Link: http://www.etc-architect.com/?p=103
From ETC-Architect » Architect Global | Infrastructure Architect, Global | Data Architect, Global
Often when talking with fellow data architects I experience that many data architects have lost the basic modelling skills and are usually just deploying the same pattern over and over. That may be excusable for a developer, but not for an architect. So usually the first thing I see is that most data architects will only touch OLAP and not OLTP, which is often a reason that OLTP applications have a crappy data architecture. Those OLAP architects will then usually also tell you that SQL is the only way while at the same time they will deploy MOLAP cubes. What they however really mean is that they think that NoSQL is actually only for losers, without understanding the concept. Usually they will follow it up with a dismissal of the old hierarchal model and then pointing out that the choice is really only between the 3NF relational and the star dimensional model. The strange thing is that they often even miss other SQL based models such as the popular dat vault, but since it was not invented 30 years ago it is discarded.
On the other hand I often meet the other and usually younger data architects that are coming in two shades. One type is constantly talking of open system data tools and even about the hardware stack such as Map Reduce and similar. Usually they try tools on a job and will then implement applications on what is just cool at the moment. Again for that you will not require an architect. The other type is usually deep into No SQL, but in that they are usually limited in similar small number of choices. They are usually well aware of the big table and of key value pair tables such as the ever attractive entity attribute
value model. Only a few then also know of the column orientated model. All database models that require a bit more math such as the semantic models on the graph theory are also left outside.
As long as data architects do not understand the full scope of all the underlying data models and the best models we will constantly encounter a lot databases that may be nicely fine-tuned, but since they are sitting on the wrong model a lot of functionality data architecture can offer is left out. For me data architecture still starts at the model with the question of the tool in a second or third place. For me at least this is tool driven design, not architecture.
