Re: optimal size for rollback

"Karen Abgarian" <abvk_at_ureach.com> wrote in message news:3D8CC2B3.E8F10610_at_ureach.com...
> > The difference between us, Karen, is that when I post I'm in control
enough
> > to make sure that what I post is comprehensible. And you do not.
>
> Try to not get personal and discuss a subject, not you or me.
>

It's not getting personal to say that what you write is incomprehensible. Its a comment on your writing, not you. You wrote that shrinks due to optimal both would and wouldn't have an effect on performance. It's difficult to discuss that proposition, as it makes no sense.

I gave you the technical explanation, and whilst I could give you the numbers (as I've done on many occassions in the past), its not actually a hard test to set up for yourself: build a 16K rollback segment with optimal set to 16K, and update 1000000 rows to force growth. Then update 1000 rows to force shrinkage. Do the same thing without optimal. Create the segment in both LM and DM tablespace.

You are correct that doing it for yourself is the only sure way to distinguish truth from fiction.

HJR
>
> >
> >
> > For example, you say "It is obvious that it will have some effect when
it is
> > present and transactions will be faster when it is absent". (I presume,
> > incidentally, that we are talking about the existence of the "optimal"
> > clause -only it's not entirely clear from your post what "it" actually
is).
> >
> > So you appear to concede that "it will have some effect".
> >
> > Only in the next breath we have "I question the existence of the
performance
> > hit".
> >
> > Well, you can't have it both ways. Either there's an effect, or there
isn't.
> > If there is, there's a performance hit. If there isn't an effect, then
> > there's no performance hit, and (by logical extrapolation) Oracle's just
> > invented a way of doing data dictionary lookups and extent allocations
and
> > deallocations for free.
> >
> > Or are you saying that whilst the "effect" is there, it isn't serious
enough
> > to warrant being called a "performance hit"?? In which case, you're on
your
> > own debating such semantics.
> >
> > To perform a shrink to optimal, even in LMT, requires visits to the data
> > dictionary, and those don't come for free.
> >
> > HJR
> >
> > >
>
> I suggested that you come up with some technical explanation or numbers
> proving the point. For me, to conclude that something is true, the
following
> things
> DON'T matter:
>
> - who says it;
> - how much he/she talks about it;
> - how loud he/she talks about it;
> - what he/she thinks of those who think the opposite.
>
> Unless you want to discuss that technically, lets close the subject.
>
>
Received on Sat Sep 21 2002 - 16:20:32 CDT