On Wed, Aug 09, 2006 at 04:54:00PM -0400, Tom Laudeman wrote:
> I'll try CLUSTER (I'm looking forward to that test), but if we really
> need speed, it will probably be necessary to create copies of the table,
> or copy portions of the table elsewhere (essentially creating
> materialized views, I suppose). I'm still trying to get my science
> compatriot here to tell me which index he most wants to improve, then
> I'll CLUSTER the table on that index.
If you enable statistics collection then you could use those
statistics to see which indexes are used the most. Those indexes
might be good candidates for clustering.
http://www.postgresql.org/docs/8.1/interactive/monitoring-stats.html
--
Michael Fuhr
