Greetings,
* Vincenzo Romano (vincenzo.romano@notorand.it) wrote:
> Sorry, my bad: I confused V10 with v11.
> But accordingly to a discussion with Bruce Momjan, table partitionin V10 is
> little more than syntactic sugar around old-fashioned table partitioning.
Well, it's a bit more than that since there's tuple-routing, but you're
right that the partition elimination is the same as it was in earlier
versions and based on constraint exclusion. That said, as noted in the
email you replied to, reasonable numbers of partitions aren't too bad
even with the planning cost; it's when you have many thousands of
partitions that you get into cases where planning time for queries is
really bad.
Also as noted on this thread, PG could handle this data volume, but to
be efficient there would be work to be done in normalization,
aggregation, and analyzing the system to ensure you're storing and
querying on the data efficiently.
Thanks!
Stephen