Hi,
increasing shared_buffers has improved *a lot* the number of inserts/second,
so my "problem" [1] is fixed.
But now I'm worried because of the sentence (Tom Lane):
"The partitioning code isn't designed to scale beyond a few dozen partitions"
Is it mainly a planning problem or an execution time problem?
I did a (very quick) test with 3000 empty partitions, and it took 0.5 secs to do
the planning for a simple select with a very simple where condition.
With 300 partitions, planning takes about 30ms.
That's fine for my case, as I don't expect more than 300 partitions; and I could
actually wait for .5 secs more if that helps with such large tables, and I won't
be doing joins.
So: the "scaling" problem would be more evident in case joins were taken into
account? Or there's something else I didn't get?
[1] http://archives.postgresql.org/pgsql-general/2010-04/msg00611.php