VoIP QoS should not be an issue if the network is managed correctly. This means the LAN, the WAN link and the core network of the service provider.
If the WAN link is an ADSL then it can be susceptible to congestion at the exchange though in my experience this very rarely happens, even at times of extreme network usage such as the Olympics or last summer’s cricket. Problems here typically stem from underprovisioning of bandwidth. An Internet Telephony Service Provider should also be operating an uncongested core network and a properly designed LAN should never give problems.
Problems do still occur but if you have the right tools these should be straightforward to detect and sort out. One of the ways we manage our network is by using probes embedded at key point in the network.
The diagram below shows the output of one such probe earlier this morning. We are looking at connection downtime, lost packets or packets arriving in the wrong order, jitter and latency or delay. It can be seen that latency is almost the only measurable effect. All the other numbers are too low to count. Even the worst case latency figure seen here of around 46ms is not going to be noticed by the human ear and most of the calls are below 20ms.
This is a very useful tool for IT managers having to run multiple services over a multi-site Wide Area Network and allows them to spot problems before anyone notices and starts to complain.