To: Oliver Crosby <ryusei@gmail.com>
Cc: "Joshua D. Drake" <jd@commandprompt.com>, pgsql-performance@postgresql.org
Subject: Re: Looking for tips 
In-reply-to: <1efd553a050719112126b8d981@mail.gmail.com> 
References: <1efd553a05071910427b290203@mail.gmail.com>
	<42DD3D55.9020508@commandprompt.com>
	<1efd553a050719112126b8d981@mail.gmail.com>
Comments: In-reply-to Oliver Crosby <ryusei@gmail.com>
	message dated "Tue, 19 Jul 2005 14:21:22 -0400"
Date: Tue, 19 Jul 2005 15:01:00 -0400
Message-ID: <8446.1121799660@sss.pgh.pa.us>
From: Tom Lane <tgl@sss.pgh.pa.us>

Oliver Crosby <ryusei@gmail.com> writes:
> The queries are all simple insert or select statements on single tables.
> Eg. select x from table where y=?; or insert into table (a, b, c)
> values (?, ?, ?);
> In the case of selects where it's a large table, there's an index on
> the column being searched, so in terms of the example above, x is
> either a pkey column or other related field, and y is a non-pkey
> column.

If you're running only a single query at a time (no multiple clients),
then this is pretty much the definition of a MySQL-friendly workload;
I'd have to say we are doing really well if we are only 50% slower.
Postgres doesn't have any performance advantages until you get into
complex queries or a significant amount of concurrency.

You could possibly get some improvement if you can re-use prepared plans
for the queries; but this will require some fooling with the client code
(I'm not sure if DBD::Pg even has support for it at all).

			regards, tom lane