Re: Re: SQL performance in prod

From: <l.flatz_at_bluewin.ch>
Date: Thu, 26 Jul 2018 09:41:58 +0200 (CEST)
Message-ID: <1745016935.6135.1532590918583.JavaMail.webmail_at_bluewin.ch>





Andy,
for a nested loop join the estimates in the outer loop are per execution, but on the join level (the NL Operation) the estimates reflects the result of all executions. Regards
Lothar
----Ursprüngliche Nachricht----
Von : andysayer_at_gmail.com
Datum : 26/07/2018 - 09:28 (GMT)
An : gogala.mladen_at_gmail.com
Cc : oracle-l_at_freelists.org
Betreff : Re: SQL performance in prod
Mladen,
“My assumption was that the plans for a single execution are different. I maybe wrong.”
I was referring to the execution of a single line in the plan. eg for a nested loop: there will be multiple executions of the plan lines that hit the joined table (assuming multiple rows in the driving table), the e-rows represents the number of rows the CBO expects to find on one iteration. Assuming one statement execution. A-rows will be the total number of rows flowing from that row source from the entire execution of the statement. So you would have to multiply the E-rows by the number of times the line was hit Hopefully that’s clearer now
Andy

--

http://www.freelists.org/webpage/oracle-l Received on Thu Jul 26 2018 - 09:41:58 CEST

Original text of this message