Bartosz - In real life scaling scenarios you'll add multiple JVMs on a physical box each running 1 EAR.
All these multiple JVMs are clustered together.
Anyways I've decided not to show the JVMs, but mention the design in my notes. I think that'll keep the Deployment diagram simple.
What you could do is use a deployment diagram as a logical implementation of your deployment approach.
As a guideline of how to implement this approach :
* Logical servers specified in this diagram could actually be zones - virtual operating systems - on the physical servers deployed.
You could use hardware profiles to depict which logical server is part of an actual physical server
* Containers specified in this diagram could actually be virtual machines on the logical server deployed
Your deployment approach can be used as a guideline to specify what kind of artifact ( e.g. ear, war, jar,.. ) is deployed on a specific type of container
As a result multiple JVMs could potentially be used on one logical server.
* Artifacts specified in this diagram could actually be the set of components you've chosen to package together.
Using this approach will provide you :
* Roadmap of how to transform the components specified in the component diagram into artifacts deployed on an actual set of physical servers.
* Clear view on how,from a deployment point of view, you will actually implement the quailty of service required