Porell, Chris wrote:
> Hi All,
> I've changed shared_buffers, checkpoint_segments, effective_cache_size and
> random_page_cost in an attempt to improve performance. That has helped a
> little...
Another thought. Have you looked at "work_mem" - this is probably a far
more important setting.
The units are kilobytes and you can set this on a session-by-session
basis to test without needing to update postgresql.conf or reload the
server by issuing "set work_mem to xxx;" I have individual nightly
batch-runs where I greatly improved the processing time by setting
work_mem to values in the range of 100000-500000 as needed.
Note that if you are comparing your old configuration to your new one,
this parameter used to be called "sort_mem".
Cheers,
Steve