Re: 11G Statistics Collection
From: Robert Klemme <shortcutter_at_googlemail.com>
Date: Mon, 16 Apr 2012 22:15:49 +0200
Message-ID: <9v3cvkF8t6U1_at_mid.individual.net>
On 16.04.2012 17:27, Mladen Gogala wrote:
> On Mon, 16 Apr 2012 15:18:19 +0200, Robert Klemme wrote:
>
>> But that approach does not work if the distribution changes in a way
>> that different plans are optimal at different times.
>
> That approach works. Don't fix it if it ain't broken. If the SQL performs
> well, nail it down and go about the other stuff.
Date: Mon, 16 Apr 2012 22:15:49 +0200
Message-ID: <9v3cvkF8t6U1_at_mid.individual.net>
On 16.04.2012 17:27, Mladen Gogala wrote:
> On Mon, 16 Apr 2012 15:18:19 +0200, Robert Klemme wrote:
>
>> But that approach does not work if the distribution changes in a way
>> that different plans are optimal at different times.
>
> That approach works. Don't fix it if it ain't broken. If the SQL performs
> well, nail it down and go about the other stuff.
I don't understand: if data distribution at time of "nailing" favors one index while at other times access via this index is slower than FTS how then would you call that a "working approach" - other than the query will eventually return (modulo snapshot too old or other issues)?
Kind regards
robert
-- remember.guy do |as, often| as.you_can - without end http://blog.rubybestpractices.com/Received on Mon Apr 16 2012 - 15:15:49 CDT