Shridhar Daithankar <shridhar_daithankar@persistent.co.in> writes:
> Also if you have fast disk drives, you can reduce random page cost to 2 or 1.5.
Note however that most of the people who have found smaller
random_page_cost to be helpful are in situations where most of their
data fits in RAM. Reducing the cost towards 1 simply reflects the fact
that there's no sequential-fetch advantage when grabbing data that's
already in RAM.
When benchmarking with data sets considerably larger than available
buffer cache, I rather doubt that small random_page_cost would be a good
idea. Still, you might as well experiment to see.
regards, tom lane