Solving Operational Business Intelligence with InfoSphere Warehouse Advanced Edition

86 Solving Operational Business Intelligence with InfoSphere Warehouse Advanced Edition

4.2 The logical modeling process

As discussed in the previous section, to provide the most efficient database

design with which to answer your business questions, you need to implement a

star schema rather than the 3NF schema currently employed within the OLTP

database. However, creating a model in the form of a star schema that

represents a business operation is not a simple process. To aid in this

development process, IBM has provided tools such as InfoSphere Data Architect

which can create, as a starting point, a basic logical data model from a converted

reverse engineered physical data model of your OLTP database. This base

logical data model can then be developed, still utilizing InfoSphere Data

Architect, into a form representing a star schema.

This process of developing a model representing a star schema is known as

dimensional modeling, and is most successful when implemented as a

four-stage process as described by Dr. Ralph Kimball

1

. These stages are:

򐂰 Identify the business process.

򐂰 Define the granularity.

򐂰 Define the dimensions.

򐂰 Identify the facts.

4.2.1 Identify the business process

The 3NF model implemented as the production database within a company

normally covers a wide range business functions, both for efficiency and

performance. A useful example is a production database that catered for the

needs of sales, order management, inventory, and procurement.

However, to best answer your business questions, you have to separate these

business processes into multiple dimensional models. Thus, the first stage is to

identify these business processes and their relative components within the

database.

1

Ralph Kimball et al.,The Data Warehouse Lifecycle Toolkit, John Wiley & Sons, 2008.

Alternative modeling technique: Within this book, we focus on dimensional

modeling as a technique for modeling a data warehouse.

There is, however, an alternative in the form of entity relationship (E/R) or

entity attribute relationship (EAR) modeling, which is related and aligned to

3NF and OLTP modeling.