Re: Expert tunning Database

Sybrand Bakker wrote:
> On Thu, 25 May 2006 20:38:02 -0500, John Koller
> <johnckoller_at_yahoo.com> wrote:
>
> >On Thu, 25 May 2006 17:58:08 +0200, Robert Klemme wrote:
> >
> >> Sybrand Bakker wrote:
> >>> So the real question is, why is the OP advocating this non-functional
> >>> method, which doesn't resolve anything on a *structural* basis (other
> >>> than keeping him employed)
> >>
> >> There you go. :-)
> >> robert
> >
> >Well, Donald Burleson is a consultant.
> >
> >The original is much easier to read by the way:
> >http://www.dba-oracle.com/art_dbazine_expert_tuning.htm
>
>
> We all know who Donald K Burleson is and we are all aware his advice
> is usually to be taken with many tons of salt, as it is usually easily
> to be disproved.
>
> --
> Sybrand Bakker, Senior Oracle DBA

More 'noise' from old Donnie:

"As we noted, experienced Oracle DBAs know that I/O is the single greatest component of response time and regularly work to reduce I/O. "

alter system kill session(...);

"In high-volume online transaction processing (OLTP) environments in which data is accessed via a primary index, resequencing table rows so that contiguous blocks follow the same order as their primary index can actually reduce physical I/O and improve response time during index-driven table queries."

Until, of course, you have inserts ...

"Once you have decided to re-sequence the rows in a table,..."

You should probably dust off the resume and start looking for a new employer.

"If the query returns less and 40 percent of the table rows in an ordered table, or 7 percent of the rows in an unordered table)..."

Hmm ... if the table is an IOT (the only table structure which can remain ordered after an insert) then there shouldn't BE a performance problem (except, of course, from the inserts themselves) and if this is a heap table it isn't ordered anyway (no matter what ole Donnie B. wants to think), even after this wonderful 'row resequencing' nonsense.  So, do his numbers sound fishy?

"The most common tuning for unnecessary full table scans is adding indexes."

Gee, Jonathan Lewis and others say otherwise ...

"The sole purpose of the pctfree and pctused table allocation parameters is to control the movement of blocks to and from the freelists. "

Gee, Donnie, have you read the Concepts Guide at http://tahiti.oracle.com lately?

'Nuff said.

David Fitzjarrell Received on Mon May 29 2006 - 18:48:24 CEST