Re: indexing

From: Cary Millsap <cary.millsap_at_method-r.com>
Date: Wed, 20 Feb 2013 14:56:16 -0700
Message-ID: <CAJOkrQadQ-+9hm4=xAuF_zztkYm9Q0ykNpM3k8Z+9J2dEOKZFw_at_mail.gmail.com>

...And when you trace a user's update because it takes too much response time, and you find out that a 3-table join is visiting the database buffer cache 150 times per updated row instead of just 30, it's the extra index touches that are causing the extra pain (a delete and an insert upon each index, plus the undo for these operations). You'll of course get extra redo being generated, too.

Cary Millsap
Method R Corporation

On Wed, Feb 20, 2013 at 1:24 PM, <Jay.Miller_at_tdameritrade.com> wrote:

> They finally fix that in 11.2? Thank you for the update, I still remember
> discovering that gotcha back when the feature was first introduced and I
> was (for a short time) ecstatic over it.
>
> Jay Miller
> Sr. Oracle Database Administrator
> 201.369.8355
>
> -----Original Message-----
> From: oracle-l-bounce_at_freelists.org [mailto:oracle-l-bounce_at_freelists.org]
> On Behalf Of Wolfgang Breitling
> Sent: Wednesday, February 20, 2013 9:33 AM
> To: oracle-l List
> Cc: Brian Zelli
> Subject: Re: indexing
>
> There is another little known problem with index monitoring. Until 11.2
> just gathering statistics on an index counted as "usage".
>
> On 2013-02-19, at 12:32 PM, Wolfgang Breitling wrote:
>
> > A word of caution: index usage monitoring is not foolproof.
> --
> http://www.freelists.org/webpage/oracle-l
>
>
> --
> http://www.freelists.org/webpage/oracle-l
>
>
>

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

Received on Wed Feb 20 2013 - 22:56:16 CET