Information architecture is a matter (en)(Finnish edition copy)
Take the first review of Inmon’s model compared to Kimball’s model (original links do not work any more) – both of those guys are the pioneers of the information architecture and a very well-known influencers from the early 1980’s.
Inmon: All corporate information, further processed, “one truth”. A central data warehouse. Data Warehouse including historical data – the Corporate Information Factory (CIF)
Inmon’s CIF includes the data from the operational systems, data transfer and processing processes (ETL), data warehouse, and the establishment of the special needs of smaller data warehouses (data marts). Data Marts are always the data warehouse data, “one truth”, never data from anywhere else. These are the special needs of different departments and processes, analytical solutions, etc.
Data integration is time-consuming step. This work requires the utmost diligence, discipline. The entire organization is committed to information architecture.
CIF: A Data Warehouse should be located in a normalized relational database format. History-containing structures may be permitted to also be “de-normalized”, at least to some extent.
The work is a long-term, construction will last a long time, but in return is expected to be and should be long-lasting and reliable data architecture.
Kimball’s model is considered to represent “opposite” view of how the company will design and build architecture. Kimball’s model is also called “dimensional” model (stars, snowflakes).
In this approach, dimensional data structures (data marts) come directly from the organization’s applications. The same information can be transferred to more than just a single data mart, depending on the function of the individual models.
Interest and the criterion of this approach is the speed of development. Analytical and reporting needs can be quickly implemented when there is not a target to design and build the whole enterprise on a common data repository. Inmon other hand says that Kimball’s model used for the company there is no “one truth”
Kimball’s model is clearly based on a process orientation, business process needs of each individual. So far, this approach is often a business point of view clearer, or at least in principle it seems to be easier and clearer.
Inmon: Kimball’s so-called first phase of the data model – Maintaining will be a big pain, and an even impossible task. Data is not integrated and data is duplicated a lot.
Well, each of us has a certainty our own views based on our own experience. Both models have pros and cons, depending on business requirements, the urgency and potential resourced development.
Kimball’s model is further developed in the 2000s from that of the first phase, with a conformed dimensions, which will help the integration of different data models. And more recently has become the Master Data Management (MDM). MDM represents in Kimball’s model “one truth” representing enterprise-level information architecture.
Inmon compares these two views in “A TALE OF TWO ARCHITECTURES‘ (original links do not work any more).
Inmon believes that Kimball’s model has lost its fundamental interests, the construction speed, because the criteria for MDM enterprise-level data model is no longer possible to design and build a lightly and quickly.
Inmon guess also that Kimball’s model is based on its popularity in business intelligence industry- the sale must be continuously and rapidly. CIF architecture, design and construction will take from this perspective too long. Kimball’s high-speed data mart building model has provided a better basis for fast product sales.
That may be true. I think that this is not the only aspect of the matter. Inmon’s CIF aroused much interest in Finland in the late 1990 onwards. Many companies started to build enterprise-level data models and centralized data warehouse architecture. Construction phase began, however, might be too early. Business and management support was lacking. Things were “technicians” driven. And the time and resources spent simply too much. Cash benefits could not be seen. It is very natural drift towards faster development cycles, which more closely mirrors then Kimball’s designs.
Time is now different from 10 years ago. Business and technology maturity has evolved to broader understanding of information management architecture and construction. It is also interesting to note how Inmon describes in the publication “A Tale of Two Architechtures” Kimball’s model developed. And the best things about Kimball’s may combine Inmon’s DW 2.0 concept. You agree with Bill on that?
Bill Inmon explains more about DW 2.0 concept in this article (original links do not work any more).
But what about business process modeling – I still miss the obvious link between the processes and the data models. – Why? – Well, that we would not build a massive data architecture, and then begin to look for something to use it for. As Emiel van Bockel said in his presentation in Helsinki, construction must start by first identifying the information needed to base a decision. Rather than that we are looking for any decisions that we could use the “expensive” information.