On Thursday 07 July 2011 19:54:08 French, Martin wrote:
> How up to date are the statistics for the tables in question?
>
> What value do you have for effective cache size?
>
> My guess would be that planner thinks the method it is using is right
> either for its current row number estimations, or the amount of memory
> it thinks it has to play with.
Not very up to date I'm afraid (as shown by the low estimate of deleted rows).
Table t2 has been insert-only since its re-creation (that's another story),
while t1 is your classic insert-many, update-recent.
We haven't tweaked effective cache size yet, it's on the TODO... like many
other things :/
--
Vincent de Phily