Inmon V/s
Kimball, the two DW approaches
In this blog is would try to explain the two most prevalent approaches
to DWH Architecture. Both the approached are used widely with their respective
constrains in real world. So which is better, this debate forever but here I would
try to provide some holistic view as what each has to offer and where they can
be used best.
Inmon’s approach:
It’s a top down approach where first, the normal data model is created then the
data marts for specific needs for various departments are created. The normal
data created is the Enterprise wide data warehouse. The EDW practically
contains information about all the company’s departments and serves as on
single repository.
Kimball’s
Approach: This is a bottom up approach where first the data marts are
created which them is combined to create Data Warehouse. In This approach the data
need not be in 3NF form. The data mart is generally connected in a form of star
schema, where only the related departments’ information is stored. This approach
keeps the most important business consideration in mind, hence creates data
marts first
Comparison
between Inmon’s and kimball’s architecture
1 1. Cost: Since Inmon’s methodology requires
to first create an enterprise wide data warehouse in 3NF form so it involves
more initial cost but subsequent cost will belower. InKimball’s approach the
initial cost of development is lower but each new phase adds significant cost.
2 2. Time: for the similar reason stated
above the Inmon’s approach require more time to implement when compared with Kimball’s. (Kimball can be considered a short term, quick win solution to the data issue)
3 3. Maintenance: Since initial time is spent in building the
complete exhaustive EDW hence maintenance is easier. When we analyze Kimball’s
architecture, it is difficult to maintain because there the DWH would face the
issue of redundancy quite often.
4 4. Skill Requirements: Since Inmon’s
system require an EDW, high skill set is required to develop such a structure
as compated with Kimball’s approach.
5 5. Frequency of changes required: Low in
Inmon’s and high in Kimball’s
Which is better? An Opinion
There is no one answer to which is
a better approach or which one to choose. The complete decision depends what
requirements and constrains the organization has. The scope of work is very important.
For instance in a banking sector Inmon’s approach may not be very useful as
customer’s information such as name, age, average balance, bank accounts, MF
schemes have nothing to do with average asset liability of the bank or marked
to market loss on daily basis on Fx transitions. Thus there is no need to
connect the customer information.
In another situation such a
manufacturing firm: In this every department has to talk to all other
departments to see where there is a deficit and where there is surplus. In such
a case a central EDW (Inmon’s approach) serves the purpose where the complete
DWH contains all the dynamic and static information of all the departments in
one place connected to one another.
In my opinion other deciding
factors could be such as If organization has more time to deliver (> 3
months), can gather information from all the departments, the depth of cross functional
data requirements is high, the future ability for adjustment and rework to be
low then Inmon’s is preferred.
To sum up, if the company is
settled, has skilled resources, time and expertise then Inmon’s approach would
give better yield in future as it serves the most exhaustive and sorted form of
data source.
Regards,
Amandeep Singh Kalra
amandeepkalra86@gmail.com
References:-
The Data Warehouse Toolkit: The Definitive Guide to
Dimensional Modeling, 3rd Edition
