Re: Optimiser choice of index
Posted in 2005
Colin Dawson wrote:
> Update statistics high on 1st column in each index, low on other indexed> columns is what's recommneded by Informix, what else should I run?
Recommended suite:
HIGH on lead column of each index - DISTRIBUTIONS ONLY
HIGH on first different key column if 2 or more indexes start with the same
column(s) - DISTRIBUTIONS ONLY
MEDIUM on all other columns - DISTRIBUTIONS ONLY (can be done in a single
command)
LOW for each full key list of every index
Optimizations:
For older servers (prior to 7.31xD3 and 9.30xC2) you can just start with a
MEDIUM on the whole table.
For newer servers you can gang all of the HIGHs into a single (or small
number) of statements rather than doing individual columns (this is too slow
on older servers).
Or just run my dostats utility.
Art S. Kagel
> Regards
>
> Colin
>
> There are 10 types of people in the world, those that understand binary and
> those that don't
>
>
>
>
>
>
>>From: "scottishpoet" <dryburghj@yahoo.com>
>>Reply-To: "scottishpoet" <dryburghj@yahoo.com>
>>To: informix-list@iiug.org
>>Subject: Re: Optimiser choice of index
>>Date: 23 Sep 2005 03:45:06 -0700
>>
>>1/ Bug in the optimiser
>>
>>2/ The Update Statistics strategy you used did not obtain appropriate
>>information to provide the optimser with enough detail to choose the
>>right index
>
> sending to informix-list