Re: postgresql shared buffers

Heikki Linnakangas wrote:
> Praveen Kumar N wrote:
> >Let me explain once more.
> >
> >I have two relations which are 10 times more than bufferpool size.I
> >have observed the following things when joined that two relations(it
> >using merge join to join both relations)
> >
> >1.It first accessed system catalog tables
> >2.Relation 1
> >3.Relation 2
> >
> >my doubt is one whole relation cant fit in the main memory.That too
> >when we use merge join, it should keep some part of 1st relations and
> >should scan second relation as bufferpool size is less compared to
> >size of each relation.similarly for the remainin part of 1st
> >relation.But it is not happening here.First whole Relation1 is scanned
> >and then Relation 2 is scanned. Then how is it joining two relations
> >using merge join? Am I missing something?
>
> Hmm. A hash join, maybe? You should do EXPLAIN on the query to see what
> it really does, otherwise we're just guessing.

Another option would be that the tuples are written to a sort tape, I
think.

--
Alvaro Herrera http://www.CommandPrompt.com/
PostgreSQL Replication, Consulting, Custom Development, 24x7 support