Increasing CPU on every execution of an anonymous block sent from WebLogic

Oracle 9.2.0.8 on Solaris 8.  

I've asked development to rewrite Java code that issues SQL with literals so that it uses binds. The new code dynamically constructs an anonymous PL/SQL block with binds, binds values to it, and executes it. The block consists solely of insert statements. I am aware that this would be better to implement this as a packaged procedure, but because this is a heavily-used function of a critical system, development is not willing to move the code from one tier to another.  

Each business transaction executes this block a few hundred times.  

In testing we found that transactions using this code run for over 100 seconds, at which point WebLogic times them out. (In the test, the dynamically-constructed PL/SQL block is identical every time.) The same code implemented with literals in its dynamic PL/SQL anonymous blocks performs just fine.  

Analysis of the trace files for the bind-variable version reveals that the CPU time to EXEC the anonymous block (and, to a lesser extent, its constituent INSERT statements) steadily increases. The first time through the CPU time is about 0.02 second. By the last execution before the timeout, the CPU time is just about 1.0 second. Interestingly, the logical I/O stats for each execution are quite low (<100 current reads and consistent reads) and quite consistent.  

We've made a rough copy of the anonymous block and run it several hundred times from SQL*Plus; this setup doesn't show the increasing CPU time. It's not an exact copy of the Java code, so not conclusive, but I wonder if it's possible that interaction between Oracle and WebLogic, maybe in getting the bind variable values, is somehow involved here.  

Has anyone seen such behavior? Any suggestions?

Thanks in advance.

Paul Baumgartel
CREDIT SUISSE
Information Technology
DBA & Admin - NY, KIGA 1
One Madison Avenue
New York, NY 10010
USA
Phone 212.538.1143
paul.baumgartel_at_credit-suisse.com
www.credit-suisse.com  

From: oracle-l-bounce_at_freelists.org [mailto:oracle-l-bounce_at_freelists.org] On Behalf Of Jared Still Sent: Monday, March 19, 2007 2:40 PM
To: Kerber, Andrew W.
Cc: robertgfreeman_at_yahoo.com; oracle-l_at_freelists.org Subject: Re: Why an organization would need an enterprise DB team

On 3/19/07, Kerber, Andrew W. <Andrew.Kerber_at_umb.com <mailto:Andrew.Kerber_at_umb.com> > wrote:

I think it would be more accurate to say that developers don't really have time to figure out how the database really works, and dba's don't really have time to figure out how the developers are programming.

Maybe, maybe not. A developer does not need to know in depth how the database works, but do need to know how certain features work, such as undo and redo.

A developer should know why it is bad to do simple DML in a loop, and why issuing COMMIT frequently is a bad idea.

If they are unable/unwilling to learn, then they need to trust the DBA.

That trust works both ways of course.

-- 
Jared Still
Certifiable Oracle DBA and Part Time Perl Evangelist




==============================================================================

Please access the attached hyperlink for an important electronic communications disclaimer: 

http://www.credit-suisse.com/legal/en/disclaimer_email_ib.html


==============================================================================



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