Professional Documents
Culture Documents
Data Warehousing: Kimball Model Vs Inmon Model
Data Warehousing: Kimball Model Vs Inmon Model
Allan Peda
Week 3, Kimball vs Inmon Data Warehouse Designs
Page 1 of 7
Allan Peda
Assignment
Describe a Business Problem or set of business problems that the your company
is trying to solve through a data warehouse or data mart. Compare and contrast
the Kimball and Inmon approaches to data warehouse. Pick the one that best
solves your business problem and why.
Problem Statement
The chosen business problem for this data warehouse is the determination of
enrollment trends and subject interest within a university. This information derived
from these data is of tremendous value as it will be used to develop long term
growth plans for the university including campus expansion, departmental
funding levels and potential areas for enhanced research and development
funding. The initial focal point of this from a business unit perspective will be
student registration data and library collection access records. This will define
initial data feeds for ETL processing.
Problem Resolution
Page 2 of 7
Allan Peda
Relational Multidimensional
High Performance
Considered Quicker To Implement
Gradual Development Supported
Robust
Flexible
Scalable Performance
As noted earlier, the Kimball design is developed along the requirements of the
end user. As such the data must be denormalized and refactored using a well
defined set of end user requirements. The underlying logical architecture is not
1 This step is sometimes altered slightly with data transformation done primarily within the
database. This is referred to as Extract, Load and Transform (ELT).
http://www.b-eye-network.com/blogs/linstedt/archives/2005/05/elt_and_etl_can.php
Page 3 of 7
Allan Peda
Kimball data warehouses are comprised of multiple subject oriented data marts
which each center on a few core fact tables. These tables contain data which are
easily aggregated. Surrounding these fact tables are multiple dimension tables,
which contain the associated characteristics of these facts (time being an almost
universal attribute for example). These dimension tables are considered to be the
entry points used to access and aggregate the associated facts. They are used
for filtering, grouping and labeling.
In order for queries to determine what correlations between facts and attributes, it
is desirable for dimension and fact tables to be shared among several subject
areas. The careful design and selection of this set of conforming fact and
dimension tables is referred to by Kimball as the BUS architecture.
Page 4 of 7
Allan Peda
Illustration 1: Data flow in the Kimball Data Warehouse Architecture (from The
Data Warehouse Toolkit, Kimball 2nd edition,2003)
Page 5 of 7
Allan Peda
Page 6 of 7
Allan Peda
References
Page 7 of 7