How to model a network

Hello community,

I’m a bit stuck to model a network, considering my needs I have found 3 ways to model it but none of them fully satisfied me.

I’m using Capella 1.3.1 with the view Physical Architecture. The network I’m trying to model is quite simple, I’m only interested by the signals that circulate in the network and I’m not interested by any characteristics/properties/attributes of these signals (CDB diagrams…).

Here are the current needs that i’m trying to conciliate, the network shall :
N1 : connect several components
N2 : manage thousands of signals
N3 : have an unique object in Capella to model the same information sent from 1 component to several components
N4 : let to list easily all the signals composing the network (for example by visualizing in the Semantic Browser)
N5 : let to connect with additionnal FunctionalExchanges in order to create FunctionalChains (for example, a FC composed of 1 signal allocated to a network and 1 FE allocated to a cable)

Here is what I imagined at the beginning, see image “[PAB] network idea”
Examples of different ways to model it:

Solution 1 : a network built with a component and a function
see image “[PAB] network sol1”
. FE are cuts in 2 parts --> it increases the number of FE --> N2, N3 are not respected
. N4 isn’t respected (it requires an additional work to list all the signals)

Solution 2 : a network built with a component and PhysicalPath
see image “[PAB] network sol2”
. N2 is OK (note that PP shall be added in the model)
. N3 and N4 aren’t respected

Solution 3 : a network built with Interfaces and ExchangeItems
see image “[PAB] network sol3” (EI are displayed instead of FEnames)
Then using the ModellingAccelerators “Interfaces from allocated functions”, we automatically create Interfaces with the CE’s name.
Then in a IDB diagram we display the interfaces created and their relation with components see image “[IDB] network sol3”
Problem : I imagined to factor the 3 Interfaces in a “mother interface” which could inherite EI and let me know all the EI of the “child interfaces” but the relation “Generalization” is 1…1 (not 1…N). Any idea to visualize all the EI ?
. with this problem, N4 isn’t respected
. N1, N2, N3 are respected
. N5 isn’t respected (I have 1 FE carrying several EI and so I can’t know which EI “belongs” to the FC)

Any idea or comment is welcome :slight_smile:
[PAB] network idea.jpg
[PAB] network sol1.jpg
[PAB] network sol2.jpg
[PAB] network sol3.jpg
[IDB] network sol3.jpg

Hello Pierre,

At physical Architecture level, I would imagine that the network is a subsystem of your system, made at least from one or more node/implementation components (possibly switches, routers, firewalls and so). This or these components should be linked with others by physical links (Ethernet, or wifi for example), as in solution 2 / physical links.

Then, if the network has no specific functional behavior contributing to applicative functional chains, then applicative behavioral components(C1 C2 C3) and functions (F1 F2 F3) can be linked directly to each others, and associated component exchanges should be allocated to physical paths traversing the network component, again as in solution 2. No behavioral component would be allocated to the network. The number of exchanges is minimal (N1 and N2)

If (and only if) the network has some contribution to applicative behavior (such as filtering, encryption, etc) and if you need to capture it at your engineering level, then you will have to add behavioral components and functions in the network implementation component(s), as in solution 1, with functions and behavioral components of the network contributing to functional chains and interacting with applicative components by exchanges.

Note that the “unique object in Capella to model the same information sent from 1 component to several components” should be an exchange item, referenced by all functional exchanges carrying its information (N3).

If you need to list all exchanges traversing the network, you should be able to do so through a query, looking for exchanges allocated to a functional path whose some physical links are connected to the network implementation component(s) (N4).

Hello Jean-Luc, thank you for your reply

Considering what you said, I will probably go on solution2.
Only N3 isn’t respected but the network seems optimized considering the needs, I guess I can’t get butter and butter money :slight_smile:
To summarize the reasons of my choice: the network has no contribution to applicative behavior, the number of exchanges is minimal, exchanges are easily linked with FunctionalChains, a query can list all exchanges.

A question emerges… I don’t know when is it relevant to link several FE to the same output port.
For example, with the 3rd image above I could link “signal1” to the same FOP.
In this model I’m not interested by the realized FOP and the EI’s FE, but is there other points in the the Arcadia way of think which could drive my choice ? (-> one or several FOP?)


Copyright © Eclipse Capella, the Eclipse Capella logo, Eclipse and the Eclipse logo are Trademarks of The Eclipse Foundation.