From: Luke H. Crouch
Date: August 4 2004 5:32pm
Subject: RE: performance question
List-Archive: http://lists.mysql.com/cluster/273
Message-Id: <075C170BC7219F439DD3FE8C880648C1AB0A78@lakota.red-man.com>
MIME-Version: 1.0
Content-Type: text/plain;
	charset="iso-8859-15"
Content-Transfer-Encoding: quoted-printable

okay, yeah...that was a good idea.=20

when running a count with a where clause, the MyISAM table took 18.19 =
seconds, and the same query on the cluster took 35.97 seconds.=20

this still seems like a pretty big difference...are there special table =
design optimizations for the cluster tables? or it may be that I did not =
allocate enough memory to indexes?

-L

> -----Original Message-----
> From: Paul Weiss [mailto:pgweiss@stripped]
> Sent: Wednesday, August 04, 2004 12:25 PM
> To: Crouch, Luke H.; cluster@stripped
> Subject: Re: performance question
>=20
>=20
> On Wed, 4 Aug 2004 11:47:40 -0500, Crouch, Luke H. =20
> <Luke.Crouch@stripped> wrote:
>=20
> > I thought the tables in the cluster are of type NDB, so=20
> they aren't =20
> > InnoDB or MyISAM...the table just happened to be in MyISAM on the =20
> > regular mysqld.
> >
>=20
> You were comparing select count(*) of a NDB table to a MyISAM=20
> table.  The =20
> MyISAM table is, of course, not in the cluster.  I was merely=20
> suggesting =20
> that if you instead compared it to an InnoDB table that was=20
> not in the =20
> cluster, the comparison would be more fair.
>=20
> > do you know how I can remove the record count on the=20
> regular mysqld =20
> > MyISAM tables? then I could see how long it actually takes=20
> to do the =20
> > full count on MyISAM and it may be similar in length of=20
> time to the =20
> > count on the NDB tables.
> >
> > is there not a way to have the cluster keep a count record=20
> as well? in =20
> > the same kind of way as MyISAM?
> >
>=20
> I don't think you can do either of these things.  Howewer,=20
> you might be =20
> able to trick mysqld into not using the count in the MyISAM table by =20
> adding a where clause that just happens to be true for all records.
>=20
>=20