by Lars Rönbäck, Stockholm University

Audience: Architects, DBA's and Application Developers.   Level: Intermediate

Link to the presentation S432. New Trends in Datamodelling

We need to rethink the concept of a data model. We live in a world where the available information changes more rapidly for every day that passes.

This is true both for the content and structure of the information.

On top of that, search requirements are also becoming inherently harder to predict.

In such a turbulent environment, current data models do not stand the test of time. At some point it is simpler to build a new model than to modify the existing one.

We must make it easier to reflect external changes in the model. The model must be flexible and adaptable. Furthermore, when information change we should handle the legacy data without causing any downtime.

Many of these problems are particularly obvious in the field of data warehousing. Some of the effects are:

- Current data warehouses are money sinks, as changes get more and more expensive to implement during its lifetime

- Current data warehouses do not give the desired return of investment, with maintenance constituting most of the total cost of ownership

- Current data warehouses cannot answer the really interesting questions, as they are modeled to answer a predefined set of queries and not ad-hoc searches

- Current data warehouses degenerate and become haphazard collections of tables, due to divergence from the initial modeling methodology

If we are to rethink database modeling, let us also try to do away with some other obstacles. Why is there a non-intuitive step in transforming a conceptual model to a physical one? Why is it difficult to make a generalized object-relational mapping?

In Anchor Modeling, a recent modeling technique, all this has been taken into consideration. Nevertheless, breaking into a scene ruled by old paradigms take time. To speed things up we realized that you need a tool, a free tool, and some new thinking on how you interact with your model. One in which a model is not built to last, but built to change. Only then can it truly last in an ever changing environment.

Lars Rönnbäck is a consultant at Resight, with a degree in mathematics from Uppsala University, who is specializing in customer intelligence. He has been working with some of the largest companies in the Swedish insurance and retail businesses, with ten years experience from the field of Business Intelligence and five years with Anchor Modeling. More precisely, he has been building data warehouses, designing SQL-driven ETL frameworks, and lately worked with customer analysis and data mining