Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Usenet -> c.d.o.server -> Re: determining extent size, etc.

Re: determining extent size, etc.

From: Jonathan Lewis <jonathan_at_jlcomp.demon.co.uk>
Date: Wed, 12 Jun 2002 07:56:40 +0100
Message-ID: <1023865065.19477.2.nnrp-01.9e984b29@news.demon.co.uk>

I agree with Daniel and Niall, though being computer-literate (you may substitute your own choice of adjective at this point) I prefer to use multiples of 16 rather than multiples of 10.

    So 64K, 1M, 16M

Of if the spread is very wide and you want to keep to just 3 sizes, go in multiples of 32 -

    64K, 2M, 64M

I tend to avoid more than a few dozen
(should that be hexes) extents per segment, and would think about moving a segment from one tablespace to the next size up byt the time it was getting on for two extents of the next tablespace size.

And (as another poster suggested) more than a couple of thousand extents per tablespace - but the latter is more of an issue with DMTs than LMTs.

Also, if you have just a couple of extremely large object, then I would consider a couple of tablespaces dedicated to each of them - always try to make it easy to monitor the I/O on big objects

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

Next Seminars
        UK            June / July
        Australia      July / August
http://www.jlcomp.demon.co.uk/seminar.html

Niall Litchfield wrote in message
<3d0663d9$0$8512$cc9e4d1f_at_news.dial.pipex.com>...

>"Daniel Morgan" <dmorgan_at_exesolutions.com> wrote in message
>>
>
>I think I'd start with small = stripe size (usually 64 or 128k) = multi
>block read count, and then as Dan says go roughly in orders of magnitude.
>This is almost certainly based on a complete misunderstanding of how
>controllers read data from disks though.
>
Received on Wed Jun 12 2002 - 01:56:40 CDT

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US