Choose the right WAN technology for VoIP
If you're implementing a WAN and you plan to use Voice over IP, you should give special consideration to the technology you use. Each technology has some advantages and some disadvantages and they also have different levels of efficiency because, among other things, they all have different header structures. There is an inverse relationship between header size and efficiency: the larger the header the lower the efficiency.
To understand this, remember that the smaller your packets, the less time it takes to process them. This is particularly important with delay. Also, there is a finite amount of bandwidth on any given link. The more bandwidth is consumed by headers, the less bandwidth can be consumed by the payload (your data). Efficiency is good, so, with all other things being equal, the technologies with smaller headers will yield better quality voice than technologies with larger headers.
As an example, by taking the estimated number of voice packets per second, multiplied by the output rate of the CODEC plus the headers, we can calculate how much data we will transmit per second. So a Point-to-Point Protocol PPP frame has a 6-byte header and a Frame Relay header is 4 bytes. An ATM cell is fixed at 53 bytes; headers take up 5. If we compare these three common WAN technologies using the G.729A CODEC at a default rate of 50 packets per second, we see that a single conversation will take 25.6 Kbps of bandwidth on a Frame Relay link, 26.4 Kbps of bandwidth on a PPP link, and a whopping 42.4 Kbps of bandwidth on an ATM link.
Of course, efficiency isn't everything. All that header-space in ATM buys you something fairly deterministic, and in a congested network, that can be much more important than the relative efficiency.
Thomas Alexander Lancaster IV is a consultant and author with over ten years experience in the networking industry, focused on Internet infrastructure.