[Just a quick note here; a more thorough discussion of my test results
will be posted to -hackers]
On Tue, 13 Apr 2004 15:18:42 -0400, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>Well, the first problem is why is ANALYZE's estimate of the total row
>count so bad :-( ? I suspect you are running into the situation where
>the initial pages of the table are thinly populated and ANALYZE
>mistakenly assumes the rest are too. Manfred is working on a revised
>sampling method for ANALYZE that should fix this problem
The new method looks very promising with respect to row count
estimation: I got estimation errors of +/- 1% where the old method was
off by up to 60%. (My test methods might be a bit biased though :-))
My biggest concern at the moment is that the new sampling method
violates the contract of returning each possible sample with he same
probability: getting several tuples from the same page is more likely
than with the old method.
Servus
Manfred