Re: Strange issue with GiST index scan taking far too long

On Mon, 2008-06-09 at 11:49 -0400, Tom Lane wrote:
> Mark Cave-Ayland <mark.cave-ayland@siriusit.co.uk> writes:
> > So by adding in an extra function around the subselect result, we have 
> > speeded up the index lookup by several orders of magnitude, and the 
> > speedup appears to be coming from somewhere within the index scan?!
> 
> Is the value you are fetching from the geography table large enough to
> be toasted?  I'm thinking you might be looking at the cost of repeated
> de-toastings.

So you are saying it is de-toasted 32880 times, in this case? If not,
where are the repeated de-toastings happening?

Sounds like we need some stats on how well toast is working for us.

-- Simon Riggs           www.2ndQuadrant.comPostgreSQL Training, Services and Support