Re: Info about statpack analysis

On Tue, 25 Oct 2005 20:51:10 +0200, "Michel Cadot" <micadot{at}altern{dot}org> wrote:

>"evvivaiciccioni" <cicico_at_cicciobello.it> a écrit dans le message de news: 435e74e9$0$8481$5fc30a8_at_news.tiscali.it...
>| Hi,
>| i'm using statpack in order to improve the performance of a Oracle 9.2.0.6
>| Server on HPUX, in the report, i found this:
>|
>| Buffer Gets Executions Gets per Exec %Total Time (s) Time (s) Hash
>| Value
>| --------------- ------------ -------------- ------ -------- --------- ----------
>| 21,680,275 78 277,952.2 22.9 90.69 90.38
>| 3666391050
>| Module: ? @SSSS (TNS V1-V3)
>| DELETE FROM DDDDD WHERE DDDDDDD.rowid=:"SYS_B_0"
>|
>|
>| I'm confused, infact as far i know, the rowid access should be the fastest
>| method, it contains the physical address of the row (maybe some other
>| consideration may apply in case of RAC, but it is not my case) but this
>| query gets more than 21.000.000 buffers for 78 executions ....... i expect
>| to see... about 78 buffers... (in case of chained row, it may be greather).
>|
>| What's wrong ???????????
>|
>| Thanks.
>|
>|
>
>DELETE FROM DDDDD WHERE DDDDDDD.rowid=:"SYS_B_0"
>is not a valid statement, are you sure you posted exactly what you have?
>
>Regards
>Michel Cadot
>

This is a result of the OP using cursor_sharing=force. This will result in artificial bind variables SYS_B_<n> He needs to use event 10046 on level 12, to see the actual values.

--
Sybrand Bakker, Senior Oracle DBA