Обсуждение: Re: [HACKERS] include/config.h FOLLOWUP

Integration wrote:
>
> ps. why not allow for larger tuples in general?  Do we take a speed hit?

Using large blocks is bad for performance: by increasing block size
you automatically decrease number of blocks in shared buffer pool -
this is bad for index scans and in multi-user environment!
Just remember that Informix (and others) use 2K blocks.
(Actually, I would like to have smaller blocks, but postgres lives
over file system...)

As for having big tuples - someone said about multi-representation
feature of Illustra (automatically storing of big fields outside
of tuple itself - in blobs, large objects, ...): looks very nice.

Vadim

On Mon, 5 Jan 1998, Vadim B. Mikheev wrote:

> Just remember that Informix (and others) use 2K blocks.

    So we're 4x what the commercial ones are as of right now?

Marc G. Fournier
Systems Administrator @ hub.org
primary: scrappy@hub.org           secondary: scrappy@{freebsd|postgresql}.org

>
> On Mon, 5 Jan 1998, Vadim B. Mikheev wrote:
>
> > Just remember that Informix (and others) use 2K blocks.
>
>     So we're 4x what the commercial ones are as of right now?

That is because they do not use the file system, so they try to match
the raw disk block sizes, while we try to match the file system size.

--
Bruce Momjian
maillist@candle.pha.pa.us

On Sun, 4 Jan 1998, Bruce Momjian wrote:

> >
> > On Mon, 5 Jan 1998, Vadim B. Mikheev wrote:
> >
> > > Just remember that Informix (and others) use 2K blocks.
> >
> >     So we're 4x what the commercial ones are as of right now?
>
> That is because they do not use the file system, so they try to match
> the raw disk block sizes, while we try to match the file system size.

    Irrelevant to my question...our tuples...are they 4x the size of the
commercial vendors, or is Vadim talking about something altogether different?

    If we are 4x their size, then I think this whole discussion is a joke since
we are already *way* better then "the others"

Marc G. Fournier
Systems Administrator @ hub.org
primary: scrappy@hub.org           secondary: scrappy@{freebsd|postgresql}.org

>
> On Sun, 4 Jan 1998, Bruce Momjian wrote:
>
> > >
> > > On Mon, 5 Jan 1998, Vadim B. Mikheev wrote:
> > >
> > > > Just remember that Informix (and others) use 2K blocks.
> > >
> > >     So we're 4x what the commercial ones are as of right now?
> >
> > That is because they do not use the file system, so they try to match
> > the raw disk block sizes, while we try to match the file system size.
>
>     Irrelevant to my question...our tuples...are they 4x the size of the
> commercial vendors, or is Vadim talking about something altogether different?
>
>     If we are 4x their size, then I think this whole discussion is a joke since
> we are already *way* better then "the others"

That's a good question.  What is the maximum tuple size for Informix or
Oracle tuples?

--
Bruce Momjian
maillist@candle.pha.pa.us