Re: TOAST performance (was Re: [GENERAL] Delete

Andrew McMillan wrote:
>
> On Tue, 2001-11-20 at 10:11, Hannu Krosing wrote:
> >
> > > Tom Lane wrote
> > >
> > >> I did some experimentation here and found a rather surprising
> > >> dependency: the time to delete a bunch of data is pretty much
> > >> directly proportional to the disk space it occupies. This says
> > >> that we're paying through the nose for having XLOG make copies
> > >> of about-to-be-modified pages.
> > >
> > Can't we somehow WAL only metadata and not the actual pages for
> > DELETEs - as delete is essentially (though currently not technically)
> > just metadata it should be a possible thing to do.
>
> Is it possible to do ordered writes, the way ext3 does?

I remember it being discussed on this list that you have very little
control over writing order if you operate above filesystem/cache level.

> http://www-106.ibm.com/developerworks/linux/library/l-fs7/

I guess that is the article that sparked the idea of journalling only
metadata for deletes (including the delete half of update)

Using the Journaling Block Device described there could actually be
a good (though currently not portable) solution if you run linux.

-------------
Hannu