Date: Sun, 28 Sep 2014 15:30:15 -0300 (ART)
From: Gerardo Herzig <gherzig@fmed.uba.ar>
To: pgsql-general@postgresql.org, pgsql-sql <pgsql-sql@postgresql.org>
Message-ID: <796319273.990931.1411929015224.JavaMail.root@fmed.uba.ar>
In-Reply-To: <784256159.990526.1411927128224.JavaMail.root@fmed.uba.ar>
Subject: how to see "where" SQL is better than PLPGSQL
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: quoted-printable
Precedence: bulk
Sender: pgsql-sql-owner@postgresql.org

Hi all. I see an entire database, with all the stored procedures writen in =
plpgsql. Off course, many (if not all) of that SP are simple inserts, updat=
es, selects and so on.

So, i want to test and show the differences between doing the same function=
 in pgpgsql vs. plain sql.
Im getting statistics (via collectd if that matters) and doing a modified v=
ersion of the pgbench tests, just using pl (and sql) functions instead of t=
he plain query:

\setrandom delta -5000 5000
BEGIN;
SELECT pgbench_accounts_upd_pl(:delta, :aid);
SELECT get_pgbench_accounts_pl(:aid);
SELECT pgbench_tellers_upd_pl(:delta, :tid);
SELECT pgbench_branches_upd_pl(:delta, :bid);
select pgbench_history_ins_pl(:tid, :bid, :aid, :delta);
END;

At first, pgbench is showing a difference between the "pl" and de "sql" ver=
sions:

(pl.scripts own the "PL" version, sql.script owns the "SQL" version of the =
test)
(This is a tiny netbook, with a dual core procesor)

gherzig@via:~> pgbench -c 2 -C -T 300 -f pl.script -U postgres test
duration: 300 s
number of transactions actually processed: 13524
tps =3D 45.074960 (including connections establishing)
tps =3D 75.260741 (excluding connections establishing)

gherzig@via:~> pgbench -c 2 -C -T 300 -f sql.script -U postgres test
starting vacuum...end.
duration: 300 s
number of transactions actually processed: 15125
tps =3D 50.412852 (including connections establishing)
tps =3D 92.058245 (excluding connections establishing)

So yeah, it looks like the "SQL" version is able to do a 10% more transacti=
ons.=20
However, i was hoping to see anothers "efects" of using sql (perhaps less l=
oad avg in the SQL version), at the OS level.=20

So, finnaly, the actual question:
=C2=BFWich signals should i monitor, in order to show that PGPLSQL uses mor=
e resources than SQL?

Thanks!
Gerardo


--=20
Sent via pgsql-sql mailing list (pgsql-sql@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-sql