On Wed, 3 Dec 2008, Mark Wong wrote:
http://207.173.203.223/~markwkm/pgsql/default_statistics_target/q2.png
http://207.173.203.223/~markwkm/pgsql/default_statistics_target/q9.png
http://207.173.203.223/~markwkm/pgsql/default_statistics_target/q17.png
http://207.173.203.223/~markwkm/pgsql/default_statistics_target/q18.png
http://207.173.203.223/~markwkm/pgsql/default_statistics_target/q20.png
I've updated the charts to include results from setting
default_statistics_target from 20-90.
Interesting.
Q2: Explodes hard when the target hits 100, but not before. Would be
interesting to compare the explain plans at, say, 50 and 150 to see if
it's possible to nail down what is shifting so much.
Q3: goes bonkers as soon as the target hits 20, so the interesting plans
to compare are 10 and 20.
Q17: our one from the original set that improved a bunch with the larger
target gets all that behefit just from going to 20. Would be interesting
to compare the plans at 10 and 20 to see what changed so much with such a
small difference.
Q18: looks like it was in some sort of local bad area around 100-400
before, with some more context that one doesn't look interesting anymore.
Q20: also doesn't look very interesting anymore. The results at 10 were
nice, and the ones at 100 were among the unusuallly bad ones, but it's
pretty random--if there was something inherantly bad related to the stats,
there wouldn't be low points around 200.
Out of those, the most interesting one to me (as someone who is trying to
defend raising the target some but not going crazy with that) is Q3. The
reason I say that is that everything else is better or basically the same
raising the target from 10, as long as you don't go too high (>=100).
That one falls apart immediately with a larger target which seems weird.
--
* Greg Smith [EMAIL PROTECTED] http://www.gregsmith.com Baltimore, MD
--
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers
