Representing versions and changes over time

[davidclark]

Hi
Our company uses EA (Archimate metamodel) to document our software, systems and architecture. We need some guidance on how to represent changes to our systems over time.
For instance, we use 5 software systems currently that implement some business functions. In a year, this will drop to 3 systems, and in 5 years, this will be only 2 systems. In these 3 different scenarios (current state, transitional state, and future state), the software systems map to differing subsets of the (same) business functions.
How is it best to map this in EA?
We can create 'versions' of EA elements, but how can we best re-use the same EA elements so that common relationships are maintained, but we can still show the change over time as things change?
Thanks
David

[qwerty]

This question rises quite often and the answer is always the same: there is no general approach. Simply because the matter is very complex and depends on the circumstances. You may try to search for similar threads here but I guess you will not find them too useful. I guess that baselining is something you will find quite often. I'm not really a friend of that since I don't think there is much sense in comparing models. Each model is a world in its own and the difference might be even more complex than each model by itself. That's especially the case if you have long term models as you indicate. You should have 3 teams each modeling independent models and talking to each other. You might extract some common elements you can share in a separate model (then maintained by a 4th group and read only by all other 3 groups).

You see: the matter is complex. Good luck anyway.

q.

[Ian Mitchell]

I agree with Q - this is a really, really hard problem, and the there are zillions of use cases.
I looked at the idea of a 'time aware EA' for a client a while back. We connected EA 'things' to 'time' things e.g a particular network 'Node' element to a '1Q2015' element, to say that this element only exists from first quarter of 2015.
Then using diagram scripts to create a diagram which can show 'stuff which exists in 2Q2015'. This worked OK for simple lists of 'what exists at any point in time'...
...but we fell down when thinking about connectors: hard to connect a connector to something, though possible e.g. with a GUID-typed tagged value. This is what you'd need if you want to be able to say that, though both elements exist at a point in time, they don't get linked together until another date.
All in all, very hard, but a problem which, if you could develop an approach, would get a LOT of people very interested.
Have a go, and let us know!