Customer Reviews

The Data Model Resource Book, Vol. 3: Universal Patterns for Data Modeling

5 star:		(12)
4 star:		(1)
3 star:		(0)
2 star:		(0)
1 star:		(0)

 

 

As an analyst for a large manufacturing company's ERP implementation, I was responsible for a very complex and critical area called Classifications. Classifications was the place where all products, vendors, or customers were grouped into buckets based on similar behavior. For example, if this company manufactured vehicles, there could be classifications for hybrids, sports cars, SUVs, minivans, etc. To better understand classifications, I dived into screens, help files, and actual database tables and after several weeks, completed a classifications data model. The model I produced was very similar to the data model that appears on page 224 of "The Data Model Resource Book Volume 3: Universal Patterns for Data Modeling" by Len Silverston and Paul Agnew.

This book contains a collection of patterns, which are general building blocks that could be used as the basis for just about any type of data modeling within any industry. Classifications is one example, and there are a collection of others such as roles, statuses, and contact mechanisms. Whereas Volumes 1 and 2 in The Data Model Resource Book series contained models for common business processes or industries, this volume contains patterns that cross through all processes and industries. Consistent with the series however, the purpose of this text is to save the modeler time so instead of starting from scratch, the modeler can start from a reliable and proven foundation. Realizing these patterns exist and making them work for your particular modeling assignment can result in a higher quality data model and a greater level of consistency within your organization.

A majority of the book is dedicated to chapters which describe how to model a pattern at different levels of generalization. Level 1 is the most concrete and this is where terms and rules a business analyst are familiar with are shown, such as email address and telephone number. Level 2 through 4 go through increasingly more generized levels with Level 4 being the most generalized. The Classifications example I encountered in the ERP package for example was a Level 3 model, very generalized so that it can be leveraged by any industry. The book makes an important point that there are situations where one level is more appropriate than another, and sometimes the modeler must trade the familiarity and business rule enforcement of a Level 1 with the flexibility available in a Level 2, 3, or 4. For example, a phone number and email address from a Level 1 model would be generalized into contact mechanism data in a Level 2 model. This extra flexibility allows for accommodating other ways of contacting someone that may not have been specified (for example, via a person's "voice over IP" or Skype number). The book also makes the point that sometimes on a single model you can combine different levels for the same requirement (i.e. a hybrid approach).

Chapter 1 introduces the concept of a universal pattern as well as the terms and symbols used throughout the book. The goals for the book are also clearly articulated, in addition to the intended audience and a summary of each chapter. There is a wonderful furniture analogy used to distinguish a universal data model from a universal pattern. Universal data models (the subject of the first two volumes of The Data Model Resource Book), are similar to already constructed standard tables and chairs. The consumer can obtain this furniture instead of build the tables and chairs from scratch. Similarly, the modeler can reuse an inventory or claims universal data model instead of building it from scratch. Universal patterns are similar to the dovetail joints of the furniture, common pieces that exist in already built tables and chairs as well as custom furniture. Universal patterns are the building blocks such as the roles and statuses behind any modeling project.

Chapters 2 through 8 each focus on a particular pattern. Chapters 2 and 3 focus on parties and roles; Chapter 2 on declaration roles and Chapter 3 on contextual roles. A party is a person or organization of importance to the business, and declaration roles are those roles that are independent of any business event while contextual roles are dependent on a particular business event. For example Bob the person can have a declarative role of `Doctor', yet when an insurance claim is filed, they can also have the contextual role of `Primary Care Physician'. Chapter 4 focuses on similar structures for relating data including hierarchies, aggregations, and peer-to-peer relationships. Chapter 5 focuses on taxonomies and classifications, and Chapter 6 on patterns for states that business concepts go through. Chapter 7 contains patterns for getting in touch with parties, such as those patterns for modeling telephone number and email address. Chapter 8 focuses on how to model business rules including the rule itself, the factors involved in the rule, and the outcomes of the rule.

