Re: Data Warehouse temptation

Rollback as in updated?

You might end up with a fragmentation problem unless you enter a default value (e.g. -1 for unknown) where you currently have nulls or make sure your PCTFREE is high.

From an architectural stand point, it might better if you received a brand new source file from the third part containing all the commodity codes. This would allow you to simply run an update with a couple of joins and a where clause that would only touch the unknown or null values.

AJ

<tawnyz_at_hotmail.com> wrote in message news:7qne6s$jl0$1_at_nnrp1.deja.com...
> This is a question to people, engaged in such activities as modeling of
> medium-sized (millions of records) data warehouses. I need your opinion
> or, if you encountered this, your experience with the following:
>
> I have a fairly banal data model in Oracle 7.3.4, split into three
> tiers:
> Loading - where incoming data is cleaned and translated
> Storage - where data is split into a relational structure
> Marts - summary objects, mostly denormalized
>
> Data comes in monthly from an increasing number of sources. The
> tuned data flow from Loading to Marts takes about several hours (single
> data chunk is about 500K of records, approx. 250MB)
>
> Now, the enterprise comes with an initiative to populate one of the key
> fields, which is some commodity code, with values, since they are about
> 40% blank.
> Suggested is that we extract unique records from "Storage", where this
> field is blank, submit them to the third-party source, receive them back
> with filled codes (they hold a standard for this) and roll the values
> back in all objects that contain this field in "Storage" and "Marts".
> This process is supposed to be carried on a regular basis (monthly or
> more frequently)
>
> I wonder, what are your thoughts or practices, regarding how such
> processes may affect the system in general.
>
>
>
> Sent via Deja.com http://www.deja.com/
> Share what you know. Learn what you don't.
Received on Tue Sep 07 1999 - 21:50:20 CDT