There were nearly 10 million VoIP subscriber licenses sold by Voice-over-Broadband server vendors in 4Q07. There are many countries in the world that would require less than that to serve the entire population. While the broadband telephony marches on there are certainly some limitations.
I have, for example, not received a satisfactory answer to a couple of questions. (1) It is still a best efforts service and therefore how do you make sure that you maintain QoS and call priority across multiple backbones and access networks? (2) How do you make sure that you keep the dialtone going during ultra busy hours or at times of emergencies?
You would get MPLS type answers thrown at you for the first question. But to think that all the carriers involved in transporting your international VoIP call will agree upon the packet priorities is truly impractical. But I will mainly discuss the second issue today. In the PSTN world your local exchange was engineered with a 1:10 ratio i.e. 10 outgoing circuits (from local exchange to regional exchange) for every 100 subscribers. So if there are 10 simultaneous active calls going on and you pick up the phone, you will not get the dialtone. In the broadband telephony world, the service provider typically manages similar ratio so that if the number of simultaneous call sessions exceeds 10 percent of the subscriber base, you will get no dialtone (and if you do, it is likely to be a false dialtone).
In the dial-up days there was an obsession around the switch congestion problem caused by Internet access. This was the case particularly in the US, since local calls were not metered and Internet users engaged their phone lines for hours. Regulators like the FCC made a lot of fuss about the issue. I don’t see anyone highlighting similar problems in the VoIP world yet. One thing is clear: VoIP networks, like PSTN, will not be engineered with 100% guarantee of a dialtone either. But you can perhaps do certain things to improve the situation. I thought of the following:
1. You could do dynamic negotiation of the compression codecs so that when there is higher pressure of calls on the system, it negotiates down to the lowest bit rate in order to accommodate more simultaneous calls.
2. Ask subscriber to pay higher price if he/she wants guaranteed dialtone. Let such subscriber have privileges to replace an active call session
3. Over-provisioning bandwidth around the call controller node
4. Block bandwidth hungry applications like video during busy hour sessions
5. With the use of FMC, force incoming calls onto other devices like mobile phones in order to make more room for outgoing calls
6. If during the dialtone crisis there is more incoming traffic and FMC systems are not able to handle them, perhaps have ENUM servers establish privileges for certain subscribers for their incoming calls so that they receive them in emergencies as well
7. Nodes connecting access network and backbones could be congestion points during dialtone crisis. So some form of dynamic over-provisioning of bandwidth at these points
8. Since broadband telephony is hosted, you can cluster several call controllers together so that Singapore based server is able to serve Italian customers during dialtone crisis in Italy
9. Through partnerships with PC-to-Phone companies like Skype offload your excess traffic over to them transparently
10. Integrate callback into your VoIP system so that if your subscriber is not getting through to your system, at least he/she is served with a dialtone via callback