This page is a placeholder for the readers of this blog to suggest topics of future blog posts. I can’t guarantee I will be able to write an entire blog post about your topic suggestion, but I can guarantee that I will read it.
Leave your topic suggestion in a comment and stay tuned…
Hi Greg,
The topic i wanted to discuss has been prevalent in almost all the corporations i have worked so far. And i would like to hear your perspective of addressing it.
Generally engineers deliver software builds to QA and they do functional testing and then perform load/stress testing. Problems that are caught mostly in production like oracle choosing bad index vs. good index for the query are not generally caught in QA’s performance cycles. I know oracle has made good attempt to fix this in 11g using adaptive cursor sharing but in 10gR2 (.4), how do we avoid these scenario’s where oracle doesn’t automatically recover from using bad plan to good plan. I have seen most cases were load test environment will not have representative dataset to begin with there by everything working hunky dory on performance labs but discover later to see problems in production.
What is the best way to simulate production dataset to avoid surprises going into production. I believe if load performance test is done on representative dataset, then oracle would generate appropriate stats thereby choosing optimum or sub-optimal execution plans which greatly helps to figure out the problem before going into production environments.
Even if this topic don’t qualify for blog spot, if you could give your ideas of how one should address these issues would be appreciated.
Thanks
Steve
@Steven Andrew
Hello Steve
Oracle 11g Enterprise Edition has an option called Real Application Testing (RAT) that would help you capture production workload and replay it in test environment. It is composed of two components Database Replay and SQL Performance Analyzer. You can learn more about RAT at:
http://www.oracle.com/database/real-application-testing.html
and
here:
http://download.oracle.com/docs/cd/B28359_01/server.111/e12253/toc.htm
There are many requirements to using RAT, you can learn about all those on the above two sites. I hope this is useful.
Thanks
-Prasanna
Hi Greg,
Good blog and i have learned quite a few things over here. I have couple of questions. Is Exadata just Oracle RAC ? I think its not. Is it a shared nothing architecture ? Please guide me to some documentation.
Thanks
Krish
@Krish
Thanks for the comment and glad you have found the blog a good learning tool. That’s my intent ;)
Exadata is modular, scalable, intelligent storage for Oracle databases. I guess one could call it shared nothing but it is no more (or no less) shared nothing than any modular scalable non-Exadata storage. Exadata storage does not require Oracle RAC, it just happens to be paired with RAC in the Oracle Database Machine, but it can be used with any Linux host (RAC or not).
Point your browser to http://oracle.com/exadata and expand the right hand columns (Technical Information, Data Sheets & FAQ, etc.) for numerous papers.
Thanks Greg. Appreciate your prompt response.
Hi Greg,
Can you please explain about dbms_monitor package using CLIENT_IDENTIFIER session variable?
Thanks
Regards,
Senthil
@senthil
Any chance you could provide a bit more context for your question? What would you like explained? Details, details, please…
Greg -I would like to know, how client_identifier session variable is useful for session tracing. In what circumstance we will be using it and why?
I have used dbms_monitor for individual session tracing like the one given below, but I would like to know what are the other options available in the package.
exec dbms_monitor.session_trace_enable(session_id => 1234,serial_num => 56789,waits => true,binds => true);
Hope you understood my question :)-
@senthil
Sorry, I’m failing to see how I can provide any information on this topic that isn’t clearly stated in the documentation.
http://download.oracle.com/docs/cd/E11882_01/appdev.112/e10577/d_monitor.htm#i1002256
Greg, Would it be possible to discuss on the topic of histogram when we set methed_opt=AUTO in 11GR2 ? Especially when the column stores varchar2 or number values which differs only in the last positions and of size more than 12 (bytes/chars) in total ?
What I am trying to get is that , does the dbms_stats care (bout histogram) only about the first 6 characters or digits when it collects histogram and if so , what is the impact on columns such as the one in my example ? How do we handle it for large partitioned tables where the histogram on non indexed columns influences the join methods if they happened to be used with predicates.
Thanks
Sundar R
Greg ,
Im working as tester , I would like to shift to Oracle dba .
My question to you is , as iam nothing now ( had done oracle dba 9i certification )
in this website , As new comer like me . from where to start ? I knew valuable this website is so much of value added information to the career improvement..
Kindly suggest me , from where to start
Thanks & Regards
Srinivas
I’d start with the DBA Essentials (2 Day sections) in the Oracle on-line documentation. After that, you probably need to find a DBA mentor and get your hands dirty with real DBA work.