RE: 10gR2 Upgrade .. Watch out

The following has not been tested objectively, it is full of subjective commentary.

While I haven't tested it specifically in a contrived scenario, the sheer fact of upgrading (from 10.1.0.5) with no other changes decreased the run for my nightly data warehouse batch by 25%. I have logging around all statements and none of them were so affected as to skew the entire result.

While this could be some several 10g enhancements I "believe" the new sorting and hashing algorithms are the cause as I have written and tuned all the sql / pl/sql code myself and it was as close to optimal as was realistic. My code is full of long hashes and sorts.

As for the person that complained that oracle goes out its way to break existing apps, the choice to upgrade is ultimately in the hands of the user and it is highly unlikely that someone at oracle intentionally writes bugs to make our life more complicated.

I have written bugs as a dba and a developer, sometimes out lack of forethought other times just making an error or a typo that still compiled, other times for no reason that I could discern after the fact.

As much as we think Oracle is one big amorphous entity, it is full of developers just like us probably complaining about bugs in lower level code just like us.

My 2 cents.

Ken

-----Original Message-----
From: oracle-l-bounce_at_freelists.org [mailto:oracle-l-bounce_at_freelists.org] On Behalf Of Mladen Gogala
Sent: Wednesday, December 27, 2006 8:17 PM To: gkatteri_at_fastmail.fm
Cc: oracle-l
Subject: Re: 10gR2 Upgrade .. Watch out

On 12/27/2006 07:27:02 PM, GovindanK wrote:
>
> >hash aggregation - if you have code that gets the data in
> >the right order after doing a group by without an order
> >by, you may now find the data coming out in the wrong
> >order because oracle has switch to "hash group by"
> >instead of "sort group by". Add order by clauses -
> >or disable hash aggregation by setting
> > _gby_hash_aggregation_enabled = false
>
> 'Group By' Does Not Sort If You Don'T Use Order By In 10g
>
> Note:345048.1
>

Yes, there is a new thing called "HASH GROUP BY" which seems faster to me, but I'm unable to
come up with a proper benchmark. That is probably something Jonathan could prove or disprove.

-- 
Mladen Gogala
http://www.mladen-gogala.com

--
http://www.freelists.org/webpage/oracle-l


--
http://www.freelists.org/webpage/oracle-l