DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=beta; d=gmail.com;
	h=received:message-id:date:from:to:subject:mime-version:content-type:content-transfer-encoding:content-disposition;
	b=fEYuGqQL0ZRsPPrczbV+7UJF/2ZxaolXhFFZPzFK8qf2mmCCOaBkfjdSPz5bJMKtH9JYhIPNGVdvAS4NmI66eBFa0PaJwdWgJOy2Eswx0l8IR/AiQaxPoKOrf5Y3THmRa34Sd2R5X1pSR1bEQV1ten4sPF7E8JgnM4epb8OsxeQ=
Message-ID: <b42b73150610091117n29e8d9d5yed07abe8a228593e@mail.gmail.com>
Date: Mon, 9 Oct 2006 14:17:24 -0400
From: "Merlin Moncure" <mmoncure@gmail.com>
To: "Pg Performance list" <pgsql-performance@postgresql.org>
Subject: odd variances in count(*) times
MIME-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit
Content-Disposition: inline

I have two systems running 8.2beta1 getting strange difference of
results in count(*).  Query that illistrates the difference is
count(*).  this is a synthetic test i use to measure a sytems's cpu
performance.

System A:
2.2 ghz p4 northwood, HT
win xp
vanilla sata (1 disk)

System B:
amd 64 3700+
linux cent/os 4.4 32 bit
4 raptors, raid 5, 3ware

explain analyze select 5000!;
A: 2.4 seconds
B: 1.8 seconds

explain analyze select count(*) from generate_series(1,500000);
A: 0.85 seconds
B: 4.94 seconds

Both systems have a freshly minted database.  By all resepcts I would
expect B to outperform A on most cpu bound tests with a faster
processor and linux kernel.  memory is not an issue here, varying the
size of the count(*) does not effect the results, A is always 5x
faster than B.  the only two variables i see are cpu and o/s.

Also tested on pg 8.1, results are same except pg 8.2 is about 10%
faster on both systems for count(*).  (yay!) :-)

anybody think of anything obvious? should i profile? (windows mingw
profiling sucks)

merlin