On Wed, 2005-09-14 at 21:06, John Fabiani wrote:
> On Wednesday 14 September 2005 08:23, Scott Marlowe wrote:
>
> > OK. But how many are you updating between regular vacuums? That's the
> > real issue. If your regular vacuums aren't often enough, postgresql
> > starts lengthening the tables instead of reusing the space in them that
> > was freed by the last updates / deletes.
> >
> > Keep in mind, that in postgresql, all updates are really insert / delete
> > pairs, as far as storage is concerned. So, updates create dead tuples
> > just like deletes would.
> >
> > > Is my use of indexes correct?
> >
> > Seems good to me.
>
> Ok but this does seem to be a not a lot of records. Even if the user updated
> 500 times a day (500 * 200) will only add 100000 records. I would not expect
> that performance would suffer adding 100000 per day for at least a week.
> Even if the number was double (in case I mis-read the user prior emails)
> 200000 or 1000000 at the end of the week would not account for the slow down?
> Or am I miss reading?
I think he was saying he updated 200 at a go, but he was doing a LOT of
updates each day. Not sure, I don't have the OP in my email client
anymore.
