bgwriter

I understand that the bgwriter code will not be changed until 8.1 (I'm
referring to this issue:
<http://archives.postgresql.org/pgsql-hackers/2004-12/msg00733.php> ).

After reading through the thread, it's my impression that there exists
some performance problem with the current method of using a percent of
dirty buffers as a GUC parameter under some (but not all) situations.

So, what situations should I watch out for to avoid the performance
problem with 8.0? Is it actually a performance regression from 7.4, or
just not as good as we hoped? Is it a relatively minor issue, or can it
really bog down the server in some situations?

Regards,
Jeff Davis