Anyone using Archimate add-in?

[Thelonius]

Sparx provide it. I've had a look at it, and at the Archimate specification. It has not been updated, and looks a bit daggy. I'm thinking that anything I could do using Archimate, I could do using basic UML.

Having a hard time seeing the value, really. In theory, the value assertion seemed interesting, but in practice - now that I've had a closer look - I'm not getting an epiphany from it.

What am I missing?

[Geert Bellekens]

I had exaclty the same feeling about this archimate thing.

Geert

[Sunshine]

I've used ArchiMate notation to do some enterprise architecture modelling for around 6 months now so have gained some experience with it. I've read the book 'Enterprise architecture at work" by Marc Lankhorst et al and all the white papers published on http://www.archimate.org/ along with the ArchiMate standard v1.0 published by the Open Group. I like the simplified metamodel but it does have gaps. I think some of the views that help link the business layer to the application layer and ultimately the technology layer are great. However, I have to concur I do feel BPMN and UML with stereotypes can cover everything that ArchiMate can do.
  
Also I've found when passing my ArchiMate models on to my colleagues further down the food chain they have to change them to another modelling language. For instance the Business Analysts transpose the ArchiMate models into BPMN to do their swimlane workflows as they add the next level of detail. The solution architects translate the ArchiMate models into UML to add further design detail. So in conclusion there  seems to be more effort involved when using ArchiMate as it does not provide a base from which to add more detail.  

I've also found amongst the various documents published on ArchiMate there are a few inconsistencies which don't really help when trying to learn it.

Finally one thing that might well sink the ArchiMate ship is that the Open Group license states that whilst its free to use for internal purposes there is a charge to organisations who wish to use it for commercial purposes.  Where UML and BPMN are free to use and so in the current financial climate these are likely to be favored over ArchiMate.

I think to sum up I'll be looking at doing my next Enterprise Architect assignment using BPMN and UML to save my colleagues some grief.

Well that's my 2 cents worth.  [smiley=wink.gif]

[AlexY]

I tend to agree with your conclusions about Archimate.

In the same time,  while looking for the "right mix" of tools for my current projects, which involve a lot of modelling of existing systems to reverse engineer the configuration & design, I came to conclusion that the best approach is to use "instance-based" UML models to model the existing systems.
After googling the idea I found quite interesting research on the topic. The authors describe such "purely" UML-based instance modelling approach to model systems & Enterprise Architectures.
One from papers can be found here:
http://softeng.polito.it/torchiano/papers/APSEC2000.pdf

(1) It seems to me you guys are on the similar path to use "pure UML". Can you please share references (if any) to the approach you use? or maybe describe in a few sentences?

(2) How suitable is EA to implement "instance based" modelling as described in the paper? Ideally, the the tool should support verification of instance model (system model) conformance to schema (domain model) and it would be even better it would do such a verification "real time" - i.e. at the time of adding of new element/relationship into the model and not as just an verification operation to run when the model is ready.
Is this implementable with EA? I anticipate that yes - EA should allow for this in this or other way, and I will look into this later.. but anyway, any information about the extent of the current support in EA, clues and suggestions about "how" would be very helpful.

Thanks,
Alex.

[Thelonius]

Alex

Interesting. And thanks for sharing your knowledge and views.

The referenced paper said something that I'm not quite sure about:

"Current object-oriented formalisms, such as UML, focus on
describing class models and use instance models only for
depicting scenarios. Little attention is being devoted to
defining how complex systems can be structured in order to
conform with those class models..."

I would have thought UML would be quite devoted to defining how complex systems can be structured to confom with class models.

Even in view of the fact that the paper was written back in 2000.

In reading all the stuff I can about Archimate - I have not yet found a really cogent explanation that tells me "Well, you could use basic UML to achieve the same thing that you can model with Archimate - BUT HERE'S WHY USING ARCHIMATE IS BETTER."

It just seems to me, from my naive perspective, that Archimate is "yet another" modeling branch off the base UML tree. Which may not be a good thing.

