On Wed, 2 Oct 2002, Manfred Koizar wrote:
> As nobody knows how each of these proposals performs in real life
> under different conditions, I suggest to leave the current
> implementation in, add all three algorithms, and supply a GUC variable
> to select a cost function.
I'd certainly be willing to do some testing on my own data with them.
Gotta patch? I've found that when the planner misses, sometimes it misses
by HUGE amounts on large tables, and I have been running random page cost
at 1 lately, as well as running cpu_index_cost at 1/10th the default
setting to get good results.