ERwin supports the concept Design Layers, including support for synchronisation between models source and model target. However, this support is incomplete: only 1:1 synchronisation is supported (one source entity/table to one target entity/table; one source attribute/column to one target attribute/column).
In order to really support design layers we do need Persistent Transforms: transforms which are supported by sync with model source. In my opinion, design layers without persistent transforms are too cumbersome, since we cannot use tools like sync with model source to keep them up to date.
For example: in a data vault environment entities are divided over at least 2 tables, in a consistent way during the complete life cycle of the entity. So we must be able to say: attributes 1-3 go to table 1, attributes 4-10 go to table 2, etc. When something changes, we should be able to automatically propagate this from logical to physical model etc. In other words this transformation must be registered (persisted) in ERwin, and must remain available for reuse during the complete life cycle of this data (persistent transform).
I propose Persistent Transforms, to start with the following types of transforms:
- Roll up: combine attributes from more than one source object into one target object
- Roll down: divide attributes from more one source object into more than one target object
- Concatenation: concatenate more than one source attribute/column into one target attribute/column
- Split (“de-concatenate”): split up one source attribute/column into more than one target attribute/column
And obvious the current non-persistent transforms can or should be included as well.
By the way, I would like Persistent Transforms not only between Logical and Phsyical, but between any layer: also from Logical to Logical, and from Physical to Physical.
Please sign in to leave a comment.