Re: Defragmentation

IIRC = If I Recall Correctly - which I obviously didn't

thanks for the information
"Howard J. Rogers" <howardjr_at_www.com> wrote in message news:39461f41_at_news.iprimus.com.au...
>
> "Niall Litchfield" <niall.litchfield_at_doial.pipex.com> wrote in message
> news:8i4tcu$4va$1_at_soap.pipex.net...
> > "Howard J. Rogers" <howardjr_at_www.com> wrote in message
> > news:39449619$1_at_news.iprimus.com.au...
> > >
> > > ""marco pinzuti"" <mpinzuti_at_virgilio.it> wrote in message
> > > news:A5B2E7DDB0044D1178C40005B83A896E_at_mpinzuti.virgilio.it...
> > > If you have at least Oracle 8, you can set MINIMUM EXTENT for a
> > > tablespace -nothing you specify at segment level can override it, and
 hence
> > > you are at least assured of a few standard extent sizes within the
> > > tablespace, rather than a whole bunch of completely random sizes.
> >
> > IIRC (and I can't be bothered to go to the docs because we just run with
 std
> > extent sizes).doesn't MINIMUM EXTENT just specify the min value for
 INITIAL
> > & NEXT and so will cause random sizes above a certain value. IMO you
 should
> > not be specifiying storage at the segment level but at the tablespace
 level
> > (with INITIAL=NEXT and PCTINCREASE 0).
> >
> >
>
> Agree with your last sentence. I always advise against segment-level
> specification of extent sizes. However MINIMUM EXTENT is *not* as you
> describe. If you set MINIMUM EXTENT to, say, 500K, then if you were to
> create a segment specifying INITIAL=40K, you would have a 500K extent
> allocated regardless (because it is designed NOT to be overridden by
> segments). However, if you specified an INITIAL of, say, 600K you will
> actually be allocated an extent of 1M. And if you asked for an extent of
> 1200K, you would be given one of 1500K.
>
> In other words, you always get extents of size MINIMUM EXTENT *or
 multiples
> thereof*, and the system always rounds up to the next multiple -not the
> nearest multiple.
>
> What you end up with, as I said earlier, is a limited variety of extent
> sizes (500, 1000 and 1500K and so on) -which is at least better than a
 whole
> bunch of -eg- 43K, 57K, 150K, 325K, 73K, 64K and so on and on and on ad
> infinitum.
>
> Nothing random about the extent sizes at all, in other words.
>
> Incidentally, what does IIRC mean?
>
> Regards
> HJR
>
>
>
>
>
>
> > --
> > Niall Litchfield
> > Oracle DBA
> > Audit Commission UK
> >
> >
>
>
Received on Fri Jun 16 2000 - 00:00:00 CDT