Comments for Striving for Optimal PerformanceStriving for Optimal Performance http://www.antognini.ch Tue, 08 May 2012 17:38:55 +0000 hourly 1 Comment on Scripts to Download Documentation by Christian Antognini http://www.antognini.ch/2011/04/scripts-to-download-documentation/#comment-482 Christian Antognini Tue, 08 May 2012 17:38:55 +0000 http://antognini.ch/?p=1502#comment-482 Hi Johannes "Troubleshooting Oracle Performance", see http://antognini.ch/top/ Cheers, Chris Hi Johannes

“Troubleshooting Oracle Performance”, see http://antognini.ch/top/

Cheers,
Chris

]]> Comment on Scripts to Download Documentation by Johannes http://www.antognini.ch/2011/04/scripts-to-download-documentation/#comment-481 Johannes Tue, 08 May 2012 17:35:50 +0000 http://antognini.ch/?p=1502#comment-481 What is the name of the oracle book eddy mentioned? What is the name of the oracle book eddy mentioned?

]]> Comment on COMMIT_WAIT and COMMIT_LOGGING by Hemant K Chitale http://www.antognini.ch/2012/04/commit_wait-and-commit_logging/#comment-537 Hemant K Chitale Mon, 09 Apr 2012 06:45:13 +0000 http://antognini.ch/?p=1766#comment-537 Useful post. Thanks ! Hemant Useful post. Thanks !
Hemant

]]> Comment on COMMIT_WAIT and COMMIT_LOGGING by Jonathan Lewis http://www.antognini.ch/2012/04/commit_wait-and-commit_logging/#comment-536 Jonathan Lewis Fri, 06 Apr 2012 18:08:10 +0000 http://antognini.ch/?p=1766#comment-536 A couple more little points to watch out for: a) when you change commit_logging from immediate to batch Oracle changes the content of the redo entries it generates. (You can see this in the stats "redo entries" and "redo size" - and confirm it with a log file dump. b) the standard pl/sql commit is NOT the same as any of the four combinations available through these two parameters; there is a fifth code path. (ditto re stats and dumps) A couple more little points to watch out for:
a) when you change commit_logging from immediate to batch Oracle changes the content of the redo entries it generates. (You can see this in the stats “redo entries” and “redo size” – and confirm it with a log file dump.
b) the standard pl/sql commit is NOT the same as any of the four combinations available through these two parameters; there is a fifth code path. (ditto re stats and dumps)

]]> Comment on Troubleshooting Oracle Performance by Oracle Books Recomented by Jonathan Lewis | shanojkumar http://www.antognini.ch/top/#comment-25 Oracle Books Recomented by Jonathan Lewis | shanojkumar Fri, 06 Apr 2012 14:24:03 +0000 http://antognini.ch/wordpress/?page_id=3#comment-25 [...] about how to identify and address performance problems. I think I’d probably rate this as the top Oracle tuning book on the market at present (UK link) (US [...] [...] about how to identify and address performance problems. I think I’d probably rate this as the top Oracle tuning book on the market at present (UK link) (US [...]

]]> Comment on COMMIT_WAIT and COMMIT_LOGGING by Marcin Przepiorowski http://www.antognini.ch/2012/04/commit_wait-and-commit_logging/#comment-535 Marcin Przepiorowski Fri, 06 Apr 2012 11:00:05 +0000 http://antognini.ch/?p=1766#comment-535 Hi Chris, Did you ever try to set commit_write to BATCH and left commit_wait default value ? My colleague spotted (and I have confirmed) that commit write set to BATCH is invisible changing commit_wait to NOWAIT So in that configuration <pre> SQL> show parameter commit NAME TYPE VALUE ------------------------------------ ----------- ------------------------------ commit_logging string commit_point_strength integer 1 commit_wait string commit_write string BATCH </pre> user process is not waiting for LGWR confirmation (semtimedop) but it is returning commit command immediately even if LGWR process is stopped with kill -SIGSTOP command. There is also no sight of "log file sync" in 10046 trace file. I have tested it in 11.2.0.2 but I'm wondering if it happen as well in your configuration. regards, Marcin Hi Chris,

Did you ever try to set commit_write to BATCH and left commit_wait default value ?
My colleague spotted (and I have confirmed) that commit write set to BATCH is invisible changing commit_wait to NOWAIT

So in that configuration

SQL> show parameter commit

NAME                                 TYPE        VALUE
------------------------------------ ----------- ------------------------------
commit_logging                       string
commit_point_strength                integer     1
commit_wait                          string
commit_write                         string      BATCH

user process is not waiting for LGWR confirmation (semtimedop) but it is returning commit command immediately even if LGWR process is stopped
with kill -SIGSTOP command. There is also no sight of “log file sync” in 10046 trace file.
I have tested it in 11.2.0.2 but I’m wondering if it happen as well in your configuration.

