RE: Query tuning exercise: what to look for in a 10053 trace
- From: Bernard Polarski <bpolarsk@xxxxxxxxx>
- To: oracle-l@xxxxxxxxxxxxx
- Date: Tue, 25 Jul 2006 23:27:31 -0700 (PDT)
Here we find again the holy trinity : DUC
Diagnose-Understand-Cure
D : Why is my DB so slow --> Oracle wait interface (Oracle press)
U : Why the CBO is doing that --> Cost based optimisation (JL)
C : What should I do now --> Dan Tow, SQL tuning (Oreilly)
You many not be a fan of SQL diagraming, but it is the only systematic
approach that tries to get aways from the try/error scheme while it gives a
methodology to join complex queries. I read Garrison book on SQL tuning and
was very disapointed. There are little to help you when you are faced with the
brute fact of an SQL with 20 tables and 40 joins/filter while with Dan Tow
method I still have a general plan of what to do, goal to reach. At this
stage, JL book will help me understand why the optimiser ignored my will and
correct.
B.Polarski
http://www.smenu.org
From: Schultz, Charles [mailto:sac@xxxxxxxxxxxxx]
Sent: Tuesday, 25 July, 2006 9:05 PM
To: Allen, Brandon; oracle-l@xxxxxxxxxxxxx
Subject: RE: Query tuning exercise: what to look for in a 10053 trace
That makes sense. Since the long-running query has not yet completed, it is
rather hard to know the "real" number of rows at each step (as opposed to the
estimate). The cardinality reported by autotrace (and the 10053 trace) do not
differ by much, but I have to assume that this little difference is having a
huge impact on the ordering. Perhaps the straw that broke the camel's back.
Thanks for the feedback.
From: Allen, Brandon [mailto:Brandon.Allen@xxxxxxxxxxx]
Sent: Tuesday, July 25, 2006 1:34 PM
To: Schultz, Charles; oracle-l@xxxxxxxxxxxxx
Subject: RE: Query tuning exercise: what to look for in a 10053 trace
Yep, read them both - and saw his presentation at Hotsos :-) I'm not sure
on this, but I think the join order will be a result of the estd cardinalities
as well - with a NL join, it will start with the row source that it expects to
return the smallest number of rows, then probe the other table for matching
rows. So, if it underestimates the cardinality, then it will end up doing a
lot more probing than it expected to do.
From: Schultz, Charles [mailto:sac@xxxxxxxxxxxxx]
Sent: Tuesday, July 25, 2006 11:25 AM
To: Allen, Brandon; oracle-l@xxxxxxxxxxxxx
Subject: RE: Query tuning exercise: what to look for in a 10053 trace
I see you have been reading his "tuning by cardinality feedback" paper. Or
perhaps "under the hood..."? =)
I have read them, but it takes a long time for me to absorb the wonderful
insights these guys have.
From an academic standpoint, I am trying to figure out why the CBO is using
a different join order - perhaps this is a vain and useless endeavor, but I am
curious.
Privileged/Confidential Information may be contained in this message or
attachments hereto. Please advise immediately if you or your employer do not
consent to Internet email for messages of this kind. Opinions, conclusions and
other information in this message that do not relate to the official business
of this company shall be understood as neither given nor endorsed by it.
Other related posts:
