Networking

How coffee, traffic jams and hotel pools can crash networks

Organisations can do as much network capacity design as they like, but people and events inevitably conspire to create largely unpredictable challenges that put paid to the best-laid plans.

The clustering of people and things naturally mutates into largely unpredictable network challenges in time, scale, location and type. Photo: Shutterstock

Written in my Singapore hotel room and despatched to TechRepublic at 60Mbps over a wired LAN.

We can extrapolate the progress of technology with reasonable accuracy, but we have little idea what people will do with it once it's been deployed.

For example, you only have to look at the rise of texting, instant messaging, social networks and the app culture. No one foresaw their evolution precisely or their unexpectedly adverse impact on networks.

The ideal would be to model people and machine behaviours to predict future traffic patterns, which could then be included in a network design. And while we got close to doing these things successfully in the era of telephony, it's an approach that has not been applied in recent decades.

Even getting the averages correct has become a challenge, and it is the higher-order moments and temporal data that is missing.

So, I think we can safely say that no one understands the internet or mobile networks.

Beyond the traffic characteristics of applications and services being warped by human activities, there are further unknowns associated with machines and inanimate objects that also play a role. A wider perspective goes something like this.

Fixed-line telephone networks were dimensioned for three to four calls per day, each lasting three minutes. But then radio and TV voting and phone-ins took off and this simple design framework was significantly distorted

Similarly, mobile networks were not designed to support text messaging, which started life as an engineering-support facility and not a service.

The mobile 3G network was never designed to support large numbers of people watching movies or interacting continuously on social networking. And then comes the category of unintended consequences.

A morning session at a conference is well attended and there's very little communication on delegates' mobile devices. At 10.30am coffee arrives, and the session breaks. Some 500 people try to make calls and get online, and the network crashes under the load. Coffee has become a strange attractor that crashes networks.

People driving along a stretch of freeway may generate one or two calls every mile. But when an accident occurs, a jam rapidly builds up. With traffic at a standstill, suddenly thousands try to phone or text home and the office in each one-mile section.

A YouTube video goes viral minutes after uploading, and the word spreads by email, text, instant messaging and social networks. Within an hour, millions are trying to view the clip, and of course the net struggles to cope.

The clustering of people and things naturally mutates into largely unpredictable network challenges in time, scale, location and type. Moreover, it seems mostly impossible to predict, and we know this problem is not about to go away.

So here is a new phenomenon to add to my list of network traffic disruptors. I'm in a hotel pool on the 57 floor at over 200m above ground level with a spectacular view over the city. As you might predict, people are posing and taking lots of pictures and making movies. But how about bathers in the water with tablets, Skyping their friends and videoconferencing around the planet?

I have just counted nine people with tablets in the pool at the same time. This sort of activity is unpredictable, and no doubt the wi-fi and 3G network is feeling the strain.

Today, we know a lot about the sociology of people and nothing of the sociology of things. So be prepared for some significant network surprises to come. And the way round these problems? Stop optimising. Instead, waste bandwidth and overprovision everywhere. It is the only viable option.

But the good news is that bandwidth is the lowest cost commodity we can produce and deploy - it costs virtually nothing per Gbit. Once the civil engineering costs are paid and the fibre is in the ground, then it really can be open season on bandwidth.

About

Peter Cochrane is an engineer, scientist, entrepreneur, futurist and consultant. He is the former CTO and head of research at BT, with a career in telecoms and IT spanning more than 40 years.

Editor's Picks