Mastering Data Warehouse Design: Relational and Dimensional...

Features

• Cover Type: Paperback with 456 pages
• Published by: Wiley August 8, 2003
• Written in: English
• ISBN 10 Number: 0471324213
• ISBN 13 Number: 978-0471324218
• Book Dimensions: 8.9 x 7.3 x 1.1 inches
• Weighs: 1.6 pounds

Book Description

• A cutting-edge response to Ralph Kimball's challenge to the data warehouse community that answers some tough questions about the effectiveness of the relational approach to data warehousing
• Written by one of the best-known exponents of the Bill Inmon approach to data warehousing
• Addresses head-on the tough issues raised by Kimball and explains how to choose the best modeling technique for solving common data warehouse design problems
• Weighs the pros and cons of relational vs. dimensional modeling techniques
• Focuses on tough modeling problems, including creating and maintaining keys and modeling calendars, hierarchies, transactions, and data quality

Back Cover Copy
At last, a balanced approach to data warehousing that leverages the techniques pioneered by Ralph Kimball and Bill Inmon

Since its groundbreaking inception, the approach to understanding data warehousing has been split into two mindsets: Ralph Kimball, who pioneered the use of dimensional modeling techniques for building the data warehouse, and Bill Inmon, who introduced the Corporate Information Factory and leads those who believe in using relational modeling techniques for the data warehouse. Mastering Data Warehouse Design successfully merges Inmons data ware- house design philosophies with Kimballs data mart design philosophies to provide you with a compelling and complete overview of exactly what is involved in designing and building a sustainable and extensible data warehouse.

Most data warehouse managers, designers, and developers are familiar with the open letter written by Ralph Kimball in 2001 to the data warehouse community in which he challenged those in the Inmon camp to answer some tough questions about the effectiveness of the relational approach. Cowritten by one of the best-known experts of the Inmon approach, Claudia Imhoff, this team of authors addresses head-on the challenging questions raised by Kimball in his letter and offers a how-to guide on the appropriate use of both relational and dimensional modeling in a comprehensive business intelligence environment. In addition, youll learn the authors take on issues such as:

• Which approach has been found most successful in data warehouse environments at companies spanning virtually all major industrial sectors
• The pros and cons of relational vs. dimensional modeling techniques so developers can decide on the best approach for their projects
• Why the architecture should include a data warehouse built on relational data modeling concepts
• The construction and utilization of keys, the historical nature of the data warehouse, hierarchies, and transactional data
• Technical issues needed to ensure that the data warehouse design meets appropriate performance expectations
• Relational modeling techniques for ensuring optimum data warehouse performance and handling changes to data over time

Reader Reviews
If you want to build a Corporate Information Factory (CIF) I suppose this book is better than many of the previous attempts at teaching how to accomplish that goal. However, like many of the previous Inmon/Imhoff books, there is too much theory (unfocused at that) and not nearly enough practical/tactical content. If you are on the CIF bandwagon however, you will find the book very helpful as compared to most of the previous books on the topic. But that begs the question. Many a CIF or enterprise-wide project has been launched... yet most are cancelled long before reaching the finish line. This is reality. In the REAL world we have REAL deadlines and REAL budgets imposed by REAL business executives who have REAL problems to solve and it involves... oh by the way... REAL MONEY! We have to deliver NOW! Well, ok, maybe not quite that fast, but you get the idea. The hard part is getting the data! Or is it? Using simple tools and a powerfully designed, highly detailed dimensional database, we have, for example, clients pulling their own data sets ready for import into statistical and mining packages. They think they have died and gone to heaven! Foist a third normal form (3NF) design on them and their eyes roll... "Now, which of the available join paths is the right one for this business question?" and "Why is it taking so long for the query?" and "Will you pull the data for me?" Now we hear... "Instead of spending 80% or 90% of my time getting the data prepared, I spend 5% or 10% of my time doing that... so I have that much more time to actually think about the business." We have seen clients' ability to understand and drive their business expand beyond their own wildest imagination in very short order. It shows on their bottom line and they are very happy with that! The whole point of BI - beyond all the data capture and cleaning and integrating and turning "data into knowledge", and making it easy for the user without dumbing it down, and all that stuff - the point of BI can be distilled down to one word: "Publish!" Booksellers don't hand you a photocopy of a handwritten manuscript. They do a lot of work with the "raw data" - typesetting and page numbers and table of contents and indexing and so on - and turn it into something accessible and useable... something we call a book. That's the point of BI. This book doesn't get it. Too many CIF or "enterprise" projects have imploded under their own weight to slavishly duplicate the same mistakes. Too many dimensional systems have succeeded with huge return on investment to relegate the ideas to a dark corner. If we stop the religious discussions (Mac vs. Windows, or the "Inmonites vs the Kimballites") and get to see how truly successful Business Intelligence (BI) systems work, we find the emphasis must be on using proper theory (not arguing it) and applying techniques that work NOW. More often than not, can you say "Dimensional!" Yes, CIF and all that has its place... but not nearly to the degree that this book would have you believe. The most successful clients have been the ones who bypassed all the "modeling wars" and used the data bus architecture of conformed dimensions. They didn't pick and chose a modeling idea or two; they actually studied Kimball and did it the right way. Dr. Codd, while addressing this question one day, asked me this question: "Would you run an OLTP system against a dimensional model?" My obvious answer was: "Of course not." "Why then," he asked, "do so many people try to do the opposite?" The biggest "problem" with the dimensional approach is that people who do not truly understand it try to pick and chose techniques from it and graft those into their current ways... and fail... and bash it. Or, they don't understand it at all. Uh, sorry, it isn't the technique that is the problem. The book purports to "answer" a message reply that Ralph Kimball posted on a discussion board some time ago. It does not. One can be certain that Ralph Kimball did not give permission to use his name on or in the book, as is done. Instead, the book does a very poor job of showing how to design and use dimensionally designed databases as a part of a larger architecture, illustrates a complete lack of understanding of the underlying principles, and then criticizes and limits the technique and its application. This does a terrible disservice to the reader... especially a reader who is trying to decide how to meet a real business need and is new to BI. I dislike speaking impolitely like this, but the truth is more important in this context. Also, on the back cover, they state that Ralph Kimball's "letter" was a challenge. It was not. It was merely a listing of many of the crucial issues in a useful BI environment addressed to an individual who had asked legitimate questions about BI. As for addressing these issues "head-on", the book does not do this at all. Does this matter? Of course it does. Real people buy this book and are led down a path that rarely leads to success. I realize that much of this review is not directly about specific details of the book. The details in the book are inconsistent, often unfocused, and sometimes downright misleading. The larger issue, and thus the focus of this review, is that the entire book is based on a premise that the CIF is "The Way" and that dependent dimensional data marts are grudgingly "ok". This is not the reality that many of us see in the business and education worlds. Comment (1) | | (Report this)