How to optimize server for cartesian join? RESULSTS

From: Jack <Jack_at_none.com>
Date: Fri, 22 Feb 2008 20:12:13 +0200
Message-ID: <47bf107f$0$14994$9b536df3@news.fv.fi>

"Jack" <none_at_INVALIDmail.com> kirjoitti viestissä:Kneoj.40$Ll.3_at_read4.inet.fi...
> We have business need to do some countings.
>
> It is using MERGE JOIN [CARTESIAN], it will go fast with 10.000 count,
> but
> with 200.000 end sum it will take very long time.
> It is using Reports 6i (a bit old;)
>
> It does utilize very complez views etc, and plan is long as you would
> assume.
> I am not going to send plan.
>
> One solution would be redesing it, but not quite sure it would help.
>
> Server is 9.2.0.8 (multi prosessor, quite good one), parameters are now
> very small, and
> not optimized.
>
> Is there some good parameters in server side which could help?
>

"Doug Miller" <spambait_at_milmac.com> kirjoitti viestissä:laosj.115$Mw.13_at_nlpi068.nbdc.sbc.com...
> In article <Kneoj.40$Ll.3_at_read4.inet.fi>, "Jack" <none_at_INVALIDmail.com>
> wrote:
>>We have business need to do some countings.
>>
>>It is using MERGE JOIN [CARTESIAN], it will go fast with 10.000 count,
>>but
>>with 200.000 end sum it will take very long time.
>>It is using Reports 6i (a bit old;)
>>
> The only way to truly "optimize" a Cartesian join is to eliminate it, by
> creating an index for at least one of the tables, on the column(s) used in
> the join.

Hi

Thanks Doug for your advice.
It does have good indexes. (but not in remarksfield ;)

Now after some basic parameter changes it is running 3,3x faster.

It does not like more memory or paraller query.

Any good advices, what next? Received on Fri Feb 22 2008 - 12:12:13 CST

Original text of this message