Re: Index skip scan
- From: Mladen Gogala <gogala@xxxxxxxxxxxxx>
- To: fairlie_r@xxxxxxxxx
- Date: Wed, 21 Sep 2005 09:43:20 +0000
On 09/21/2005 02:02:20 AM, fairlie rego wrote:
> Hi,
>
> Due to the following bugs your best bet is to use the INDEX_SS hint as
> mentioned by Mladen before
>
> 3354231Index skip scan is sometimes estimated at very high cost fixed in 9206
>
>
> 4050226Index skip scan may not be chosen when it is a good choice fixed in
> 9207
>
> Regards,
> Fairlie
>
Quite frankly, I am not sure that I want index skip scan to be chosen just
like that, when CBO feels like it. Index skip scan is one thing that goes
against the traditional philosophy which says that index is usable in a query
if its leading columns are in the "WHERE" clause. This principle, which will
obviously cease to apply in the very near future, was used as a base for our
indexing schemes. In particular, for large reports and queries that we didn't
want to use an index, we omitted the leading column from the "WHERE" clause.
Now, index might even be chosen despite that. I didn't know about the
"features"
of the optimizer, but it was my experience that INDEX_SS was never chosen,
unless
the hint was given. This state of the affairs did not bother me at all. As a
true
blue conservative, I am little worried about that.
--
Mladen Gogala
http://www.mgogala.com
--
//www.freelists.org/webpage/oracle-l
Other related posts:
