Yes, because you only swap a sort to disk in the first place if it doesn't
fit in the sort_area_size; and when you do swap to disk, regardless of the
mechanism involved in actually getting it to disk, it's the contents of the
sort_area_size that are swapped. So what it gets swapped to needs to be
organised in sort_area_sized chunks, or multiples of sort_area_size.
Regards
HJR
--
------------------------------------------
Resources for Oracle : www.hjrdba.com
============================
"G.Ong" <onggs_at_acslink.aone.net.au> wrote in message
news:dyYn8.1326$Go6.108418_at_ozemail.com.au...
> Hi,
>
> I like to ask if the use of sort_area_size as a factor in the uniform
extent
> is still relevant with direct sort now being used ?
>
> Ghee
>
> Howard J. Rogers wrote in message ...
> >No. That advice applies to temporary tablespaces which happen to be
> >dictionary managed, because the initial extent has to allow for the
segment
> >header block.
> >
> >But the poster has indicated that he is using 8i temporary tablespaces in
> >locally managed tablespaces, and in such tablespaces, the advice from
> Oracle
> >is that the uniform size parameter should *equal* the sort_area_size or a
> >multiple thereof, *without* the addition of the extra block. Since
extent
> >allocations are handled differently in locally managed tablespaces, there
> is
> >*no* need for the extra space required in dictionary managed versions of
> >temporary tablespace.
> >
> >So, to answer the original poster, no... stick to the exact setting of
> >sort_area_size (or an integer multiple thereof), and don't pad it with
> extra
> >bytes at all. You will pay a performance penalty if you do.
> >
> >Regards
> >HJR
> >--
>
>
>
Received on Tue Mar 26 2002 - 11:50:56 CST
