From: Nicolai Petri <nicolai@catpipe.net>
Organization: catpipe Systems ApS
To: pgsql-hackers@postgresql.org
Subject: Re: More thoughts about planner's cost estimates
Date: Fri, 2 Jun 2006 11:01:30 +0200
User-Agent: KMail/1.9.1
Cc: Josh Berkus <josh@agliodbs.com>,
 Tom Lane <tgl@sss.pgh.pa.us>
References: <14821.1149104747@sss.pgh.pa.us> <447F23A9.1090503@agliodbs.com>
In-Reply-To: <447F23A9.1090503@agliodbs.com>
MIME-Version: 1.0
Content-Type: text/plain;
  charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
Content-Disposition: inline
Message-Id: <200606021101.31487.nicolai@catpipe.net>

Hi All,

Just a small comment from a mortal user.

On Thursday 01 June 2006 19:28, Josh Berkus wrote:
> 5. random_page_cost (as previously discussed) is actually a funciton of
> relatively immutable hardware statistics, and as such should not need to
> exist as a GUC once the cost model is fixed.
It's correct that the hardware statistics are quite immutable - per 
tablespace. This suggests to me that the GUC should be default value only and 
then overridable per tablespace. It's quite normal to have primary (current 
data) and secondary (historical data) storage for larger databases.
This can also fit nicely into the pg_tmp storage if tie support for multiple 
tmp folders together with tablespaces.

>
> 6. We haven't added any way to estimate rows returned from SRFs.
This would also be very cool - currently the planner can really get annoyed 
when joining SRF functions with tables.

---
Nicolai