[Tile-serving] osm2pgsql, -G and multipolygons

[Paul Norman]

I hope no one minds my announcements of me having written up my 
benchmarking work, but I figure this is the best place to announce
 it and stimulate discussion.

I wrote up some of my benchmarking on the -G option for osm2pgsql, 
and what its splitting MULTIPOLYGONs into POLYGONs does: 
http://www.paulnorman.ca/blog/2014/03/osm2pgsql-multipolygons/

The short version is it decreases performance by 5.5% overall. 
It has minimal impact on zooms 13 and above, but has more impact 
on the higher zooms, of which most of the workload is.

I'm doing some investigations to see if there are ways to recover 
some of the lost performance, but nothing to report yet.