|
By Len Silverston |
|||
|
If you'd like to get a jump start on effective reliable Relational Data Models for any application, you need to read Len Silverston's, The Data Model Resource Book, Revised Edition, Volume 1, A Library of Universal Data Models for All Enterprises. I can't think of a single more essential book for application development managers, CIO's, DBA's, and even infrastructure managers. If I had to keep only one book in my IT library, this would be it. This book takes the error factor out of data modeling, puts an expert modeler on your team, and lets you begin with a solid, proven foundation for your data. Data Modeling has historically been portrayed as an art more than a science, with every developer and user having an opinion on how the data should be structured. Most users don't understand relational database theory, and see everything from a "flat-file" or spreadsheet perspective. Even enterprise PIM's and contact managers rarely have good relational data models implemented. Is Your Model Busted? Can your application track information for someone who works for several companies in different capacities without having to enter the same individual more than once? Can you have several contact numbers for a company, which can be organized by the role or purpose for the number? Can you have several different addresses for a company, without having to make duplicate records? Can you hold more than a fixed number of telephone numbers for a contact? If you can't answer yes to all these questions, then your application may be built on a flawed model. Bad Models Can Be Expensive Application-dependent data models can make changes and enhancements very expensive and difficult. Especially as needs shift over time, eventually the business rules, logic, and data requirements combine to make the application so complex that further development is no longer cost effective. A good model should be independent of the application that it is built to serve. The danger is that this seems opposite to what it should be. Here are Len’s own words on the subject of custom versus universal models: “On the surface, it may seem that a more specific data model may serve the needs of the organization better because many of the data constructs and data names may be more specific to that particular enterprise. As stated in Chapter 1 of Volume 2, there are many reasons that an enterprise might find the broader data model more useful. To summarize these reasons, this model provides a framework that can accommodate a broader vision for that enterprise, more integration across various parts of the enterprise, more stability should changes occur in the enterprise, structures that are easier to modify for that industry because one is working from a broader perspective, and finally, these broad industry models often include specific structures, when necessary, to handle special requirements for very specific types of companies.” Not Invented Here? Some people like to start every project from a blank sheet of paper. Others like to build on what has been done in the past. Data modeling is more like mathematics and science than art. There are only a fixed number of truly relational data models for the needs of a given set of data. Nearly always they appear to be counter-intuitive to one untrained in relational theory. The art comes in explaining how it works to the users (and even developers). This one issue can drive users (and ultimately, the development team) crazy, especially if they are allowed to be part of the data design process (beyond defining data requirements that is). Too many times I have seen development projects fail or incur serious cost overruns because the data model was imposed by users from a flat-file mindset.. |
Do You Know What You Don't Know? Even professionals in IT may suffer from this affliction, as data modeling and design is a specialty area. Other fields, like Medicine, have specialists recognized as the only ones trusted and competent to perform certain procedures. An internist would hardly think to perform heart surgery, but users, programmers, and project managers dabble in relational database design. Our field is too young, and relational databases have only been available for about 15 years. Many working databases are converted hybrids from legacy applications, and can’t pass muster of a fully normalized relational model. Ted Kowalski, Author of ‘Opening Doors: A Facilitator’s Handbook’ and formerly head of data administration for A I M Management Group, Inc. says, "Within a short period of 10 business days, Mr. Len Silverston delivered a corporate data model repository using his extensive library of universal data models and data warehouse designs that was tailored to A I M Management Group, Inc., a 30 billion dollar organization. Mr. Silverston’s expertise proved invaluable to the success of our projects." The book is written in an easily understandable style, not like a pithy research paper. It is laid out in separate knowledge areas, and each area discusses information requirements and issues specific to that application area. The chapter listing is: Introduction, People and Organizations (Parties), Products, Ordering Products, Shipments, Work Effort, Invoicing, Accounting and Budgeting, Human Resources, Creating the Data Warehouse Data Model from the Enterprise Data Model, A Sample Data Warehouse Data Model, Star Schema Designs for Sales Analysis, Star Schema Designs for Human Resources, Additional Star Schema Designs, Implementing the Universal Data Models. The parties model of Chapter 2 is brilliant in its simplicity, and bears going over several times to understand all of the nuances of its elegant flexibility. The Appendix contains, among other things, a complete list of the fields and attributes of every table. It is tough to come to the realization that universal data models might be the best solution for any complex application. I have personally spent many hours going through Len’s book and have finally come around to the opinion that nearly every application should be built from universal data models as a starting point. Len draws several examples for each area of knowledge, and shows alternate ways of modeling the data. These are very helpful in understanding the optimal model for a particular problem. Working from these models puts a relational theory expert on your development team. Who wouldn’t be able to benefit from having a brilliant thinker like Len on their side? All you have to do to get Len’s expertise is to buy this book. With apologies to American Express, I won’t leave for the development meeting without it. Silverston’s follow-up companion book, The Data Model Resource Book, Volume 2, A Library of Universal Data Models for Industry Types will be reviewed in a future article. Both books come with a CD containing SQL scripts and supplementary data which will also be reviewed in an upcoming article. Hyperlink to more info on this book: http://www.wiley.com/WileyCDA/WileyTitle/productCd-0471380237.html LEN SILVERSTON (lsilverston@univdata.com) is founder and owner of Universal Data Models, LLC (universaldatamodels.com), a Colorado-based firm providing consulting and training for helping enterprises customize and implement "universal data models" and develop holistic, integrated systems. Mr. Silverston has over 20 years’ experience in delivering data integration, database and data warehouse solutions to organizations. Theo Gantos (  )
is President of TEKA, a technology consulting firm. Theo is a member of
the AITP Chicago board, AITP webmaster and editor of the AITP Chicago
Newsletter. |
||
Questions about this site? Send mail to: 
