Re: Oracle DB performance tuning training

From: Stefan Koehler <contact_at_soocs.de>
Date: Fri, 21 Aug 2015 23:29:54 +0200 (CEST)
Message-ID: <151481292.98741.1440192594294.JavaMail.open-xchange_at_app02.ox.hosteurope.de>



Hi Raza,
please don't get me wrong. I am not claiming to type long hand commands or to do a lot of manual analysis just to satisify some ego or to prove anything. It does not matter, if you use Enterprise Manager, SQL Developer, TVD$XTAT, Snapper, SQLd360 or whatever.

> at the end of the day, your time actually belongs to the organization paying for your time. They'd rather have you complete the task expeditiously

I totally agree with you and in course of this it is important to focus on the business needs. You need to gather properly scoped diagnostic data and execute the tasks with the greatest net pay off to the business. But this is all about methodology and not primarily about tools. If you just get trained on how to interpret an Enterprise Manager graph in case of CPU, I/O consumption or whatever, you probably gonna fail with this task, especially as some (possible) important components are not accounted as database time at all, but may hurt the end users the most.

> There is a business reason why tools are created. Unless you want to work for yourself..your choice is what serves your employers interests first
> and foremost.

Absolutely and i do not claim not to use tools, but it makes no sense just to train these tools. My field experience (with clients using Enterprise Manager for example) shows that a lot of work time is just spent to reduce CPU time or wait event <X> without even knowing how much time is spent on each component for the whole business process. Does it make sense to reduce just a few secs of CPU time for a SQL (because Enterprise Manager shows high CPU usage for this SQL), if most/all of the end user response time is spent in the application layer or in inter-communication between app and database / app implementation? I hope you would agree with me that it makes no sense, even if some graphs show high CPU bars ;-)

Funnily this is the same discussion and persuading that i have to do with clients who just get trained on some tools, but in the end the properly gathered and scoped diagnostic data are convincing :-)

Finally all i wanted to say is the following:

  1. Train methodology and understanding of performance
  2. Train the prober tools for point 1

Best Regards
Stefan Koehler

Freelance Oracle performance consultant and researcher Homepage: http://www.soocs.de
Twitter: _at_OracleSK

> raza siddiqui <raza.siddiqui_at_oracle.com> hat am 21. August 2015 um 21:16 geschrieben:
>
> Though we don't want to get into a shouting match, but I'd prefer to frame the situation as follows:
>
> if you have access to an electric and manual drill / screwdriver etc, which one are you going to pickup and use ?
>
> Yes - there is a valid argument for understanding what is being done and why, but at the end of the day, your time actually belongs to the
> organization paying for your time. They'd rather have you complete the task expeditiously, than satisfying your ego because you typed the command
> longhand, rather have a tool generate it for you.
>
> Another key example.
>
> You need to recover your crashed database. RMAN will require minimum of 3 commands - and it'll get it done right, whereas figuring-out indvidual
> commands, and correct sequence to issue them...well you can explain the mess to your boss.
>
> There is a business reason why tools are created. Unless you want to work for yourself..your choice is what serves your employers interests first
> and foremost.
>
> My $0.02

--
http://www.freelists.org/webpage/oracle-l
Received on Fri Aug 21 2015 - 23:29:54 CEST

Original text of this message