Alvaro Herrera <alvherre@commandprompt.com> writes:
> Martin Marques escribi�:
>>> seq_page_cost = 5.0 # measured on an arbitrary scale
> What people generally do around here is mess with random_page_cost, and
> leave seq_page_cost alone.
It's also worth pointing out that having seq_page_cost >
random_page_cost is simply not sane, and is very likely to result in
performance problems in other queries.
regards, tom lane