Re: The problem with denormalization.

"Laconic2" <laconic2_at_comcast.net> wrote in message news:Ncednc27yMJ7qMXcRVn-hA_at_comcast.com...
>
> "Tony Andrews" <andrewst_at_onetel.com> wrote in message
> news:1096291773.101549.254100_at_h37g2000oda.googlegroups.com...
<snip>
> Take a look at star. It isn't the silver bullet. But it's good.

It is often used for OLAP to rehost data from a mostly-normalized structure to a star schema, but you can get the benefits of a star schema in an OLTP system as well with virtual stars. With the average RDBMS, it is often too much overhead on the sytsem to have your aggregate data computed on the fly or even to use the size of SQL Views you would need with the number of virtual fields (stored procedures or whatever your database permits in this regard) you would want for a star. But there are some systems that perform well for both OLTP and for OLAP by using virtual stars (the ones I know are against non-1NF data structures, as some of you might have guessed).

[An aside: The U2 products from IBM are and example of this approach and the reason this data model has lasted since the mid-60's, with U2 business growing.]

--dawn Received on Tue Sep 28 2004 - 02:11:16 CEST