Theories of Kimball and Inmon About Data Warehouse Design
Bill Inmon and Ralph Kimball are the two pioneers that stated different philosophies in enterprise-wide information gathering, information management, and analytics for decision support. They have formed a great debate about data warehousing. They approach the problem with different viewpoints, design techniques, and implementation strategies. Inmon suggests creating a single enterprise by data warehouse for producing an overall business intelligence system. Kimbal suggests creating several small data marts for producing department level analysis and reporting.
Architecture
Bill Inmon’s opinions recommend starting with building a centralized enterprise-wide data warehouse by several databases to the analytical needs of departments, which are later known as data marts. Hence this approach has received the top-down title.
The core of Bill Inmon’s model is an enterprise data warehouse. A central repository of data integrated from all the organizational operational systems. This is the definitive consistency of all the representation of business data. This means all the companies decide the naming definitions of which data is true and their conflicting values and all other data cleaning operations are performed before enters the warehouse. The data is stored in the normalized form and the warehouse is not created directly. instead, the data is fed into different data marts which the data is filtered down to the subset of specific needs. As an example, the marketing department will have data that only used by the marketing team. Applications that are to retrieve the data will be connected to these data marts.
Ralph Kimball’s philosophy recommends building several data marts that serve the analytical needs of departments followed by virtually integrating these data marts for consistency through an information bus, this approach is called the bottom-up approach.
The idea is when data is being normalized into an often called star schema. This format makes query writing fast and simple and getting to report as quickly as possible. first, the data from the specific subject area or a process is put into a star schema. this is the data mart of the Kimbal design. multiple data marts can be connected with shared attributes and this forms a dimensional data warehouse.
Both these architectures share the same mutual features nevertheless each of these has a single integrated atomic data repository. When Inmon’s architecture was called an enterprise data warehouse, Kimballs architecture was called a dimensional data warehouse. In Inmon’s method, the data mart is physically separated from the enterprise data warehouse and it has been built for department uses. Nevertheless, it’s not necessary to separate data marts from the dimensional data warehouse in the Kimballs method.
Structuring Data
Besides the difference in approaches, They are proposing different methods for structuring the data. Inmon’s method uses the relational (ER) model which is the 3NF in the enterprise data warehouse whereas Kimbal’s approach uses a multidimensional model which is a star schema and snowflakes to organize data in the dimensional data warehouse. Inmon argues that data is in a relational model it will attain enterprise-wide consistency. Which makes it easier to spawn off the data marts in the dimensional model. Kimbal argues that actual users can understand, analyze, aggregate, and explore data inconsistencies more easily if the data is structured in a dimensional model additionally to enable the information bus. Data can be accessed directly from analytics systems in Kimble’s approach. In Inmon’s approach, analytics systems can only access data from the enterprise data warehouse through data marts.
Which approach suits you?
Each of these methods has its benefits in different scenarios. Inmon’s approach is time-consuming nevertheless easy to maintain. Kimball’s approach takes lesser time but hard to maintain and often redundant and subjects to revisions. While Kimball’s favors initial business support requirements and individual business data integration requirements, Inmon favors strategic business support requirements and enterprise-wide data integration requirements. When you have a generalistic small team, Kimball’s approach suits you, and when the team is bigger and specialized can go to Inmon’s approach. When you urgently need the first data house with low start-up costs, you can go for the Kimbles method. But if you have a high start-up cost and more time to meet your business needs, you can use the Inman method.
