Inmon architecture and Kimball architecture are two of the most
popular data warehouse architecture models.
The architecture of the Inmon Approach
In the perspective of Bill Inmon, the first step should be to create a
consolidated enterprise-wide data warehouse. This can be made by
connecting numerous databases to the analytical needs of departments,
which are later referred to as data marts. As a result, this method is known
as the top-down approach.
Data Structure :A normalized central data warehouse to reduce
redundancy and improve integration
Inmon’s Approach to Data Warehouse Designing mainly consists of the
following three steps:
Step 1: Specifying the Primary Entities of the Organisation.
Step 2: Develop a detailed Logical Model of each entity.
Step 3: Development of the Final Physical Model.
Advantages of the Inmon Approach
Because it is the only source for data marts and all data in the data
warehouse is integrated, the data warehouse genuinely acts as the
enterprise's single source of truth.
Due to the limited redundancy, data update abnormalities are avoided.
This simplifies the ETL(Extraction, transformation, and loading)
procedure and reduces the risk of failure.
Because the logical model represents the distinct business entities, We
may easily understand the business processes.
Very versatile — As business requirements change or source data
changes, updating the data warehouse is simple because everything is
in one location.
Can handle a variety of reporting requirements within the organisation.
Disadvantages of the Inmon Approach
As more tables and joins are added, the model and
implementation can grow increasingly complicated.
We'll need people knowledgeable in data modelling and the
business in general. These resources might be difficult to come
by and can be rather costly.
Management should be aware that the initial setup and
delivery will take longer.
More ETL work is required as the data marts are developed
from the data warehouse.
A vast team of professionals is required to manage the data
environment efficiently.
The architecture of the Kimball Approach
the bottom-up method calls for the creation of many data marts to
satisfy the analytical needs of departments, followed by virtual
integration of these data marts for consistency through an
information bus.
Data Structure : Denormalized Data Marts for faster queries and
improved performance
Kimball’s Approach to Data Warehouse Designing mainly consists of
the following three steps:
Step 1: Identifying the critical business processes and queries.
Step 2: A dimensional model is created by loading data into it.
Step 3: Development of the final Physical Model.
Advantages of the Kimball Approach
The first part of the data warehousing project will be provided
fast because it is simple to set up and build.
The star schema is simple to understand for business users
and implement for reporting. Most BI (business intelligence)
solutions are compatible with the star schema.
The data warehousing environment has a small footprint; it
takes up less space in the database and makes system
maintenance more manageable.
A small group of developers and architects is all that is
required to keep the data warehouse running smoothly.
Because data marts are designed for departmental or business
process-level reporting, it works very well for department-level
metrics.
Drill-across, in which a BI tool generates a report by traversing
various star schemas, may be achieved successfully with
conformed dimensions.
Disadvantages of the Kimball Approach
Because data is not fully integrated before addressing
reporting purposes, the core of the 'one source of truth’ is lost.
Over time, redundant data might produce data update
abnormalities.
Adding columns to the fact table can slow things down. This is
because tables are built to be very deep. The fact table will get
significantly more prominent and perform poorly if more
columns are added. As a result, changing the dimensional
model as the business requirements vary is difficult.
Because the model targets business processes rather than the
enterprise, it cannot meet all enterprise reporting demands.
Integrating legacy data into a data warehouse is a time-
consuming procedure.
Kimball Vs Inmon
Each of these strategies offers its own set of advantages in different
contexts. Although Inmon's method is time-consuming, it is simple
to maintain. Kimball's method saves time but is difficult to maintain,
typically repetitive, and subject to modifications.
Inmon prefers strategic and enterprise-wide data integration
requirements, whereas Kimball prefers initial business support and
individual company data integration requirements.
Kimball's approach is best when you have a small generalist team,
while Inmon's is best when you have a more extensive and more
specialised group.
The Kimball technique can be used when you quickly require a first
data house and a reasonable cost. On the other hand, the Inmon
technique can be used if you have a high startup cost and require a
longer time to meet your business needs.
Incremental Data Load vs Full Data Load
