Message-ID: <bcd915ceb9bb65f1a4e00b673371071c63146a4f.camel@gmail.com>
Subject: Re: Catching up with performance & PostgreSQL 15
From: Mladen Gogala <gogala.mladen@gmail.com>
Reply-To: gogala.mladen@gmail.com
To: Alvaro Herrera <alvherre@alvh.no-ip.org>
Cc: pgsql-performance@lists.postgresql.org
Date: Tue, 29 Nov 2022 19:44:23 -0500
In-Reply-To: <20221129180914.pdrqrbqtls3g6yoq@alvherre.pgsql>
References: <20221129180914.pdrqrbqtls3g6yoq@alvherre.pgsql>
Content-Type: multipart/alternative; boundary="=-8Kw0Z9dGgAsBeUVbrmRF"
User-Agent: Evolution 3.46.1 (3.46.1-1.fc37) 
MIME-Version: 1.0
Archived-At: 
 <https://www.postgresql.org/message-id/bcd915ceb9bb65f1a4e00b673371071c63146a4f.camel%40gmail.com>
Precedence: bulk

--=-8Kw0Z9dGgAsBeUVbrmRF
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

On Tue, 2022-11-29 at 19:09 +0100, Alvaro Herrera wrote:
> On 2022-Nov-29, Mladen Gogala wrote:
>=20
> > Hmmm, I think I will run pgbench with and without JIT on and see
> > the
> > difference.
>=20
> I doubt you'll notice anything, because the pgbench queries will be
> far
> below the JIT cost, so nothing will get JIT compiled at all.=C2=A0 Or are
> you
> planning on using a custom set of queries?
>=20

Nope. I am planning to set jit_above_cost parameter to 5. That should
take care of the pgbench problem. Other than that, you're right: JIT
should not be used for OLTP. However, pure OLTP or DW databases are a
rarity these days. Reporting is a crucial function and almost every
OLTP database that I've seen also has reporting function, which means
that there are complex queries to be executed.
--=20
Mladen Gogala
Database Consultant
https://dbwhisperer.wordpress.com


--=-8Kw0Z9dGgAsBeUVbrmRF
Content-Type: text/html; charset="utf-8"
Content-Transfer-Encoding: quoted-printable

<html><head></head><body><div>On Tue, 2022-11-29 at 19:09 +0100, Alvaro Her=
rera wrote:</div><blockquote type=3D"cite" style=3D"margin:0 0 0 .8ex; bord=
er-left:2px #729fcf solid;padding-left:1ex"><div>On 2022-Nov-29, Mladen Gog=
ala wrote:<br></div><div><br></div><blockquote type=3D"cite" style=3D"margi=
n:0 0 0 .8ex; border-left:2px #729fcf solid;padding-left:1ex"><div>Hmmm, I =
think I will run pgbench with and without JIT on and see the<br></div><div>=
difference.<br></div></blockquote><div><br></div><div>I doubt you'll notice=
 anything, because the pgbench queries will be far<br></div><div>below the =
JIT cost, so nothing will get JIT compiled at all.&nbsp; Or are you<br></di=
v><div>planning on using a custom set of queries?<br></div><div><br></div><=
/blockquote><div><br></div><div>Nope. I am planning to set jit_above_cost p=
arameter to 5. That should take care of the pgbench problem. Other than tha=
t, you're right: JIT should not be used for OLTP. However, pure OLTP or DW =
databases are a rarity these days. Reporting is a crucial function and almo=
st every OLTP database that I've seen also has reporting function, which me=
ans that there are complex queries to be executed.</div><div><span><pre>-- =
<br></pre><div>Mladen Gogala</div><div>Database Consultant</div><div><a hre=
f=3D"https://dbwhisperer.wordpress.com">https://dbwhisperer.wordpress.com</=
a></div><div><br></div></span></div></body></html>

--=-8Kw0Z9dGgAsBeUVbrmRF--