Re: COPY Fillfactor patch

From: Simon Riggs <simon(at)2ndquadrant(dot)com>
To: Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
Cc: Bruce Momjian <pgman(at)candle(dot)pha(dot)pa(dot)us>, Hannu Krosing <hannu(at)tm(dot)ee>, pgsql-patches(at)postgresql(dot)org, Neil Conway <neilc(at)samurai(dot)com>
Subject: Re: COPY Fillfactor patch
Date: 2005-04-20 16:05:55
Message-ID: 1114013155.16721.2275.camel@localhost.localdomain
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-patches

On Wed, 2005-04-20 at 09:42 -0400, Tom Lane wrote:
> Simon Riggs <simon(at)2ndquadrant(dot)com> writes:
> > On Wed, 2005-04-20 at 09:03 -0400, Bruce Momjian wrote:
> >> My guess is that the FILLFACTOR will actually be a GUC variable rather
> >> than a clause to CREATE INDEX or REINDEX or CLUSTER.
>
> > I hope not, but it would be easier to set like that and a lot easier to
> > code up that way.
>
> I don't like that either. It seems highly unlikely that every table in
> a database would need the same fillfactor. It should be a per-table
> attribute, maybe with a DB-wide default, much like we handle per-column
> statistics targets.
>
> My thought about this is that the case where extra free space really
> helps is UPDATE: if heap_update can put the new tuple version on the
> same page where the old one is, there's a noticeable savings. So what
> you'd probably want is that heap_insert respects fillfactor (ie, won't
> eat the last fillfactor percent of a page) but heap_update doesn't,
> if it can thereby put the update on the same page. As long as you
> vacuum before a particular page is entirely filled by updates of its
> own tuples, you have a steady-state situation.

Agreed. Thats the best approach.

Best Regards, Simon Riggs

In response to

Browse pgsql-patches by date

  From Date Subject
Next Message Oliver Siegmar 2005-04-21 07:10:02 Re: BUG #1609: Bug in interval datatype for 64 Bit timestamps
Previous Message Alvaro Herrera 2005-04-20 13:58:59 Re: FW: Implementation of SQLSTATE and SQLERRM variables