Oracle FAQ | Your Portal to the Oracle Knowledge Grid |
![]() |
![]() |
Home -> Community -> Usenet -> c.d.o.misc -> Re: 7.3 Optimizer diffs from 7.2
In article
<19971114012801.UAA00980_at_ladder02.news.aol.com>,
crazylongs_at_aol.com (Crazylongs) wrote:
>
> We just upgraded to 7.3 from 7.2. Complex queries
run the same or even
faster
> than with 7.2. However, primary keys/indexes
seem to be ignored, and do ing
> FT scans. We either have to force the rule based
optimizer in init.ora
(which
> makes complex queries run slow) or force it by
deleting the statistics
before
> loading (which does the primary key lookups) and
then re-estimating after.
> What did Oracle do to the optimizer to have this
effect, has anyone else
seen
> it, and does anyone have any other suggestions?
>
> Note: The first thing we did was compute
(estimate) statistics for every
> object (using 30% of rows), thinking they should
be re-estimated for the new
> optimizer, but this had little or no effect.
I too, have seen different access paths chosen by
the cost based optimizer
between 7.2 and 7.3. I haven't looked into whether
this involves the primary
key of a table or not.
One occurence I do recall is where we have a hint in
a statement to use a
particular index in one table in the join of three
tables. The hint improved
the response time by convincing oracle to use a
particular access path. When
we moved the same database to 7.3, oracle used a
different access path, and the
statement ran longer. We did compute statistics on
both databases too. I
don't know the reason for the difference.
Another problem we encountered was when we estimated
statistics using the
default # of rows (1064 I believe) on the 7.2
database. We discovered the hard
way that the statistics generated this way can
really throw off how the
optimizer chooses its access paths. Once we ran a
compute statistics, the
entire system seemed to run much better. I don't
know if looking at 30% of the
rows might cause the same problem, but I assume that
it might if some of your
data is very skewed.
John Reimer
Database Administrator
Worthington Steel
-------------------==== Posted via Deja News ====----------------------- http://www.dejanews.com/ Search, Read, Post to UsenetReceived on Fri Nov 14 1997 - 00:00:00 CST
![]() |
![]() |