The protocols behind VoIP

Posted on 9 Jul 2010 at 16:25

The details behind the four main VoIP architectures

VoIP systems most often make use of four pieces of network architecture: QoS, multicast, SIP and H.323. The first and most important realisation about these pieces is they’re not from the same rule book: it’s a bit like linking together the smell of asparagus, tax rebates, luck and sunshine, because they can all be experienced by a human.

All four of the vital VoIP service and traffic types are related to a network, but to see how disconnected they actually are let’s start with the two protocols SIP and H.323: one comes from the ITU (International Telecommunication Union), the other from the IETF (Internet Engineering Task Force). Both have a wider scope than the use to which they’re put by VoIP, and neither come with handy testers showing you a red or a green light in any given situation.

Most LAN engineers are used to pinging and hitting email servers with Telnet; the toolbox for proving H.323 and SIP connectivity is nowhere near as well developed or understood.

The other two – multicast and QoS – are terms that cover a raft of technologies. These aren’t protocols so much as capabilities, so if H.323 and SIP are comparable to different models of car, then QoS and multicast are comparable in the same metaphor to books of law that cover how to use the roads.

So we have some simple definitions of these four architectures:

QoS (Quality of Service)

A means of defining how different types of network traffic are prioritised, for situations where you want to guarantee transport times for specific types of traffic. Voice chat without QoS in a busy network will stutter.

Multicast

An entire concept for using any IP network, up to and including the whole of the internet, for a kind of transport that goes against the grain of the rest of the system. Everything about networks is about making sure that packets reach their intended target; multicast is about ensuring packets go to every target in the scope of the network. VoIP systems use multicast for sharing group status data around a population of phones, which is a highfalutin way of saying that it helps you to see that the switchboard is ringing.

SIP (Session Initiation Protocol)

A data structure for initiating sessions of messages between two parties. In VoIP, SIP appears mostly as a means of communicating and authenticating with a proxy server (also called SIP Proxy) that puts call participants into contact with one another.

H.323

This has the simplest definition in the VoIP framework: it ships packets of audio data between VoIP resources, although not always in the same format and not in absolutely every implementation.

Finally, we need to mention Power over Ethernet (PoE), which isn’t a protocol as such but a means of delivering a voltage over a network. PoE is a convenient way of running small devices on the end of network cables, where a huge deployment of little fat power transformers under every desk would be both cumbersome and prone to failure (as well as spectacularly un-green).

PoE is delivered by network switches, or “power injectors”, which leads many people to wonder if they’re dangerous. They’re not. The PoE-providing switch detects whether the network lead is plugged into something that needs power or not. You won’t blow up any laptops by getting a PoE switch.

Author: Steve Cassidy