Re: DB CPU is much lower than CPU Time Reported by TOP SQL consumers

  • From: goran bogdanovic <goran00@xxxxxxxxx>
  • To: exriscer@xxxxxxxxx
  • Date: Mon, 26 Nov 2012 00:26:31 +0100

The difference you see IMO is coming from wait time in CPU run-queue - as
Karl has already mentioned.
The reported value by oracle for CPU consumption of statement (or process)
doesn't differentiate if the process was all the time on CPU or on-CPU,
on-CPU-run-queue and then again on-CPU. Oracle just reports let say total
<xzy>ms CPU-time.
Do you have OS statistics (OSWatcher for example) from this period to
compare?
On Tue, Nov 13, 2012 at 9:00 PM, Ls Cheng <exriscer@xxxxxxxxx> wrote:

> Hi
> I also thought of that and in fact one statement was called by another but
> the rest doesnt seem the case:
>
> See following data and attention to the module, they are all different so
> cant be double counting:
>
>
>
> *    CPU                  CPU per             Elapsd
> Old
>   Time (s)   Executions  Exec (s)  %Total   Time (s)    Buffer Gets  Hash
> Value
> ---------- ------------ ---------- ------ ---------- ---------------
> ----------
>    4407.08       20,294       0.22   51.6   10006.66   1,228,369,784
> 3703299877
> Module: JDBC Thin Client
>
>    3943.14      157,316       0.03   46.2    6915.60   1,034,202,723
> 1127338565
> Module: sel_ancomm_vss_06.tsk@c2aixprod (TNS V1-V3)
>
>    2358.20      269,711       0.01   27.6 4095.76 1,508,308,542
> 1995656981
> Module: sel_zuteiler_alert.tsk@c2aixprod (TNS V1-V3)
>
>    1305.21        9,932       0.13   15.3    2483.90         331,327
> 1310406159
> Module: sel_verwaltung.tsk@c2aixprod (TNS V1-V3)*
>
> These 4 statements already adds up to 12013 CPU Seconds and the DB CPU is
> 8464 seconds.
>
> Also look this text from top sql statmenet section:
>
> -> Total DB CPU (s):           8,539
> *-> Captured SQL accounts for  232.5% of Total DB CPU*
> -> SQL reported below exceeded  1.0% of Total DB CPU
>
> Capture SQL is 232% DB CPU! How can this be possible?
>
> Thanks
>
>
> --
> //www.freelists.org/webpage/oracle-l
>
>
>


--
//www.freelists.org/webpage/oracle-l


Other related posts: