Archives of the TeradataForum

Message Posted: Tue, 20 May 2003 @ 14:49:06 GMT

	<Prev	Next>		<<First	<Prev	Next>	Last>>

Sanjay got it right - if the questions are known beforehand, Star stuff is ok, I guess, otherwise not.

IMHO normalization is the standard and Star stuff is the aberrant. Because "most of the RDBMS in a warehouse" are Star is due to the inability of other software and hardware to support normalization, and not because Star is the standard. They simply can't handle queries against an RDBMS that today reaches into terabyte size or masses of concurrent users launching less than optimum SQL, so the design has to be compromised.

We have an enterprise warehouse with everything from financials, bus and electronic maintenance, bus routing, student systems, HR, all in 5th normal form with summary tables for quick aggregations. Our goal is to answer any question about any function at any time, and 5th normal has been doing it good, every day, all the time, for the last 15 years. The proof that normalization works is not in a "white paper", but in successful implementations.

We don't use our data warehouse for OLTP. You need two tiers, one tier for OLTP and one for the DW. It's a case of the right tool for the job and not "one tool fits all." The objectives of OLTP and data warehousing are at odds and you must compromise the data warehouse to use if for OLTP and visa versa. If you look in DB2 Magazine, Quarter 3, 2002, Vol 7, No 3, "Tuning Up for OLTP AND Data Warehousing", you will notice the use of the word "compromise" in the article itself explaining how to fine-tune DB2 to do both OLTP and data warehousing from the same box and RDBMS.

There is a 5-page "white paper" of sorts in Teradata Magazine 3Q 2001, "The Benefits of Being Normal", written by Tom Russell and Rob Armstrong which may help you out.

Maybe someone on the List has electronic versions of these two articles. If not send me your snail mail address off line and I'll send you a copies.

John Balchunas
DBA and SysAdmin
Jefferson County Public Schools

	<Prev	Next>		<<First	<Prev	Next>	Last>>