Unlock a world of possibilities! Login now and discover the exclusive benefits awaiting you.
Hello all,
I'm working on a Datamodeldesign that should be easily extensible (flexibility/maintance) and provide good performance (GUI) at the same time. The idea I have is to use tables that holds the properties of an object and link to the related object by inserting the ClassIDs to the object. There might be quite a number of classtables but just one objecttables. So for each classtable there is a column for this ClassIDs in the ObjectTable. The question I have is wether there are any concerns regarding this design. I'm clear about QkliViews preferred concepts (Dimensions/Facts/Starschema etc.). Below is an exhibit that should make it more clear what I mean.
Thanks a lot ...
Exhibit:
Hi
Just
I think structure for transaction system should comply it's requirements as the structure for analytics should comply its. In general there is no structure which is good for both systems. When you load data from transaction system into analytical system in general you do data transformation.
When the system is simple (as in your example) actualy there is no big difference. But as system becames more sophisticated, some rules must be met.
Generally speaking data structure of transactional system must be normalized and data structure for BI should be denormalized.
CHeers
Darius
I'm not sure I'm totally following your line of thought, but the overall structure looks a bit too normalized to me. I think your structure will multiply the number of links that will have to be resolved in the run-time, slowing down performance.
My instincts tell me - "keep it simple"...