The key factor inhibiting the growth of VoIP is Quality of Service (QoS). In a nutshell, QoS is how voice traffic is given priority over other data traffic such as e-mail, JPEGs or MP3s that traverse a network.
QoS is a very complex issue. No one provider has it right. It currently works more by luck than by good design because service providers rely on all these Internet "clouds", which are public, to deliver their services. And there is no QoS on public networks because there are no cost benefits to the private network providers to implement it. Why would an ISP implement QoS when it has its competitors traversing its network?
But quality will come down to some key areas:
• The last mile -- the link between your home and ISP
• The ISP access to the network core
• The core segment
The last mile
The last mile is where users will encounter the most probable congestion point. The reason is that this is where you have the narrowest "pipe". Users may have a 512/256Kbps subscription, so if they send or receive large files it is very easy to fill that pipe with non-voice data.
The easiest way to implement QoS for a residential line is for users to use the Internet exclusively for phoning when making a call and not do any other form of Internet surfing during the call. But that is not practical and can be difficult if the Internet connection is shared among many users in a home or small business environment.
Many modems on the market are now tailored for VoIP use -- the modem identifies voice and gives it priority. Although VoIP is still in its early stages as a mass market technology, the QoS has improved in the past few years.
Access layer to ISP
The end user has no control at the access layer (ISP) from the ISP to the core of Internet traffic which is shared by hundreds of people. Presently, ISPs are not providing voice-aware QoS. However, what you will start seeing is ISPs offering voice to their customers, with their big differentiator being "if you buy voice and data I will give you QoS".
In the core of a network there is no QoS. However, QoS is not so much a problem here because the pipes are significantly bigger. In the core, a voice packet is in there with e-mail, MP3, JPEGs, TIFFs, PowerPoint files and so forth. The thing voice has going for it is that voice packets are small. So chances are, routers in the core will drop MP3s and e-mails before voice. This way the voice call experiences no latency and gets to the receiver as fast as possible.