Re: Comparisons Oracle 11.1.0.7 to Oracle 10.2.0.4

From: Charles Hooper <hooperc2000_at_yahoo.com>
Date: Thu, 22 Oct 2009 04:27:43 -0700 (PDT)
Message-ID: <2f1b3d5a-4807-4bba-a791-e0973e38fcbd_at_o21g2000vbl.googlegroups.com>



On Oct 22, 5:37 am, "Arne Ortlinghaus" <Arne.Ortlingh..._at_acs.it> wrote:
> I have made some performance comparison tests on test computers and it has
> turned out that the Oracle 11 databases are in general slower compared to
> the Oracle 10 databases by 10 to 30%. The databases are running on Windows
> Vista and Windows 2003 64 Bit. We hope that the release 2 of Oracle 11 gives
> some enhancements because in general there are not many advantages regarding
> new features useable for us.
>
> Has someone made similar comparison results?
>
> Arne Ortlinghaus
> ACS Data Systems

I have tested the performance of 10.2.0.4 and 11.1.0.7 on 64 bit Windows. If I recall the test results correctly, you will find that with STATISTICS_LEVEL set to ALL you will see that 11.1.0.7 is

significantly faster than 10.2.0.4 in all but the most simple SQL
statements - in such a case 10.2.0.4 might consume a lot of system
(kernel) mode CPU time, while 11.1.0.7 consumes considerably less.

While testing 10.2.0.4 and 11.1.0.7, I found that 10.2.0.4 was typically slightly faster than 11.1.0.7 when the exact same execution plan was obtained for a SQL statement and the STATISTICS_LEVEL was set to TYPICAL. 11.1.0.7 occasionally found a different execution plan than 10.2.0.4, partially influenced by the presence (or absence) of system WORKLOAD statistics.

If you would, please determine if the execution plans are identical between the two Oracle releases. You might also check the system statistics in SYS.AUX_STATS$, and the various initialization parameters to make certain you are comparing the two Oracle releases with the same inputs. The location of the data files for the two releases could also influence performance - more data passes under the read-write heads per revolution on the outer tracks of the hard disks, so that could yield better performance for the Oracle release with data files located closer to the outer hard disk tracks.

What do you find is slower? Are you checking the execution time of a process with a stop watch, checking the wait events, monitoring the CPU utilization, creating a Statspack Report, or something else?

Charles Hooper
IT Manager/Oracle DBA
K&M Machine-Fabricating, Inc. Received on Thu Oct 22 2009 - 06:27:43 CDT

Original text of this message