Cédric Villemain<cedric.villemain.debian@gmail.com> wrote:
> 2010/4/23 Robert Haas <robertmhaas@gmail.com>:
>> Since all your data is probably fully cached, at a first cut, I
>> might try setting random_page_cost and seq_page_cost to 0.005 or
>> so, and adjusting effective_cache_size to something appropriate.
>
> that will help worrect the situation, but the planner is loosing
> here I think.
The planner produces a lot of possible plans to produce the
requested results, and then calculates a cost for each. The lowest
cost plan which will produce the correct results is the one chosen.
If your costing factors don't represent the reality of your
environment, it won't pick the best plan for your environment.
-Kevin