From: Josh Berkus <josh@agliodbs.com>
Reply-To: josh@agliodbs.com
Organization: Aglio Database Solutions
To: pgsql-performance@postgresql.org
Subject: Re: PG8 Tuning
Date: Wed, 17 Aug 2005 13:39:47 -0700
User-Agent: KMail/1.8
Cc: Michael Stone <mstone+postgres@mathom.us>
References: <1536.217.45.209.171.1123763001.squirrel@www.gradwell.com>
	<200508160912.32263.josh@agliodbs.com>
	<20050816173047.GC19080@mathom.us>
In-Reply-To: <20050816173047.GC19080@mathom.us>
MIME-Version: 1.0
Content-Type: text/plain;
  charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
Content-Disposition: inline
Message-Id: <200508171339.47399.josh@agliodbs.com>

Michael,

> Well, you don't have to spend *quite* that much to get a decent storage
> array. :)

Yes, I'm just pointing out that it's only the extreme cases which are 
clear-cut.  Middle cases are a lot harder to define.  For example, we've 
found that on DBT2 running of a 14-drive JBOD, seperating off WAL boosts 
performance about 8% to 14%.  On DBT3 (DSS) seperate WAL (and seperate 
tablespaces) helps considerably during data load,but not otherwise.  So it 
all depends.

> That's a different creature from
> a data mining app that might really benefit from having additional
> spindles to accelerate read performance from indices much larger than
> RAM. 

Yes, although the data mining app benefits from the special xlog disk 
during ETL.  So it's a tradeoff.  

> At any rate, this just underscores the need for testing a 
> particular workload on particular hardware

Yes, absolutely.

-- 
--Josh

Josh Berkus
Aglio Database Solutions
San Francisco