I was impressed with the consistency and comprehensiveness of each of these chapters. These chapters follow a similar format of demonstrating each of the four levels of detail. Each chapter begins with an explanation of the pattern and a discussion of its importance. Then for each of the four levels, there is a section on the reason for the level, how the pattern works (with lots of examples), when the pattern should be used, and the weaknesses of the pattern. I found the charts and tables to be extremely useful in the text, especially the Summary of Patterns table at the end of each chapter.

Chapter 9 focuses on how to apply these patterns in many types of efforts including both relational and dimensional modeling efforts and both application and enterprise areas of scope. As with the other chapters, there is a great summary at the end on the strengths and weaknesses of patterns with each type of effort. Chapter 10 adds the human dynamics side to incorporating patterns, as success or failure is heavily connected with people's perception or trust. Four principles are discussed, that will help acceptance and usage of the patterns: Understand motivations and work toward meeting them, Develop a clear, common, compelling vision, Develop trust, and Manage conflict effectively.

To summarize, under every data model is a set of common building blocks, clearly explained in "Universal Patterns for Data Modeling". I would recommend this book for every analyst, modeler, or architect who is striving for a level of information consistency within their organization. Whether you are just starting your modeling adventure or have been in the modeling for decades, you will find these patterns invaluable tools for every modeling effort.

Unlike the world of Object Oriented programming, there are very few pattern books devoted to data modeling. This is unfortunate because most business applications suffer from over-specialization & inflexibility in the data model, where change is very expensive. The few data model patterns books that do exist do not explore the depths of subject to the extent that this book does.

Based on their many years of experience, Silverston & Agnew chose to focus on the most common areas of interest (Parties, Roles, Relationships, Statuses, Classifications, Contact Mechanisms, & Rules) and offer a set of data model pattens for each. Each area has a set of "pattern levels" ranging from specific to general, listing benefits, drawbacks, & usage guidelines for each. The modeler can choose which pattern level best suites the enterprise based on perceived needs.

This book can be thought of as the "gang of four" book for data modeling. I would recommend this book for anyone engaging in any form of data modeling. You may not use the exact patterns, but it will at least serve as a catalyst for future thought.

As a developer, the databases I have to work with are sometimes thrust upon me. I've often felt that the inflexibility of the design limited the creativity of the developers forced to use them. This book gave me a good understanding of universal patterns that I could take advantage of when designing sub modules where I could control the DB design. It's a clear, straightforward resource that makes you think about how to improve your work immediately. 2 Chapters of particular interest to me were chapters 8 (Business Rules) and 10 (Socializing the patterns). The Socializing chapter was interesting, because you know there'll be someone who fights accepting a new idea on your team. This is not a chapter you usually see in a technical manual.

Don't assume that you can fly through this book (unless your very familiar with the concepts). It's worth the extra time to fully understand the notations laid out in the beginning and to get a good understanding of each chapter before moving on to the next. I recommend this book to anyone who wants to get a better understanding of data modeling, particularly developers. If you don't understand it, you can't use it effectively.

"Universal Patterns for Data Modeling" is the third volume in The Model Resource Book series.

In "Universal Patterns", authors and data modeling experts Len Silverston and Paul Agnew examine the recurring patterns that are essential for architects to understand in building and maintaining enterprise data models. The book covers key patterns, including those for party, role, hierarchy, classification, status, communications, and business rules.

Modeling methodology guides sometimes are overly prescriptive. However, a strength of the Silverston/Agnew book is that it recognizes that different organizations have different requirements. Each pattern is examined across four "levels." A level represents the degree of generality or specificity with which the particular pattern is implemented. Thus, there is no underlying "one-size-fits-all" assumption, but rather a framework for thinking about the relative tradeoffs in choosing the level of generality to match given business requirements.

Anyone with a serious interest in data modeling should have all three volumes on their bookshelf. However, Volumme 3, "Universal Patterns" is definitely the volume to start with if you have not already bought Volumes 1 and 2. I've joked with Len and Paul that Volume 3, is the "prequel." For those who are new to this series, start with Volume 3, then go on to Vol. 1 and 2.