regards,
Marcin

]]> Comment on Parallel Processing With Standard Edition by How does Oracle Database Standard Edition (One) benefit from running on a dual CPU server? - Admins Goodies http://www.antognini.ch/2010/09/parallel-processing-with-standard-edition/#comment-442 How does Oracle Database Standard Edition (One) benefit from running on a dual CPU server? - Admins Goodies Fri, 06 Apr 2012 10:33:56 +0000 http://antognini.ch/?p=1258#comment-442 [...] possible that you could improve load performance by leveraging more CPUs using something like the DBMS_PARALLEL_EXECUTE package or by simply running multiple loads simultaneously. Answered by Justin [...] [...] possible that you could improve load performance by leveraging more CPUs using something like the DBMS_PARALLEL_EXECUTE package or by simply running multiple loads simultaneously. Answered by Justin [...]

]]> Comment on COMMIT_WAIT and COMMIT_LOGGING by Log Buffer #266, A Carnival of the Vanities for DBAs | The Pythian Blog http://www.antognini.ch/2012/04/commit_wait-and-commit_logging/#comment-534 Log Buffer #266, A Carnival of the Vanities for DBAs | The Pythian Blog Fri, 06 Apr 2012 06:01:19 +0000 http://antognini.ch/?p=1766#comment-534 [...] Chris Antongnini blogs about the purpose of COMMIT_WAIT and COMMIT_LOGGING parameters from the real world experience. [...] [...] Chris Antongnini blogs about the purpose of COMMIT_WAIT and COMMIT_LOGGING parameters from the real world experience. [...]

]]> Comment on System Managed Extent Size – 11g Improvements by Autoallocate and PX « Oracle Scratchpad http://www.antognini.ch/2009/08/system-managed-extent-size-11g-improvements/#comment-235 Autoallocate and PX « Oracle Scratchpad Thu, 05 Apr 2012 18:22:12 +0000 http://antognini.ch/?p=537#comment-235 [...] Antognini reports an enhancement arriving (silently) in 11.1.0.7 that addresses this issue. Share this:TwitterLike this:LikeBe the [...] [...] Antognini reports an enhancement arriving (silently) in 11.1.0.7 that addresses this issue. Share this:TwitterLike this:LikeBe the [...]

]]> Comment on COMMIT_WAIT and COMMIT_LOGGING by Christian Antognini http://www.antognini.ch/2012/04/commit_wait-and-commit_logging/#comment-533 Christian Antognini Thu, 05 Apr 2012 16:08:33 +0000 http://antognini.ch/?p=1766#comment-533 Hi Martin As Marcus pointed out, it works as documented. Here is an example where I do not set the parameters: <pre>SQL> show parameter commit NAME TYPE VALUE ------------------------------------ ----------- ------------------------------ commit_logging string commit_point_strength integer 1 commit_wait string commit_write string SQL> !./commit.sh chris ian 2> /dev/null ***** Server Process ***** % time seconds usecs/call calls errors syscall ------ ----------- ----------- --------- --------- ---------------- nan 0.000000 0 8 1 semtimedop ------ ----------- ----------- --------- --------- ---------------- 100.00 0.000000 8 1 total ***** Log Writer ***** % time seconds usecs/call calls errors syscall ------ ----------- ----------- --------- --------- ---------------- 100.00 0.018192 25 737 pwrite ------ ----------- ----------- --------- --------- ---------------- 100.00 0.018192 737 total</pre> Cheers, Chris Hi Martin

As Marcus pointed out, it works as documented. Here is an example where I do not set the parameters:

SQL> show parameter commit

NAME                                 TYPE        VALUE
------------------------------------ ----------- ------------------------------
commit_logging                       string
commit_point_strength                integer     1
commit_wait                          string
commit_write                         string

SQL> !./commit.sh chris ian 2> /dev/null

***** Server Process *****

% time     seconds  usecs/call     calls    errors syscall
------ ----------- ----------- --------- --------- ----------------
   nan    0.000000           0         8         1 semtimedop
------ ----------- ----------- --------- --------- ----------------
100.00    0.000000                     8         1 total

***** Log Writer *****

% time     seconds  usecs/call     calls    errors syscall
------ ----------- ----------- --------- --------- ----------------
100.00    0.018192          25       737           pwrite
------ ----------- ----------- --------- --------- ----------------
100.00    0.018192                   737           total

Cheers,
Chris

]]>