Any stat missing in a CBO situation will be set to the hard-coded default
values found in the source code. One example (out of about 40 or so
values) is default selectivity for equality predicates. Before 7.3 it was
0.125 (meaning that a where= condition would assume you needed to scan
1/8th of
Are these defaults recorded somewhere that
mere mortals have access to?
Jared
On Thursday 07 June 2001 00:25, Mogens Nørgaard wrote:
Any stat missing in a CBO situation will be set to the hard-coded default
values found in the source code. One example (out of about 40 or so
values) is
Jared:
In the Oracle 8.1.6 Docs:
Oracle8i Designing and Tuning for Performance
8 Gathering Statistics
Table 8-2 Default Table and Index Values When
Statistics are Missing
Kevin
-Original Message-
Sent: Thursday, June 07, 2001 6:21 AM
To:
Probably. I'm not responsible for posts I answer
in the middle of the night.
Jared
On Thursday 07 June 2001 07:21, Connor McDonald wrote:
Didn't I already say Chapter 8
:-)
Connor
--- Jared Still [EMAIL PROTECTED] wrote:
Are these defaults recorded somewhere that
mere
Ya know, having read that entire chapter last week
you would think I could remember this.
Jared
On Thursday 07 June 2001 05:30, Toepke, Kevin M wrote:
Jared:
In the Oracle 8.1.6 Docs:
Oracle8i Designing and Tuning for Performance
8 Gathering Statistics
Hi Jared,
My post was to meant to imply that (say) if (in a
single query - and thats the important bit I missed
out) there were two tables, one with stats, and one
without, then the one without will use defaults (as
quoted in chapter 8 of the same manual)
I was trying to imply that if you've
Got it, I misunderstood your previous post.
If nothing else, I got some more good reading
out of the tuning manual. :)
Jared
On Wednesday 06 June 2001 02:36, Connor McDonald wrote:
Hi Jared,
My post was to meant to imply that (say) if (in a
single query - and thats the important bit I
Sounds like nonsense to me...If you have half
statistics, then the optimizer is going to use
defaults for the others, which if anything, would (or
should) lead to less optimal access paths
It sounds like a weird attempt to get toward a first
rows kind of approach, ie they want a rule type
On Tuesday 05 June 2001 01:20, Connor McDonald wrote:
Sounds like nonsense to me...If you have half
statistics, then the optimizer is going to use
defaults for the others,
Conner,
Are you sure about that?
From the Optimizer chapter in the tuning manual:
If the data dictionary contains
: Compute statisics
On Tuesday 05 June 2001 01:20, Connor McDonald wrote:
Sounds like nonsense to me...If you have half
statistics, then the optimizer is going to use
defaults for the others,
Conner,
Are you sure about that?
From the Optimizer chapter in the tuning manual
10 matches
Mail list logo