Nassib Nassar <nassar(at)renci(dot)org> wrote:
> In this example it looks to me like the planner is choosing a Seq
> Scan resulting in 18x running time compared to running it with
> enable_seqscan = 'off'.
I would try these settings:
random_page_cost = 2
cpu_tuple_cost = 0.02
Based on your estimated cost versus actual run times, there's a good
chance they'll better model your environment, and serve you well in
general.
-Kevin