Subject: Re: SELECT LIMIT 1 VIEW Performance Issue
From: Simon Riggs <simon@2ndquadrant.com>
To: K C Lau <kclau60@netvigator.com>
Cc: pgsql-performance@postgresql.org
In-Reply-To: <6.2.1.2.0.20050922221846.02c45fb8@localhost>
References: <6.2.1.2.0.20050615100524.04f9d6a8@localhost>
	<6.2.1.2.0.20050922102035.07bcf2e0@localhost>
	<1127378456.4145.76.camel@localhost.localdomain>
	<6.2.1.2.0.20050922181009.02c36ca8@localhost>
	<1127393284.4145.109.camel@localhost.localdomain>
	<6.2.1.2.0.20050922221846.02c45fb8@localhost>
Content-Type: text/plain
Organization: 2nd Quadrant
Date: Thu, 22 Sep 2005 17:16:23 +0100
Message-Id: <1127405783.4145.113.camel@localhost.localdomain>
Mime-Version: 1.0
Content-Transfer-Encoding: 7bit

On Thu, 2005-09-22 at 22:39 +0800, K C Lau wrote:
> >Is a 51ms query really such a problem for you?
> 
> Unfortunately yes, as our target performance is in the high hundreds of 
> transactions per sec. And 51 ms is already the best case for a single 
> select, with everything cached in memory immediately after the same select 
> which took 390 ms on a quiet system.

If the current value is used so often, use two tables - one with a
current view only of the row maintained using UPDATE. Different
performance issues maybe, but at least not correlated subquery ones.

Best Regards, Simon Riggs