Network Engineering for Audio Engineers - Part 3: Wide area networks (WANs) and the Internet

2.4 WIDE AREA NETWORKS AND THE INTERNET
We said we weren't going to delve much into WANs, since AoIP is intended to be confined to LANs. On the other hand, when we get into VoIP telephony and IP codecs later, we are unavoidably talking about the big wide world of WANs and the Internet, so it will be necessary to touch on the topic here.

By the way, some WANs are, in effect, LANs. For example, you could use an Ethernet radio to extend your studio LAN to the transmitter site. Because the radio works at network layers 1 and 2, there is no IP routing involved. There will also be very good quality of service. So, while the distance is certainly "wide," the two sites are linked in such a way as to effectively comprise a LAN.

2.4.1 The Internet
The original motive for the development of the Internet was to link up the local networks at a few university and military computing centers. Clearly, the designers' goals were modest in light of what their unpretentious project has since become!

But, thankfully, the spirit of the pioneer designers lives on. They were "get on with it" types who preferred to write code and try it out in the real world, rather than engage in lengthy theoretical debates.

More important, they wanted to construct the network in a way that was open and extensible, rather than locking it down in a closed and constrained fashion. (Tellingly, Internet standards documents are called RFCs, "Requests for Comments.") There is not much chance the designers had audio/video streaming in mind back when the Internet was getting started, but their approach to the design lets us do it today.

For our purposes, as audio engineers, the Internet has two characteristics:

• It's everywhere.
• It's unreliable.

The first offers tantalizing opportunity; the second, frustration.

Internet service providers (ISPs) are not able to offer any guarantees with regard to quality of service because most of the time they don't control the end-to-end path. It is atypical that both ends of a connection are being served by the same ISP.

The common case is that traffic must traverse at least two vendors' networks, with an Internet Exchange Point (IXP) or one or more third-party networks interposed between the two. IXPs are notorious for being overloaded, causing dropped and delayed packets. The third-party networks are often overloaded as well.

Economics plays a starring role in shaping the characteristics of the Internet. The Internet is cheap and unmetered precisely because it offers no guarantees.

As we've seen, the Internet relies on statistical multiplexing, with the expensive long-haul lines that form its backbone being dynamically shared. You might have a 4-megabit DSL line, but that definitely does not mean that you will be assured anything like this data rate end-to-end. It would be prohibitively expensive and impractical to build a network that could handle all subscribers running flat out at their full rate. This would be like the city zoo being sized to have room for everyone in town visiting on the same day. (Perhaps motivated by a particularly compelling Discovery Channel episode?) This is what statistical multiplexing is all about—making assumptions and observations about the nature of typical traffic patterns that can be used to guide a network's design, all within the framework of the ever-present, keep-the-customer-satisfied versus keep-the-cost-down trade-off.

There is nothing wrong with this. Indeed, as we've said, the Internet would not exist without taking advantage of this tactic. But it does mean that you can never count on the Internet as a 100 percent reliable transport for audio. It can reach "good enough" status for some purposes, but only when audio devices are designed for the inescapable unreliable network conditions. We'll meet such devices in Chapter 7.

You can improve your chances of achieving smooth-flowing audio by ensuring that both ends of a transmission are being delivered by a common ISP, thus avoiding the troubles caused by IXPs and third-party-caused bottlenecks. And some ISPs are better than others, surely. Each has its own idea as to where to set the satisfaction versus cost compromise. But to get a guarantee, you will need to arrange some kind of private or "virtual" private network.