Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Mailing Lists -> Oracle-L -> Re: Load and Query question

Re: Load and Query question

From: jaromir nemec <jaromir_at_db-nemec.com>
Date: Sat, 29 Apr 2006 22:29:04 +0200
Message-ID: <00a701c66bcb$8fc344f0$3c02a8c0@JARAWIN>


Hi,

> . would you rather have a relational model which would help really large
> load
> volumes and get the data out there faster or go with a totally
> denormalized structure
> and don't care about when the data is delivered ?

I would definitively postpone the processing of descriptions in the presentation layer (in this case the join argument is IMO not very valid; I assume here the dimension considered here are more or less statically as in the currency example).

In case of slowly changing dimensions (SCD) the demoralisation could be useful but again I would consider primarily the denormalization of other dimension keys related with the SCD rather then the denormalization of descriptions (e.g. storing in the fact table redundantly to customer identification also the customer segment valid to the time of the transaction). The motivation to do so needn't be the pure "join cost to much" argument. The tricky part could be simple to find the right version of the SCD to be joined.

only my .02

Regards

Jaromir

> Hi List,
>
> I have a semantic question in a DWH environment. At the current place I
> work, the tables have been de-normalized to such an extent that they may
> not
> get qualified as 2NF even. Sure there are PK but they are there more to
> provide a security implementation rather than follow the relational
> priciples. Out of 200+ tables there is not one dimension or a fact or even
> one referential key. Each table has about 130+ columns out of which equal
> quantities are codes and descriptions.
>
> The argument in favor the current design is that joins are expensive.
> However from a DBA's perspective would you rather have a relational model
> which would help really large load volumes and get the data out there
> faster
> or go with a totally denormalized structure and don't care about when the
> data is delivered ?
>
> On an average day, the load takes more than 10+ hours and volumes are
> presently at 8mil+ rows everyday. The env is Solaris 9, Oracle 10.2.0.1.0
>
> Thanks
>
> Shreeni
>
> --
> http://www.freelists.org/webpage/oracle-l
>
>
>
>

--
http://www.freelists.org/webpage/oracle-l
Received on Sat Apr 29 2006 - 15:29:04 CDT

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US