Message-ID: <4360F8C1.4040202@archonet.com>
Date: Thu, 27 Oct 2005 16:56:49 +0100
From: Richard Huxton <dev@archonet.com>
User-Agent: Mozilla Thunderbird 1.0.7 (X11/20051013)
MIME-Version: 1.0
To: =?ISO-8859-1?Q?Sidar_L=F3pez_Cruz?= <sidarlopez@hotmail.com>
Cc: pgsql-performance@postgresql.org
Subject: Re: how postgresql request the computer resources
References: <BAY23-F28F5DD23A4A73C7BFE3CD1CE680@phx.gbl>
In-Reply-To: <BAY23-F28F5DD23A4A73C7BFE3CD1CE680@phx.gbl>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: quoted-printable

Sidar L=F3pez Cruz wrote:
> Is there something that tells postgres to take the resorces from=20
> computer (RAM, HDD, SWAP on linux) as it need, not modifying variables =

> on postgresql.conf and other operating system things?

Ah, and how is it to know what to share with other processes?

> A days ago i am trying to show that postgres is better than mssql but=20
> when execute a simple query like:
>=20
> (1)
> select count(*) from
> Total runtime: 134552.325 ms
>=20
> (2)
> select count(*) from fotos where archivo not in (select archivo from=20
> Total runtime: 26747.236 ms
>=20
> (3)
> select count(1) from fotos f where not exists (select a.archivo from=20
> Total runtime: 89765.714 ms
>=20
> (4)
> SELECT count(*)
> Total runtime: 114893.116 ms

> WITH ANY OF THIS QUERIES MSSQL TAKES NOT MUCH OF 7 SECONDS....

In which case they make a bad choice for showing PostgreSQL is faster=20
than MSSQL. Is this the only query you have, or are others giving you=20
problems too?

I think count(*) is about the weakest point in PG, but I don't think=20
there'll be a general solution available soon. As I'm sure someone has=20
mentioned, whatever else, PG needs to check the row for its visibility=20
information.

 From the start of your email, you seem to suspect your configuration=20
needs some work. Once you are happy that your settings in general are=20
good, you can override some by issuing set statements before your query. =

For example:
	SET work_mem =3D 10000;
might well improve example #2 where you had a hash.

--
   Richard Huxton
   Archonet Ltd