degrade performance. Their reasoning, allegedly based on empirical
> evidence of their users, is that all their SQL is written with where
> clauses, and the indexes that exist will be used, which is what is
> required. If we run an UPDATE STATS, we might risk unexpected and
> undesirable access paths. A caveat is added that, under v7.23, some
> customers have experienced benefits from an UPDATE STATISTICS LOW, and
> that we might run it, but prepared to remove the statistics so derived
> if performance nose-dives.
>
> Well, sounds like bollocks to me, but does anyone else have any views
> or, better still, experience of this software packeage running with
> Informix?
Sounds like REALLY badly written software to me.
Art S. Kagel