How to elaborate information flow in large scale distributed system by UML

Question

I am currently Designing UML Diagrams for a distributed backed made up of 8 nodes, which handle specific operations depending on the message received from the web Client.

The problem I have is that, the initial node which acts as the "Gate of Entry" to all messages and interactions generated from the Client-end, does not contain concrete operations that could be denoted as "Standard Use cases" but only a multitude of message flows pertaining to client-end use cases operations that flow throw this point.

But I also desperately require a method to show a process mapping between the client-end Use Case and initial node Use Case processes for consistency sake. If these messages are not mapped, the following the flow of operations from the Client-end to Server-end will be somewhat complicated.

Unfortunately due to the complexity of the system I have a problem of designing the whole System in a single diagram too.

Some Possible Improvisations considered

• To create "Place holder" like Use Case Ext. notation to indicate an extension. E.g. The client-end has a place new order Use Case from which a message flows through initial node to the back-end nodes. In order to create connectivity indicate a Use Case which would read Place new order Ext. and this would signify a connectivity but would only point to a message flow.(But I am not sure if this practice is largely excepted in UML.)

• Only rely on Sequence and communication diagrams aim to show how "things" interact, each lifeline designating one of the system components. But I feel is that the correlation between Client-end and Back-end Use Case will not be very clear.

I also checked for similar problems on this forum and there were a few but did not explicitly answer my question.so I thought of posting this question. Can someone please suggest what would be the best option in showing information flow in this highly event-driven distributed information system such as this - One of the above mentioned or any other options that I may have overlooked?

Answer 1

Have you considered using component diagrams, where the components have ports and the ports accept (and produce) signals instead of operations? A signal is basically a message / event.

You can also model signals as classes at the high level, or in as much detail as you need. You could model some of the properties of an application-level signal such that you can show how a component demultiplexes based on some property before forwarding the message along to an output port.