Chris Bitmead <chrisb@nimrod.itg.telstra.com.au> writes:
> On the postgres side, what if you actually made the cache _smaller_,
> only caching important stuff like system tables or indexes. You yourself
> said it doesn't matter if you get it from the cache or the kernel, so
> why not let the kernel do it and prevent double buffering?
In fact I don't think it's productive to use an extremely large -B
setting. This is something that easily could be settled by experiment
--- do people actually see any significant improvement in performance
from increasing -B above, say, a few hundred?
regards, tom lane