I did a ps ax | postmaster but found no -B, and concluded that it uses the
value specified in /etc/postgrelsql/postgresql.conf on shared_buffers (I
saw -B was shared buffer doing a man postmaster). I'll change this to 256
and rerun the test!
Will post the results here later. Please tell if this was a too puny
increase!
Daniel Åkerud
> =?iso-8859-1?Q?Daniel_=C5kerud?= <zilch@home.se> writes:
> > Anyone know what causes the signifacant performance decrease after 7
indexe=
> > s?
>
> I'd bet that somewhere around there, you are starting to see thrashing
> of the buffer pool due to needing to touch too many different pages to
> insert each tuple. What is your -B setting? If you increase it,
> does the performance improve?
>
> regards, tom lane
>