I don 't have access to the data to back this up, but I have seen PARSE
call response time move from over a minute to less than a second by
eliminating histograms from the equation. I think when I last saw this,
it was an Oracle8 system.
Cary Millsap
Hotsos Enterprises, Ltd.
http://www.hotsos.com
* Nullius in verba *
Upcoming events:
- Performance Diagnosis 101: 3/23 Park City, 4/6 Seattle
- Hotsos Symposium 2004: March 7-10 Dallas
- Visit www.hotsos.com for schedule details...
-- --Original Message-- --
From: oracle-l-bounce@(protected)
[mailto:oracle-l-bounce@(protected)] On Behalf Of Wolfgang Breitling
Sent: Monday, March 01, 2004 10:30 PM
To: oracle-l@(protected)
Subject: Re: RE: query slow in 9i, but not slow in 8i
There surely is some "cost " involved in evaluating a histogram during
the
parse phase, but I doubt that it will be significant enough to be
measurable. But histograms change the estimates of the CBO and it is
quite
possible that the presence of a histogram on a column where the content
is
not skewed enough to really warrant a histogram can change the estimates
enough to cause the optimizer to use a different, slower access path
than
without the histogram.
At 09:14 PM 3/1/2004, you wrote:
>I am unfortunately speaking from heresay, however I heard once that
having
>histograms all over the place can slow down parsing as they have to be
>inspected. I was left with the impression that unless the data is
skewed
>enough to warrant a histogram then there actually is a negative cost
>associated with having too many histograms.
>
>Surely someone on this list can either support me or shoot me down on
this
>one?
