KIMBALL vs INMON

A presentation by W H Inmon

the essence of the difference between Inmon and Kimball

Inmon there needs to be a single version of the truth

data mart

data mart marketing

finance

sales data mart

mgmt integrated historical granular

data warehouse

single version of the truth

HR

the essence of the difference between Inmon and Kimball the question being answered what is the single version of the truth? what is corporate data?
data mart

data mart

data mart marketing

finance

sales data mart

mgmt integrated historical granular

data warehouse

single version of the truth

HR

the essence of the difference between Inmon and Kimball

Kimball a data warehouse is the union of all of the data marts

data mart

data mart finance data mart

a data mart is based on business function Ralph Kimball

sales

HR

the essence of the difference between Inmon and Kimball the question being answered how quickly can I build reports? how quickly can I do analysis?

data mart

data mart finance data mart

sales

HR

over time the architectures have evolved

1990 Inmon Single version of the truth

2000 Inmon an architecture corporate information factory

2010 Inmon DW 2.0, unstructured data

Kimball a union of data marts

Kimball conformed dimension

Kimball a need for integration

1990 Inmon Single version of the truth

2000 Inmon an architecture corporate information factory

2010 Inmon DW 2.0, unstructured data

Kimball a union of data marts

Kimball conformed dimension

Kimball a need for integration

Kimball is today where Inmon was in 1990 What has Kimball said to all of those people who followed his teachings in 1990?

the essence of the difference between Inmon and Kimball

1990 Inmon Single version of the truth

2000 Inmon an architecture corporate information factory Kimball conformed dimension

2010 Inmon DW 2.0, unstructured data

2020

Kimball a union of data marts

Kimball a need for integration

Kimball unstructured data belongs in a data warehouse

prediction in 2020 the Kimballites will discover that textual data belongs in a data warehouse

from an implementation perspective

appl

data marts
mktg sales

appl

data marts
mktg sales finance mgmt HR Engineering Production

finance
mgmt HR Engineering Production

daily refreshment of data

appl

data marts
mktg sales

appl

data marts
mktg sales finance mgmt HR Engineering Production

finance
mgmt HR Engineering Production

each of these lines must be crossed at least once a day

daily refreshment of data

appl

data marts
mktg sales

appl

data marts
mktg sales finance mgmt HR Engineering Production

finance
mgmt HR Engineering Production

n m+n

mxn

daily refreshment of data

appl

data marts
mktg sales

appl

data marts
mktg sales finance mgmt HR Engineering Production

finance
mgmt HR Engineering Production

n m+n

mxn

how many programs have to be written? have to be maintained?

daily refreshment of data

appl

data marts
mktg sales

appl

data marts
mktg sales finance mgmt HR Engineering Production

finance
mgmt HR Engineering Production

m which overnight batch processing window do you want?

n m+n

mxn

reconciliation

appl

data marts
mktg sales

appl
$32000 $1000

data marts
mktg sales finance mgmt

$32000 $1000

finance
mgmt HR

$1,009,087

HR

$1,009,087

Engineering Production

Engineering Production

in which environment would you rather do reconciliation?

appl

data marts
mktg sales

appl

data marts
mktg sales finance mgmt HR Engineering Production

finance
mgmt HR Engineering Production

in which environment would you rather add a new data mart?

from an architectural perspective

star schema (Kimball)

relational based data warehouse (Inmon)

star schema (Kimball)

relational based data warehouse (Inmon)

good for fast reports not a short term proposition good for a system of record

as an end user I am confused there are 17 data marts that have information and I dont know which one to go to. And they all have different information

every time there is a new requirement I have to start from scratch. And these darn data marts are hard to maintain. I have to build a new one every time there is a change in requirements

we have had data marts for five years now. We have 250 of them and only 10 of them are actually being used today

Ive got these auditors coming in and I dont have any data that I trust that I can show them

with Kimball, the star schema is the architecture

with Inmon, the relational foundation is only the start of the architecture

Interactive
Very current Transaction data

A p p l

A p p l

A p p l

Integrated
Current++
Textual subjects
Internal, external

Detailed
Simple pointer
S u b j S u b j S u b j S u b j

Continuous snapshot data Profile data

Captured text

Text id ......
Linkage

Text to subj

Summary

Near line
Less than current Textual subjects
Internal, external

Detailed
S u b j S u b j S u b j S u b j

Continuous snapshot data Profile data

Captured text

Simple pointer

Text id ......
Linkage

Text to subj

Summary

Archival
Older
Textual subjects
Internal, external

Detailed
S u b j S u b j S u b j S u b j

Continuous snapshot data Profile data

Captured text

Simple pointer

Text id ......
Linkage

Text to subj

Summary

the Inmon approach is a FULL architecture leading to DW 2.0. And DW 2.0 is a true full scale architecture

Interactive
Very current Transaction data

A p p l

A p p l

A p p l

Integrated
Current++
Textual subjects
Internal, external

Detailed
Simple pointer
S S S S u u u u b b b b j j j j

Continuous snapshot data Profile data

Captured text

Text id ......
Linkage

Text to subj

Summary

Near line
Less than current Textual subjects
Internal, external

Detailed
S S S S u u u u b b b b j j j j

Continuous snapshot data Profile data

Captured text

Simple pointer

Text id ......
Linkage

Text to subj

Summary

Archival
Older
Textual subjects
Internal, external

Detailed
S S S S u u u u b b b b j j j j

Continuous snapshot data Profile data

Captured text

Simple pointer

Text id ......
Linkage

Text to subj

Summary

DW 2.0 supports some really important architectural features - the life cycle of data within the data warehouse - the accommodation for very large amounts of data - the recognition that cost is the ultimate limiting factor for a data warehouse - unstructured data as an essential component - metadata as an essential component

ask Kimball how he supports unstructured data? ask Kimball how he supports metadata? ask Kimball how he supports really large amounts of data? ask Kimball how he supports archival data?

corporate data

structured data

unstructured data

the vast majority of corporate data is not structured

the Inmon architecture is complete; the Kimball architecture is not

structured data

unstructured data

Kimball

structured data

unstructured data

Inmon

Kimball Inmon
Interactive
Very current Transaction data

A p p l

A p p l

A p p l

Florida

Integrated
Current++
Textual subjects
Internal, external

Detailed
Simple pointer
S S S S u u u u b b b b j j j j

Continuous snapshot data Profile data

Captured text

Text id ......
Linkage

Text to subj

Summary

Near line
Less than current Textual subjects
Internal, external

Detailed
S S S S u u u u b b b b j j j j

Continuous snapshot data Profile data

Captured text

Simple pointer

Text id ......
Linkage

Text to subj

Summary

Archival
Older
Textual subjects
Internal, external

Detailed
S S S S u u u u b b b b j j j j

Continuous snapshot data Profile data

Captured text

Simple pointer

Text id ......
Linkage

Text to subj

Summary

South America NYC Chicago Hawaii Sao Paolo Mexico Canada

Bermuda Denver Calgary Los Angeles Gold Coast Florida Miami San Francisco Seattle

Kimball only addresses one small part of architecture. Inmon addresses a much more comprehensive picture

data mart

data mart

data mart data mart

data mart

mgmt data warehouse

HR

how Inmon/Kimball fit together