On Mon, 14 Mar 2005 21:23:29 -0500, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> I think that the "reduce random_page_cost" mantra
>is not an indication that that parameter is wrong, but that the
>cost models it feeds into need more work.
One of these areas is the cost interpolation depending on correlation.
This has been discussed on -hackes in October 2002 and August 2003
("Correlation in cost_index()"). My Postgres installations contain the
patch presented during that discussion (and another index correlation
patch), and I use *higher* values for random_page_cost (up to 10).
Servus
Manfred
