Complexities of Relational Model and Simplicities of Dimensional Model

Hello,

Technology is changing very rapidly. Sometimes new database systems are developed and the old ones are not used anymore. Normalization brings a high level of complexity in model of relational databases. Because of such complexities models of real information systems may not be understandable in the future.
Do you think migrating them to dimensional modeling maybe helpful? Keeping in mind it is simple to understand?

thanks.

Hi,

do you distinguish beetween relational and dimensional? I distinguish beetween normalized and dimensional because it's very common to use RDBMS for dimenional realisation. Is that right?

I think dimensional modeling (or not) is a design approach. Relational modeling is a technical one.

Greetings
Markus

rf001 wrote:Hello,

Technology is changing very rapidly. Sometimes new database systems are developed and the old ones are not used anymore. Normalization brings a high level of complexity in model of relational databases. Because of such complexities models of real information systems may not be understandable in the future.
Do you think migrating them to dimensional modeling maybe helpful? Keeping in mind it is simple to understand?

thanks.

Normalization in relational models exists for very specific reasons independent of any current or future technologies that may exist for storing data. Properly implemented, a normalized database ensures data integrity in an OLTP environment. It is certainly possible (and expected) that technologies will evolve that would improve how one defines, documents and supports such databases, to the point of 'hiding' the complexities of a normalized data store, but normalization itself will still (and always) be there.

A dimensional model, on the other hand, is used for a different purpose. One would not build an OLTP application on top of a database created using dimensional design. Such an application would be far more complex than it would need to be as well as unreliable. Dimensional models are geared towards high volume data analysis, not data collection and maintenenace.

Yes, dimensional model is not for this purposes, but can it be used to minimize the complexities of data model?

"Technology is changing very rapidly. Sometimes new database systems are developed and the old ones are not used anymore." What I mean by this is:

Suppose, a university has a database system using some old version of oracle, and an application developed in VB. Since it was developed lots of developments have come in the technology and now they decide to develop a new state-of-the-art system from scratch. In this case the old system will not be used anymore but it will be needed for reference purposes (legal obligations or institutional memory). It becomes static or dead as it is not going to be used in day to day activities anymore.

What can be done to keep such a system accessible? Keeping in mind, that in the future, it maybe difficult to understand the structure of the database and as a result it may be difficult or at least very time consuming to query it for personel who did not develop it or use it before. Secondly, the application may become un-usable because of the changes in the technolgy, as it is dependent on other things like the Operating system. So it maybe really needed to access the data manually, without using the applicaton developed for accessing it.

In such situations, will migration of the database to a dimensional model be useful?

In such situations, will migration of the database to a dimensional model be useful?

Depends if the business wants it that way.

Creating a data warehouse requires significant effort. If there is a perceved value in doing so, then you can garner support from the business to do it and once completed, would be utilized. If there is no recognized need or perceived value, moving the data is a waste of time.

In dimensional world, no one has suggested replacing the OLTP system with dimensional (or OLAP) system altogether. They serve different purposes and should coexist, as said by ngalemmo, no matter how useful the OLAP system will become. Based on dimensional modelling, the separation between relational and dimensional structure should be at physical level, unlike some other methodology blurring the line.

However with old OLTP system (technology) being outdated, the migration process should be firstly from the old OLTP system to a new OLTP system, maybe an ODS, where normalisation (3NF) dominates the structure. Most of the time you would find the old OLTP system was built many years ago when relational technology was just started as the standard database structure, and they were not built as properly normalised OLTP systems. Almost all the OLTP systems I have experienced need major revamp regardless of the need for an OLAP system, whereases the push for BI system does speed up the revamp or upgrade process.

Like relational technology in its early stage, the dimensional technology was introduced and band-aided to many existing OLTP systems, which further exacerbates the problem in those legacy systems.

One important point to remember when building a dimensional system is, we are building a redundant data system in different structure. The ETL sits between the two structurally different systems and does all the magic to put them in synch.

Normalized data structures are not difficult to understand. If you can read a dimensional model, you can read a normalized data model.