Instead, if you decide to use CTAS, you will create the table in
nologging mode (you obtain the same result of hint APPEND).
superG wrote:
> Sorry for my mistake, but yesterday I was very tired.
> The append hint must be used in INSERT. I suggest also the use of the
> parallel hint (set parallelism degree to <CPUs/2>, for example, if your
> server has more than 1 cpu), or, before executing the insert, run the
> following command:
>
> alter session enable parallel dml;
>
> You must estimate the degree of parallelism of the table where you are
> going to do INSERT.
>
> Bye.
>
> G. Pignatelli
>
> HansF wrote:
> > On Tue, 08 Aug 2006 14:56:30 -0700, joel garry wrote:
> >
> > >
> > > HansF wrote:
> > >> On Tue, 08 Aug 2006 20:55:12 +0200, Maxim Demenko wrote:
> > >>
> > >> >
> > >> > Actually, it seems to be a very funny thread now.
> > >> > I didn't know before, that /** append */ (i.e. a simply comment) can so
> > >> > improve performance... But , should be of course tested - and on my
> > >> > linux box - it seems, Oracle scales with the count of tests ;-)
> > >> >
> > >>
> > >> Of course, to do this properly, we should also flush the shared pool.
> > >>
> > >> And put it into a script and rerun it 6 or more times to ensure that the
> > >> results are repeatable.
> > >> (http://www.brainyquote.com/quotes/quotes/b/benjaminfr109067.html)
> > >
> > > Therefore, the essence of science is insanity :-)
> > >
> > > I think the timing results of your inserts before the truncate should
> > > be thrown out due to possible segment extension issues.
> > >
> >
> > Good call.
> >
> > --
> > Hans Forbrich (mailto: Fuzzy.GreyBeard_at_gmail.com)
> > *** Feel free to correct me when I'm wrong!
> > *** Top posting [replies] guarantees I won't respond.
Received on Wed Aug 09 2006 - 02:57:05 CDT
