Pieter,
No, but the basics of Kimball & Ross' books are summarized in some articles
Kimball wrote for Intelligent Enterprise. My summary follows. First, a
data warehouse is a relational database. You can't ignore Boyce-Codd Normal
Form just because you've decided to jump camp and call yourself a data
warehouse architect. The primary mission of a data warehouse is to support
management's understanding of the performance of the organization through
numeric measurement (metrics). Traditional relational designs tend to be
better at capturing data. There are design choices which improve the speed
and reliability of data capture that can in turn inhibit management's
ability to interpret the data being captured.
So, a Systems Analyst tasked with building a relational database tries to
understand the flow of information through a business process and in turn
design a database that efficiently supports that information flow. Some key
goals are to provide the most current picture of the business, to minimize
redundant work by attempting to only capture a piece of data once and
provide it wherever needed, and to put the right data in the right hands at
the right time.
The same Systems Analyst tasked with building a data warehouse is not
releived of his need to understand how information flows through a business.
But instead of ensuring the efficient flow of current data the task is to
ensure management has information it can use to measure the performance of
the organization and make good decisions. As a first goal a data warehouse
must be an accurate historical record of the organization's performance.
One of the early questions that has to be asked is, "What are we measuring
and from what perspective(s) (dimension(s)) to we want to view these
measures?" Based on the answers to these questions the analyst can then
begin to map out attributes of the dimensions that management is interested
in and to identify the numeric facts that will be used to measure the
organization's performance.
So, the relational schema that often evolves has at its center a fact table
containing columns of numeric measures as well as foreign keys that relate
to one or more dimensions by which the numeric measures will be viewed.
"Pieter Linden" <pi********@hot mail.com> wrote in message
news:bf******** *************** ***@posting.goo gle.com...
Alan,
since you're on the subject, know any good books on Data Warehousing
besides Kimball & Ross, "DW Toolkit" ?