Sparx Systems Forum

Enterprise Architect => General Board => Topic started by: Rob A on March 21, 2011, 05:47:20 pm

Title: 4+1 architecture
Post by: Rob A on March 21, 2011, 05:47:20 pm

Hi All,

We use a 4+1 model to verify our architectures. There has been a lot of debate internally recently about the suitability and how to achieve this within EA.

I was wondering whether anyone had any experiences / suggestions they could provide to help us in this regard?

Are there ways to do this (or best practices within EA), or are there alternative methods to perform this verification process?

Thanks

Title: Re: 4+1 architecture
Post by: Guillaume on March 21, 2011, 06:38:33 pm

Hi Rob,

I never used the 4+1 model in a given project but did some investigations a while ago for a prospect client and it doesn't seem far from what I'm doing in EA currently i.e. requirements model, analysis model (use case, ...), design model, and architecture model.
The core of my EA project is traceability e.g. from requirements to use cases, by instanciating an object from a class, etc.

In the 4+1 model, if asked to apply this, I would create the following using EA when suitable EA default models :
- a use case model (the "+1" model)
- a domain/class model = logical view
- a business process model = process view, for my activity diags
- a component model = the dev view (component diagrams)
- a deployment model = physical view (apply components to nodes)

Of course this is just the EA project template to start with. The processes involved in running your project are to be defined and followed in a manner that makes use of your EA project structure, adapting it when necessary. However the project should be lead by the use cases and its scenarios

Title: Re: 4+1 architecture
Post by: Rob A on March 22, 2011, 09:30:27 am

Thanks for the reply.

We do much the same, though we are looking into how to 'verify' what we are doing. That is, we would like to use the +1 to ensure we have all elements of our logical model covered with scenarios.

Would you model all of your use cases in EA, or just a sample? If only a sample how would you know if you have sufficient logical coverage? (or too much)?

4+1 is probably not the only way to do this, but we work in a regulated environment and such I am looking at better ways to ensure we are 'correct' - well at least as correct as our logical model is.

Title: Re: 4+1 architecture
Post by: Guillaume on March 22, 2011, 06:11:17 pm

Hi Rob,

to answer your question, I would identify all use cases in EA.
This is what I do for all my projects ; use cases allow to define the scope of your system

You don't have to go into details for each of them (e.g. describing all scenarios and their steps)
Once you have all your use case items, you can build your traceability through dependency associations with other elements of your model.