From: Ron Mayer <rm_pg@cheapcomplexdevices.com>
Subject: Re: More thoughts about planner's cost estimates
Date: Tue, 06 Jun 2006 17:32:03 -0700
Organization: Hub.Org Networking Services
Lines: 21
Message-ID: <44861E83.9070403@cheapcomplexdevices.com>
References: <14821.1149104747@sss.pgh.pa.us> <11319.1149287800@sss.pgh.pa.us>
Mime-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit
To: Tom Lane <tgl@sss.pgh.pa.us>
User-Agent: Thunderbird 1.5.0.4 (Windows/20060516)
In-Reply-To: <11319.1149287800@sss.pgh.pa.us>
To: pgsql-hackers@postgresql.org

Tom Lane wrote:
> 
> One objection to this is that after moving "off the gold standard" of
> 1.0 = one page fetch, there is no longer any clear meaning to the
> cost estimate units; you're faced with the fact that they're just an
> arbitrary scale.  I'm not sure that's such a bad thing, though.

It seems to me the appropriate gold standard is Time, in microseconds
or milliseconds.

The default postgresql.conf can come with a set of hardcoded
values that reasonably approximate some real-world system; and
if that's documented in the file someone reading it can say
  "hey, my CPU's about the same but my disk subsystem is much
  faster, so I know in which direction to change things".
And another person may say "ooh, now I know that my 4GHz
machines should have about twice the number here as my 2GHz
box".

For people who *really* care a lot (HW vendors?), they could
eventually make measurements on their systems.