Re: Advise with Oracle data warehouse

Daniel Morgan wrote:

> Insults included ... I completely disagree with you. The point of a data
> warehouse is to support querying and reporting.
>
> If you have no transactions exactly what is it you expect to generate in
> those redo logs that will be archived, that will be worth the price of a
> latte' if the whole thing blows up?
>
> Turn off archiving. Truncate all tables. Reload all data. Why take the
> performance hit?
>
> If transactions are taking place then it isn't a data warehouse.

I agree with Daniel's comments. (I differentiate between a traditional data warehouse and a new-fangled real-time BI data source which loads and/or accesses the TX data in real time.)

In addition, I suggest you consider - build on a seperate cleansing system (which can have archivelog mode turned on, if you really, really want it), use transportable tablespaces (which you back up), set the primary tablespaces read only if possible, use partitioning.

IMHO the primary reason for ArchiveLog is to ensure that irreproducible data can be recovered. The 'irreproducible' part to me means that you do not have to burden external resources (such as customer) to verify the data on recovery. In a proper warehouse I do not envision any data that can not be reproduced from internal sources and recovery time might, or might not, be faster using archives vs raw rebuild.

/Hans Received on Sun Apr 27 2003 - 16:42:42 CDT