Re: In an RDBMS, what does "Data" mean?

"Dawn M. Wolthuis" <dwolt_at_tincat-group.com> wrote in message news:canb6q$rab$1_at_news.netins.net...

> Is the star join a relational concept? I heard someone suggest that
> fact-dimension tables with star schema is bad design, but I forget the
> rationale for that and they seem to be very effective.

As near as I can make out, a "star join" is yet another join algorithm, that is added to the ones previously implemented.

Earlier join algorithms include the "loop join" and the "merge join". I could describe these in more detail, but you may already know them. They all acheive the same result: a join. They differ in performance, and different ones are better in different cases. A smart optimizer picks the best algorithm given the available information.

A star schema is not a relational concept as such. A star schema is a projection of the multidimensional model onto databases like Oracle, DB2, etc. that I still refer to as "relational DBMSes", except in this forum, where I will be scolded by the keepers of the faith if I do.

In order to implement a successful star schema, you have to unlearn most of what you learned in normalization catechism. I would have said that would be fun for you, except that you don't unlearn 1NF.

Is it bad design? It depends. For certain types of uses, it is far more useful than a fully normalized relational design. Especially reporting, warehousing, and OLAP. Like almost everything in life, sometimes it's a good idea, sometimes it's a bad idea.

But I wouldn't recommend that you run off and learn star schema immediately, although it might be useful if you could incorporate that into some of the SQL you teach in college. What I would recommend, for what it's worth, is that you learn a little MDDB and OLAP, if you haven't already. Then, I think you would find it quite easy to back your way into star schema.

It's just a blend of MDDB concepts with the relational and SQL concepts you already know. Received on Tue Jun 15 2004 - 21:49:03 CEST