Like most people, I have always considered that Internet capacity would automatically expand to cope with any growth in use. But is this really the case? A recent post on Slashdot got me thinking.

Internet bandwidth

The problem is one of bandwidth, or the volume of data passing through the network at any one time. This appears in a couple of different forms:

  1. A bandwidth usage cap, usually presented by ISPs as a monthly figure.
  2. The data transfer rate given as the number of bits, packets of information, transferred each second.

Monthly allowances can easily be converted, by dividing the bandwidth cap figure by the number of seconds in a month, to give the average monthly data transfer rate. Figures for a few typical internet applications are shown below.

Table 1: Typical monthly average data transfer rates for internet applications

Typical application Monthly bandwidth cap (GB) Average monthly data transfer rate (Mbps)
Low usage domestic broadband 2 0.006
Average domestic broadband 15 0.046
High usage domestic broadband 50 0.152
Low traffic web hosting 100 0.304

The transfer rates in Table 1 are very low when compared to a typically advertised broadband services at 1, 2 and even 8 Mbps and most corporate networks will capable of at least 100 Mbps.

What do you actually use?

Most users don’t carefully spread their use over the month and cause usage spikes when actually doing something like downloading a large file, watching video online or playing online games.

In response to new bandwidth heavy applications, check out the Web or magazines, you’ll find ISPs advertising new services with ever increasing transfer rates. This is great, but most connections aren’t capable of attaining the advertised rates.

Take my home broadband connection, I have no complaints over its reliability and it runs at a little under 1 Mbps. I find this to be adequate for my normal use and it can even cope with two of use surfing at the same time. BT think that they can squeeze 3 Mbps out of my line, but in this case the upgrading to the new 8 Mbps service on offer is pointless.

Is the limiting factor the copper wires to my house, at the exchange or at another point in the network entirely? I don’t know, but I’d be prepared to bet it isn’t the Internet backbone and online TV isn’t going to catch on in this household until it’s sorted out.

Being in contention

When using an internet connection you find that you are sharing its bandwidth with a number of other users. This is the contention ratio for that connection. Often you won’t all be using the connection at the same time and intensity, so you’ll be relatively unaware of each other.

Consequently the actions of individual users isn’t the prime concern, but rather their usage pattern on mass. For example, the mass adoption of telecommuting mooted as an emergency response in the wake of a flu pandemic is likely to result in a synchronised spike in network load. Many, I suspect rightly, expect that much of the Internet may grind to a halt, but not that it will collapse completely.

In contention terms, we’re all going to hit the connection bandwidth hard at the same time and the chances are we’ll each only come out with a small piece. Still it may be better than nothing, but as someone who has recently being using dial-up again I’m not so sure!

In the case of such emergencies, QoS features in routers already prioritise network traffic to maintain bandwidth access to more critical applications. It is possible to see that this can be applied to the Internet as a whole and may well be already.

It’s not just about the wires

The most vulnerable part of the system is likely to be the servers expected to deal with this huge increase in traffic.

Unless massive over capacity has been specified, faced with this situation the best outcome would be that most users would simply be unable to establish a connection with their destination server. The worst would be the same as a DoS attack, causing the server to crash.

Is the rapid mass adoption of remote working practices a viable response to pandemic scale infections even if you ignore the network limitations? For the majority I expect not.

Focus on the steady-state

So when pondering the strangulation of the Internet these emergency scenarios are a side-show. The main point is whether the regular adoption of higher bandwidth services can be supported by its infrastructure. In the case of streaming TV via the Internet, Google thinks not. The Internet wasn’t designed to stream this type of information in real time.

But that’s only video streaming, what about all the other web applications for which the Internet is better suited?

The term of the moment is Web 2.0; basically the adoption of a more social approach to building web content. Much of this is text, relatively compact images and the occasional audio file thrown in. None of this should cause significant trouble to the established networks as long as they keep pace with any growth in Internet uptake.

But will it do that? Who is responsible for the investment? It may be that bandwidth becomes more valuable in the future, like carbon dioxide perhaps, but there’s scope for greater efficiency in existing applications, such as:

  • Cleaner, more standards compliant code.
  • Image file formats with improved compression performance.
  • Better use of data compression at the web server.
  • More semantic code allowing machines to efficiently retrieve information.
  • Reduce the volume of web crawling required to assembly search engine indexes.
  • Get serious about reducing the amount of spam clogging the network. This includes junk email and comment spam on blogs and forums.

I suspect that these measures combined with others, identified by people much brighter than myself, would have a significant impact on our current bandwidth use. Could this issue be the Climate Change or tomorrow? We already have bandwidth trading after all.