UML for hardware specification

[Piotr Rezmer]

Hello,
I want to us EA to create document describing hardware specification of the servers. I did it using deployment model, components. Then I added tag values to the components:
model=
serial=
rackId=
CPU=
RAM=
HDD=

I changed to deployment diagram to show tag values.

Is it the proper way to make hardware specification?

best regards
Piotr Rezmer

[Luis J. Lobo]

Why don't you use "Nodes" instead of "Componentes" for Hardware?

[Piotr Rezmer]

Usually I use "node" for showing different sites. Then, inside node I show devices, with stereotype <<server>>.

[Sunshine]

I use nodes for hardware and components for applications or application components.  I use stereo types to distinguish different types of hardware on the nodes and tag values for things like RAM, CPU etc.
I use devices for end user devices like PC, laptop and execution environments for virtual machines. I use packages for sites.
Pretty sure I picked up this from a UML book like Architecture Bootcamp or something like it.
The Sparx websites have some tutuorials on using UML too
http://www.sparxsystems.com.au/downloads/whitepapers/UML_Tutorial_Part_1_Introduction.pdf

As an aside I note that ArchiMate notation does the same thing.  

If you use components for hardware what to you use for representing applications? :-?

At the end of the day the important thing is to get the message across to those that are building the system so I wouldn't go changing things until you start on the next project.

[qwerty]

I use node instances for the above plus Run State for RAM a.s.o. The node itself is the abstract form (e.g. Win Server) while the instance is a concrete device. It's also possible to nest nodes (e.g. mass storage or some device coupler).

Note that EA has a Node/Device/Execution Environment triple.

q.

[Piotr Rezmer]

I use devices for end user devices like PC, laptop and execution environments for virtual machines. I use packages for sites.
Pretty sure I picked up this from a UML book like Architecture Bootcamp or something like it.
The Sparx websites have some tutuorials on using UML too
http://www.sparxsystems.com.au/downloads/whitepapers/UML_Tutorial_Part_1_Introduction.pdf

I checked this pdf. Interesting that EA says nothing about using devices in deployment diagrams.

I noticed some advantages of nodes over devices:
-stereotypes have icons in upper right corner
-device cannot inherit from other device (nodes can)

If we set parent for a node, it will inherit tag values. So this way we can save some time by declaring required tags at parent, and only filling values at children.  However I found two weak points:
1. How to hide parent template from diagram[/list]
2. How to hide arrow to the parent
3. How  to setup parent for multiple nodes at the same time. (multiselection and ctrl+I doesn't work)

best regards
Piotr Rezmer

[Piotr Rezmer]

I use node instances for the above plus Run State for RAM a.s.o. The node itself is the abstract form (e.g. Win Server) while the instance is a concrete device.
q.

What benefit do you get when using run state instead of tag values? Did you manage to hide parent node?

best regards
Piotr Rezmer

[Piotr Rezmer]

Did you manage to hide parent node?

I managed that by creating extra deployment diagram with inheritance relation between nodes. Parent node is related to rack equipment, child node is a server. Then I drag the child node to real deployment diagram and select "as new Instance".

[qwerty]

What benefit do you get when using run state instead of tag values? Did you manage to hide parent node?

best regards
Piotr Rezmer

Well, I used that form since a long time. In former EA versions the Tags were not shown in a compartment, so not visible on the diagram. Still the run state shows without compartment which looks more pleasing to me. However, there's not real pro/con.

For hiding parent nodes: I don't understand what you mean. Can you elaborate?

q.

[Piotr Rezmer]

For hiding parent nodes: I don't understand what you mean. Can you elaborate?

I was thinking about a situation where I:
- create node
- add tags

Then I create another node and set the first one as parent to the new one. This way, new node will already have tags created in parent.  If parent is located on the same diagram as the child,it creates a mess (abstract - not physicallu existing parent and arrows between children and parent). But as I wrote in another post, this can be solved by placing parents on a separate diagram.

[qwerty]

To avoid that you can also place a diagram frame in your parent showing the child's connections.

q.