Re: locally managed tablespace

Nah, absolutely not, Jonathan. No minextents (I tend not to believe in it, except for rollbacks ;-) )

This is 8.1.5 on Solaris 5.6 (???) I can check it out on the office tomorrow for certain (there's a demo I do, and the script of it is on the website, so I'm confident I can repeat this for you tomorrow if need be!)

However, the stuff I did this arvo. on 8.1.7 also happened to be on Windows 2000. So now you've got me thinking it's an o/s thing. I don't have access to anything higher than 8.1.5 on a Unix platform, so the best test I can come up with is 8.1.6 on W2K... I'll let you know!!

(Actually, thinking about it, I can do an Intel Solaris 8.1.7, and see what happens there as well....)

Regards
HJR "Jonathan Lewis" <jonathan_at_jlcomp.demon.co.uk> wrote in message news:981538254.15078.0.nnrp-12.9e984b29_at_news.demon.co.uk...
>
> Howard,
>
> Which platform and specific version: the behaviour
> I got on 8.1.5 (specifically 8.1.5.0.0) was not as
> you describe -
>
> Under 8.1.5.0.0 on NT (and 8.1.7 on HP-UX 11),
>
> If I specify
> storage (initial 3M next 2M)
> on a tablespace with uniform storage of 1M, then
> I get 3 extents of 1M each.
>
> However, there is the case where I specify
> storage (initial 3M next 2M minextents 2)
> in which case I get 5 extents of 1M each.
>
> Or even :
> storage (initial 1M next 2M minextents 4 pctincrease 50)
> where I get 11 extents of 1M - a notional
> 1M, 2M, 3M, 4.5M (rounded up).
>
>
> In other words, my initial response was a little superficial.
> Oracle will consider the full storage clause in the traditional
> way, then convert the result into the smallest number
> of uniform extents required to contain the result. Setting
> next_exent to the uniform size after the object is created.
>
> Is it possible that your test table also
> had a 'minextents' clause ?
>
>
> --
> Jonathan Lewis
> Yet another Oracle-related web site: http://www.jlcomp.demon.co.uk
>
> Practical Oracle 8i: Building Efficient Databases
> Publishers: Addison-Wesley
>
> Reviews at: http://www.jlcomp.demon.co.uk/book_rev.html
>
>
>
> Howard J. Rogers wrote in message <3a8100a4_at_news.iprimus.com.au>...
> >Interesting -Just discovered this today. In 8.1.7 (and I believe in
 8.1.6),
> >exactly what Jonathan says should happen does.
> >
> >With uniform extent size=1m, and a segment created with initial=3m next
 =2m,
> >you get 3 initial extents of 1m each. Then you say "allocate
 extent" -and
> >you get a single 1m extent.
> >
> >What I found interesting today is that that is *not* what 8.1.5 did. In
> >8.1.5, you would have ended up acquiring 2 extra extents of 1 meg each,
 to
> >fulfil the "2m next" requirement.
> >
> >I'm slightly gobsmacked that a mere point-release should introduce such a
> >difference in something so basic as extent allocation behaviour. Mind
 you,
> >8.1.5 is a dog of a product anyway, and 8.1.6 at least is always
 infinitely
> >preferable.
> >
>
>
>
Received on Wed Feb 07 2001 - 05:38:45 CST