First up: hardcore real-time peeps will take issue with the use of ".NET" and "real-time" in the same sentence, due to .NET's non-deterministic nature ;)
Having said that, if you're just implementing a supervisory or visualisation layer over an existing real-time system (say, implementing a SCADA-type system), then .NET should be fine. Then your network architecture can boil down to two scenarios:
- Clients poll from a server: you create a centralised server which contains much of your process logic, and clients poll from this server periodically.
- Server supports a publish/subscribe mechanism: clients subscribe to the server's information, and the server sends out updates when they occur.
There's no one "right" way to do the above comms; it depends a lot on size and frequency of updates, network traffic, etc.