Re: Designing database tables for performance?

From: Bob Badour <bbadour_at_pei.sympatico.ca>
Date: Fri, 23 Feb 2007 21:44:35 GMT
Message-ID: <7fJDh.384$PV3.6587@ursa-nb00s0.nbnet.nb.ca>

jgar the jorrible wrote:

> On Feb 23, 5:24 am, "Cimode" <cim..._at_hotmail.com> wrote:
>

>>On Feb 23, 2:06 pm, Frank Hamersley <terabitemigh..._at_bigpond.com>
>>wrote:
>>
>>>Cimode wrote:
>>
>>>[..]
>>
>>>>Yep.  Last time I discussed database issues with an ORACLE guru, he
>>>>was trying to convince me that RAM was logical as opposed to Hard
>>>>drive which was physical.  To the ORACLE gurus, as soon as it is in
>>>>memory, it becomes totally logical.  A total absurdity of course...
>>
>>>He was prolly talking about the types of IO's for a query that the
>>>optimiser predicts and execution engine encounters.  Sybase uses the
>>>same terminology and weights them differently when costing out
>>>(possible) plans.
>>
>>>Cheers Frank.
>>
>>In what RAM would be less physical than HD ? For any reason, an
>>absurdity is an absurdity.

>
>
> Not an absurdity, you just aren't paying attention to how the I/O is
> counted. From Oracle's point of view, if the desired data exists in
> Oracle's buffers, that is a logical I/O.

To exactly what sort of logic does it apply? Predicate logic? First order? Second order?

If Oracle has to ask the OS
> to give it stuff to put in the buffers (or Oracle knows that it has to
> get it off a disk using its own raw I/O), that is counted as a
> physical I/O.

Why not count one as a cache hit and one as a miss? How does it help to create a new and obscure term by borrowing a completely unrelated word with an existing well-defined meaning?

[snip] Received on Fri Feb 23 2007 - 15:44:35 CST