Re: Performance of MERGE INTO statement

From: 조동욱 <ukja.dion_at_gmail.com>
Date: Mon, 19 May 2008 08:49:41 +0900
Message-ID: <43c2e3d60805181649m24321483o59acaea4b1af82b6@mail.gmail.com>


>
> Finn - I'm not sure why the optimizer thinks that there is only one row.
> I've gathered stats and the NUM_ROWS column shows 3,041,890. There are
> presently two partitions on this table. USER_TAB_PARTITIONS shows that
> one has 1,078 rows and the other has 4,089,540 rows (it's partitioned by
> month and I've been testing it with one month of data).
>
> * | 20 | TABLE ACCESS FULL | CALL_FACTS | 1 | 66
> | 1358 | KEY | KEY |
>

You did not post the predicate information. (where call_facts.c1 = ... ) Oracle applied some basic formulars for your predicate and the result was "1"(or 0).

Predicate part of the execution plan is essential when interpreting the cardinality and cost.

Dion Cho

--
http://www.freelists.org/webpage/oracle-l
Received on Sun May 18 2008 - 18:49:41 CDT

Original text of this message