Re: Normalization and Derived Information

From: Kenneth Downs <firstinit.lastname_at_lastnameplusfam.net>
Date: Mon, 04 Oct 2004 16:11:08 -0400
Message-ID: <soasjc.3is.ln_at_mercury.downsfam.net>


Dawn M. Wolthuis wrote:

> "Kenneth Downs" <firstinit.lastname_at_lastnameplusfam.net> wrote in message
> news:2pf1jc.cgg.ln_at_mercury.downsfam.net...

>> Laconic2 wrote:
>>
>> My current thinking, which I am implementing in my current project, is to
>> materialize in the INSERT and UPDATE triggers.   This is not to say there
>> are not other ways, but in this way the definitions are satisfied within
>> the transaction.   There will be howls of protest about performance, but
>> methinks the CPU time on the server to generate the value is less than
>> the wire time to send it over from another tier (I could be wrong).

>
> I could be missing something, but the only way there would be howls of
> protest would be if you decided to materialize the values synchronously,
> right? I can see reasons to update data synchronously, but no reason to
> maintain materialized derived data while a user waits. Triggering an
> asynch
> processes would make more sense, I would think. Again, maybe I'm not
> understanding either your requirements or your design. Cheers! --dawn

The howls of protest are, as you suspect, over synchronous updates, where I measured performance slowdowns of 20% in that environment. This I judge, and emphasis here is on personal judgement, to be far superior to the faster transactions, where the users waited days, weeks, and months for mistakes to be fixed in the extremely complex house-that-jack-built system that attempted to reliably and repeatably derive these values on the fly at extraction time.

>

>> --
>> Kenneth Downs
>> Use first initial plus last name at last name plus literal "fam.net" to
>> email me

-- 
Kenneth Downs
Use first initial plus last name at last name plus literal "fam.net" to
email me
Received on Mon Oct 04 2004 - 22:11:08 CEST

Original text of this message