[Thelonius]

Just found this on Archimate:

http://blogs.msdn.com/mikewalker/archive/2008/09/26/archimate-the-emerging-architecture-modeling-standard.aspx

From Mike Walker's blog. More relevant views here. I think you should be able to access this URL.

[AlexY]

Thelonius,

I would have thought UML would be quite devoted to defining how complex systems can be structured to confom with class models.

My understanding that the paper actually proposes to model real system (or enterprise) and their parts (subsystem) as instances of classifiers - i.e as objects (instances of classes) in object diagram, and to do such modeling in parallel with building of domain model consisting of the classes of the objects.
Though such approach is one from obvious UML use cases, for some reason I have not come across descriptions of practical application of the method (apart from the paper referenced earlier and a few of other papers by the same authors written back in the end 90x and beginning of the millennium).
On the other hand, we observe efforts to achieve similar goals with profile based UML extensions like SysML and now with Archimate. The latter one is not mentioned as an extension of UML, but it is obvious that UML tool vendors utilize UML profile mechanism to get ArchiMate implemented.

In reading all the stuff I can about Archimate - I have not yet found a really cogent explanation that tells me "Well, you could use basic UML to achieve the same thing that you can model with Archimate - BUT HERE'S WHY USING ARCHIMATE IS BETTER."

This is actually the question I want to find answers to as well. what's wrong with UML that we need yet another modelling language? but I ask the same question in a wider context: not only why can't we use UML for EA modelling, but why can't we use UML for system modeling or business process modelling?

One from answers I am reading in Marc Lankhorst's at al. "Enterprise Architecture at Work" - the founding book on ArchiMate from language authors.

"The ArchiMate approach can be contrasted with he original approach in UML, which we describe in Chap. 2. In this approach, semantics was explicitly left out of the program. People who used the models could develop semantics for them, but a general semantics was not supplied. This approach stemmed from the origins of UML as a combination of three existing notations that did not have formal semantics. Hence, the focus of UML was and is on notation, i.e, syntax, and not on semantics. Although some of the diagrams of the more recent versions of UML have a formal semantics (see, e.g., the token-based Petrinet-like semantics of activity diagrams in UML 2.0), there is no overall semantics for the entire language.
We have taken a the opposite approach. We do not put the notation of the ArchiMate language central, but rather focus on the meaning of the language concepts and their relations. Of course, any modelling language needs a notation and we do supply a standard way of depicting  the ArchiMate concepts, but this is subordinate to the architectural semantics of the language".

Because UML is largely semantic-less, the modeler is in charge for defining of semantic for the notation he uses in his models.. Returning to the UML vs Archimate question - nothing prevents an architect to use ArchiMate semantics with UML notation.  Additionally, the notation can be extended with UML profile based extensions to be ArchiMate compatible (and I guess, this is what Sparx has actually done to implement ArchiMate in EA).
The same can be done DIY, but such a move,  comes at a cost that an architect needs to spend a lot of time defining semantics (if no Arhimate used), mapping semantics to notation, communicating concepts to his audience.

As a conclusion:
(1) Major value of Archimate is in semantics it defines. Understanding and common understanding of semantic is the hardest part in any more or less substantial modelling effort, and ArchiMate seems well-suited to help solving of the issue in the field of EA.
(2) It provides cross domain mapping of concepts, which no other language had in scope so far;
(3) As ArchiMare is being standardized (OMG plans & actions) and eventually (at this stage 50/50, but I would bet it will happen) it will lead to adoption of the language on larger scale with all benefits of standardization to follow (focused attention of best brains in the field, availability of tools & skills, overall growth of  importance of EA as a discipline).
(4) The alternative to ArchiMate is to use UML notation instead of ArchiMate with ArchiMate semantics, but this case involves getting some DIY job done to define the mapping b/w notation & ArchiMate semantics, which at least takes time.. and if UML & ArchiMate are available as a part of the same tool in my organization, I think it make very little sense not to use ArchiMate.

By advocating ArchiMate, seems I contradict to my yesterday's statement, but should note this because ArchiMate is not directly applicable to the kind of modelling I need to do right now, which is more software system bottom-up modelling in 1 domain or maybe 1.5 domains, rather than EA modelling in 3 domains (business, appl/data, technology).