RE: Heisenberg and measurement intrusion....

Hi Ted,

I see what the person is saying. Yes, there is a measurable effect on performance, due to instrumentation. (This is really due to observer effect, not the Uncertainty Principle.)

However, without instrumentation, you don't really have a system that can be tuned or optimized. You NEED that data to understand the system and its performance profile. Without it, you have no idea which way to turn when performance becomes an issue. Therefore, system instrumentation is an integral part of the system. It's not optional. So regardless of whether measurement intrusion costs you 1% or 10% or whatever, in overall performance, it's a necessary cost.

To borrow a thought from Tom Kyte: Overhead is something you can do without. Instrumentation is NOT something you can do without. Therefore, there is no overhead cost to instrumentation. It's not an overhead, it's a requirement.

Finally, I'd say to the person making the objection: "Ok, what if we eliminate all instrumentation from the system, and discover that the system performs (for example) 10% better. So what? Are we going to remove all instrumentation, in order to save that 10%? Of course not, we NEED instrumentation so that we know what to do when something goes wrong! And, if there is no circumstance where the instrumentation can be removed, then why do we need a baseline without it? What will that accomplish, other than demonstrating the performance improvement of a system configuration that will never see the light of day?"

Just my two bits.....

-Mark

--

Mark J. Bobak
Senior Database Administrator, System & Product Technologies ProQuest
789 E. Eisenhower, Parkway, P.O. Box 1346 Ann Arbor MI 48106-1346
+1.734.997.4059  or +1.800.521.0600 x 4059 mark.bobak_at_il.proquest.com
www.proquest.com
www.csa.com

ProQuest...Start here.

-----Original Message-----

From: oracle-l-bounce_at_freelists.org [mailto:oracle-l-bounce_at_freelists.org] On Behalf Of Ted Coyle Sent: Tuesday, July 31, 2007 3:40 PM
To: 'oracle-l'
Subject: Heisenberg and measurement intrusion....

I'm on an Oracle performance project and a project participant made a statement regarding measurement intrusion.

Is the statement below accurate?
"So the Heisenberg uncertainty principle mandates we run without monitoring as a baseline."

I responded with a wikipedia link which I'll send later, but I'd like to get opinions first. :)

Regards,
Ted

--

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

--

http://www.freelists.org/webpage/oracle-l Received on Tue Jul 31 2007 - 15:54:05 CDT