"Markus Wollny" <Markus.Wollny@computec.de> writes:
>> (RAID-5 writes tend to be very slow.) Consider adding a
>> mirrored pair of disks just for the log file.
> As far as I can see, logfile isn't a problem on debug_level 0 (which I
> always set for production state) because there's hardly anything to
> write except on vacuum analyze (which usually only takes place when
> hardly anybody does anything on the system anyway).
Curt's talking about the WAL log (pg_xlog/), not the postmaster's debug
output.
However, if your problems are with SELECTs and not with update
operations, then the WAL log is not the source of your problems anyway.
regards, tom lane