Re: Reorganizing tables

This may have been mentioned already, but a table re-org can be a positive threat to performance.

If your primary tables have (for example) a three-stage life:

    initial row size is small
    row grows after a couple of weeks
    row reaches full size after another 4 weeks, then what value are you going to use for your PCTFREE when you rebuild ?

If you set it to zero to avoid wasting space for the vast percentage of rows that are full size then many incomplete rows will end up migrating.

If you set it to match the growth requirement of the newest rows, then you leave lots of empty space in the blocks that hold only old rows.

(The optimum storage answer is to set the Hakan factor before moving the table - but that's not a trivial exercise because of a bug whose number I can't remember).

Regards

Jonathan Lewis
http://www.jlcomp.demon.co.uk

For one-day tutorials:
(see http://www.jlcomp.demon.co.uk/tutorial.html )

____UK_______April 8th
____UK_______April 22nd
____Denmark May 21-23rd
____USA_(FL)_May 2nd

Next dates for the 3-day seminar:
(see http://www.jlcomp.demon.co.uk/seminar.html )

____UK_(Manchester)_May
____Estonia___June (provisional)
____USA_(CA, TX)_August

The Co-operative Oracle Users' FAQ
http://www.jlcomp.demon.co.uk/faq/ind_faq.html

• Original Message ----- From: "DENNIS WILLIAMS" <DWILLIAMS_at_LIFETOUCH.COM> To: "Multiple recipients of list ORACLE-L" <ORACLE-L_at_fatcity.com> Sent: 24 March 2003 20:53 Subject: RE: Reorganizing tables

> Thanks Jared, Rachel, Tom, Dick, Prakash, Ron
>
> Excellent points. Very much appreciated. Unfortunately at this point
people
> are asking "but have you tested it?". So I need to construct some
type of
> test that will demonstrate how much effect a reorg will have. After
I've
> answered that question, then I can move on to some of the other
issues that
> you mention. I have joked that if the results are strongly positive,
they
> won't see me much after that because I'll be touring the world
selling my
> performance solution that never occurred to anyone else.
> Our test system is cloned from an RMAN backup of production so
the tables
> should be close to production. I'm thinking of creating a new table
and
> copying the contents of a production table into it and then tracing
> full-table scans and comparing the results.
>
> Jared - is there a way to estimate block-level fragmentation?
Comparing the
> average row length with the number of blocks used?
>
> Dennis Williams
Received on Mon Mar 24 2003 - 16:25:08 CST