Re: Performance of count(*)

Am Donnerstag, 22. März 2007 16:17 schrieb Andreas Kostyrka:
> * Mario Weilguni <mweilguni(at)sime(dot)com> [070322 15:59]:
> > Am Donnerstag, 22. März 2007 15:33 schrieb Jonah H. Harris:
> > > On 3/22/07, Merlin Moncure <mmoncure(at)gmail(dot)com> wrote:
> > > > As others suggest select count(*) from table is very special case
> > > > which non-mvcc databases can optimize for.
> > >
> > > Well, other MVCC database still do it faster than we do. However, I
> > > think we'll be able to use the dead space map for speeding this up a
> > > bit wouldn't we?
> >
> > Which MVCC DB do you mean? Just curious...
>
> Well, mysql claims InnoDB to be mvcc ;)

Ok, but last time I tried count(*) with InnoDB tables did take roughly(*) the
same time last time I tried - because InnoDB has the same problem as postgres
and has to do a seqscan too (I think it's mentioned somewhere in their docs).

(*) in fact, postgres was faster, but the values were comparable, 40 seconds
vs. 48 seconds

Maybe the InnoDB have made some progress here, I tested it with MySQL 5.0.18.