Re: Oracle Myths

I don't think this is one where you can produce a definitive 'hard' proof; you can only posit a logical argument. Viz (and I'm sure you already say this to your students, Daniel):

1. I can always invent a system that requires a given number of real physical I/O requests per second, irrespective of the size of any available cache.
2. Real hardware devices can only support around 100 random I/O requests per second.
3. If I want to 10,000 I/Os per second then I need a minimum of 100 devices.

Given the free choice between 50 x 20GB drives and 100 x 9 GB drives, I'd need to have the larger number of smaller drives. On the other hand, given the choice between 100 x 20 GB drives and 100 x 9 GB drives, I'd take the larger and be very careful about how the spare 11GB per drive was used.

QED In real life, of course, the calculations are rarely made (until it's too late), and the scale of operation is rarely so extreme that throwing loads of cash at a cache fails to make the difference.

--
Jonathan Lewis
http://www.jlcomp.demon.co.uk

Author of:
Practical Oracle 8i: Building Efficient Databases

Next Seminar - Australia - July/August
http://www.jlcomp.demon.co.uk/seminar.html

Host to The Co-Operative Oracle Users' FAQ
http://www.jlcomp.demon.co.uk/faq/ind_faq.html



Daniel Morgan wrote in message <3CE987D0.A4D9D0E0_at_exesolutions.com>...


>

>On the face of it I can't disagree. But then I couldn't disagree initially

>with a number of things pointed out as myths.

>

>Has anyone actually put this to the test and gathered the numbers ... or

>is this unsubstantiated theory?

>

>The quest continues.

>

>Daniel Morgan

>