Len's books are a must-have for all data architects looking for quality improvements in their data models. The first two volumes offer solid foundational data constructs ready to jump start any data architecture initiative. I've used these models on countless occasions, and even after many years of relying on them in my daily work I'm still amazed not only by how often the models already have built-in solutions for data modeling situations I encounter day after day, but also at how resilient these data construct are at satisfying changes in business requirements without the need for changes to the data structures.

This third volume is a significant addition to the series. While applying the models from the previous volumes I always found myself deducting the data modeling patterns out of the already implemented data model. The extrapolated pattern, and not the actual model, is very often what I use to customize the models for a specific requirement.

With the third volume, deducting the data pattern out of the model is no longer necessary. It explains the main data patterns thoroughly, while providing four levels of abstraction going from very specific and inflexible, good for discussing the pattern with non-technical resources, to very abstract and flexible, good for actual physical implementation. I would say that in many cases the patterns in the third volume are actually more advanced than the patterns used in the models from the previous two volumes, so I agree with one of the other reviewers when he says it would be a good idea to start with volume 3, to ensure you are working with the latest version of the patterns.

In any case, do yourself a favor and get all three volumes. You will be greatly helping your organization in finally reaching a stable data architecture. Who knows, maybe they'll even thank you for it!

As a manager responsible for several data warehouses I found this book clearly describes fundamental data modeling patterns in a way that makes them usable in real world applications. I have found that UDM speaks to the core data warehouse design allowing to you reuse patterns across your enterprise. This book walks you through several types of patterns clearly explaining why you need it, when you should use it, and what are the pattern's weaknesses.

I highly recommend this book to any data warehouse professional who is serious about implementing good data models quickly and effectively.

Pity the poor data modeler. He or she is deluged from all sides. One the one hand, she must develop a product of high technical excellence so databases and computer applications work efficiently. One the other hand he must capture the nuance in everyday business language in such a way that regular human beings can use computers effectively. Both tasks are essential. All this in the face of the cold, brutal reality best expressed (I believe) by the statistician George Box, that "no model is correct. But some are useful."

Hence the reason that Volume 3 of Len Silverston's series on is so welcome. The Data Model Resource Book, Volume 3 (co-authored by Paul Agnew) proposes four "levels" of reusable "patterns" for the most common data model constructs such as "roles" and "classifications" and other common data modeling structures. To oversimplify, Level 1 models are for the business community and aim to make it easier for them to contribute to the modeling effort; Level 4 models are the sophisticated technicians and aim to make it easier for them to build long-lived, easily-maintained data structures. So Volume 3 can help on both the essential tasks noted above.

For some time, many of us have felt that data modeling is too much "art" (maybe even "black art") and not enough process. In the last two chapters, Silverston and Agnew begin to address the vexing questions of "when should a data modeler apply what model." No silver bullet, but solid steps in the right direction.

Thus Volume 3 is a practical and valuable resource for anyone involved in data modeling.

I have been a project manager (PM) in DW/BI for approx. 9 years and I generally find my data architects to be proficient, but not in command of data modeling concepts nearly to the depth of this book. Even as a PM, I found the book to be well written, easy to digest and incredibly thorough. I especially like the author's pros and cons on each pattern, which is very helpful for directing the modeler on how to apply the concepts in their specific situation.

Recently I worked with our data architect to design a hierarchical dimension between 3 roles in our business unit. I leveraged the book to develop a starting point, which was essentially used "as-is". I haven't seen the book since, as my architect has "borrowed" and has used it ever since.

I think this book is an excellent reference for all IT Data Professionals, including leads, which are looking to improve the quality of modeling and ultimately the product delivered to their clients.

This volume pulls together the first two and tells how to use and implement universal data models. Very applicable for what I do.

What a great reference book! If you do anything with database design or architecture, you need to buy each book in this series. Period.