We currently are in the process of installing an expressway for our collaboration edge deployment.
We are using the 2 NIC solution for the Expressway E. Currently External Users can log in, make audio and video calls. Internally Jabber users log onto the CUPS or UCM for their calls.
I may have missed something in the deployment documentation for remote and mobile deployment. But I cannot see any reference to how the jabber client handles moving from the corporate network to the cellular data network while a call is established.
My understanding is this.
Jabber client performs DNS look ups to determine what network it is located in. If it returns _cups-login or _cisco-uds it will connect internally to the CPUS or UCM servers. If those 2 request fail it will find _collab-edge which will enable edge communications mode and begin communications with the expressway E external interface.
OK got that.
In the 2 NIC expressway E scenario, When a jabber for android, mobile or even windows client is on the corporate network and establishes a call, and while on this call moves to the external Internet network, How is the call preserved?.
Currently we have a message stating the call is preserved, but nothing happens until the client returns back to the corporate network.
I can understand this in our network design as the client has established a session with its peer as peer to peer, then proceeds to the internet and establishes a connection with the expressway, but how is the media handled from there? It will not go to the expressway.
So my thoughts are that hairpinning is needed top force all clients internally to be connected to the external address of the EXPE. This would keep the client registered to the same peer when jumping between networks.
Has anyone done this? is it standard practice?
A hurdle that arises with this is making sure that the _cupslogin and _uds DNS records are not dished out to those clients that traverse both networks, only sending the _collab-edge record. SO therefor making the client always connect via the expressway external interface which requires hairpinning / NAT reflection. Then separate DNS servers or DNS filtering is required to make sure the internal only clients attached to machines that don't traverse the network are giving the _cupslogin record so they connect to the CUPS server.
Is my theory correct? Can anyone point me to some documentation that explains this in more detail.
Thankyou for your time.
