On 31-aug-2005, at 10:04, Mark Foster wrote:
I do see a risk in the move toward IP systems at the edge. At the core is a different story to at least some degree. Twas also pointed out that British Telecom are heading down the same track as Telecom NZ, and their rollout should be completed earlier. I trust therefore that it has all been thought out in terms of robustness and the like.
There are two types of VoIP: voice over a private, tightly controlled IP network, and voice over the public internet. Now obviously the latter is a risky proposition, as it imports all the limitations of the internet into the voice service. Apart from the fact that many parts of the internet aren't all that robust (but some are), this is a problem because voice and IP react differently to congestion collapse, which invariably happens to some degree in big emergencies. With IP, delays and packet loss build up, slowing everything down, but allowing many protocols to continue to work at a reduced rate. With PSTN, initiating calls starts failing more and more, but when you get through, you generally get to talk because you get a reserved piece of the scarce bandwidth. With VoIP, packet loss and delay eventually make the service useless. So VoIP fails harder than either traditional IP apps and PSTN. However, voice over a private network isn't entirely trouble-free, even though the private network can be designed such that congestion is a less fatal problem. And it does have the advantage that it allows IP routing protocols to route ongoing calls around failed parts of the network. On the other hand, in a circuit switched network you can do all kinds of interesting stuff (such as restarting all your control software) without breaking your sessions. We're only now seeing this in IP, and I think it's not really possible to reach the same levels with IP routing even in the long run. And then there is all this SIP stuff, which I'm (thankfully) only superficially familiar with, but never seemed particular robust to me. And voice over any kind of packet infrastructure introduces significant additional delays. I think in 10 years or so we'll realize that TDM isn't so bad after all.