Re: Partitioning into thousands of tables?

Joshua Tolley <eggyknap@gmail.com> writes:
> On Fri, Aug 06, 2010 at 03:10:30PM +1000, Data Growth Pty Ltd wrote:
>> Is there any significant performance problem associated with partitioning
>> a table into 2500 sub-tables?  I realise a table scan would be horrendous,
>> but what if all accesses specified the partitioning criteria "sid".  Such
>> a scheme would be the simplest to maintain (I think) with the best
>> localisation of writes.

> I seem to remember some discussion on pgsql-hackers recently about the number
> of partitions and its effect on performance, especially planning time.
> Unfortunately I can't find it right now, but in general the conclusion was
> it's bad to have lots of partitions, where "lots" is probably 100 or more.

It's in the fine manual: see last para of
http://www.postgresql.org/docs/8.4/static/ddl-partitioning.html#DDL-PARTITIONING-CAVEATS

            regards, tom lane