On Mon, May 25, 2009 at 12:10:21AM -0700, Scara Maccai wrote: > is there any chance to get the "Planner Cost Constants" right enough to get a "good" estimate in seconds of how long a query is supposed to run? > The "rowcount" estimates are always good (there is no skew data at all in the db, values are pretty much "plain" distributed) The most straightforward way I can think of would be to make sure your planner constants (seq_page_cost, etc.) reflect time in meaningful units, which would be very hard, and probably wouldn't work for reasons I don't know. Anyway, the quick answer is no, 'cause it's really hard. - Josh / eggyknap
Attachment:
signature.asc
Description: Digital signature
