Re: Re: Re: A transaction profile question - block growth always groups of 5?

<61.12.128.6 [Igor Laletin]> wrote in message news:9dabt261mfj_at_news1.newsguy.com...
> > > > You'll also note from Steve's website that fragmentation doesn't
 affect
 this
> > > > issue, or cause it, in the slightest, despite what you posted
 earlier.
> > >
> > > Wrong. Howard: read your books.
> > > http://www.ixora.com.au/tips/creation/extents.htm
> > > "Extents may be enlarged further if their allocation would otherwise
 leave
 a free space fragment of less than 5 blocks."
> > >
> > > Say you have 12-blocks free extent and you ask for 7-blocks extent.
 Oracle
 rounds up to 10 but remaining 2 blocks are a too small chunk.
> > > So you get all 12.
>
> [snip]
>
> > In other words, to avoid fragmentation in the first place, Oracle may
 decide
> > NOT to round to multiples of 5. Which is entirely the opposite of what
 you
> > stated originally ("Yes, it can get up to 4 blocks more than you
 requested.
> > _If_ the free space is fragmented in such a way Oracle couldn't find
> > anything better."),
>
> What is opposite? You already have a fragmentation. Can you get +4 blocks?
 Yes, it is possible.
>
> > and agrees with my statement that 'fragmentation does
> > not cause [rounding]' -because what you have here is potential
 fragmentation
> > *preventing* the rounding.
>
> Err... actually you said 'fragmentation doesn't _affect_ this issue
[rounding]'. It does. If you have fragmented space could happen that your extents are not multiples of 5.

Er, I was refuting your suggestion that rounding would only happen if there was fragmentation. It doesn't, as you acknowledged.

>
> > I agree that in this quantum age, concepts of
> > causality can be confusing, but not that much in this case, surely.
>
> I was not 100% right and you were not either. Let's agree on that.
>

Nope. I was not 100% right, because I ommitted to mention the 'request smaller than 5 doesn't get rounded' issue. But you were practically 100% wrong (and offensive about it to boot).

HJR
> Dixi,
> Igor.
>
> >
> > [Snip]
> >
> > HJR
>
>
> ==================================
> Poster's IP address: 61.12.128.6
> Posted via http://nodevice.com
> Linux Programmer's Site
Received on Wed May 09 2001 - 01:27:21 CDT