Saturday, September 15, 2012

How Fast Can Bandwidth Demand Really Grow?

There is a running debate in the communications ecosystem about just how fast bandwidth demand might grow, either in the wireless or untethered spheres. A decade and a half ago, most executives and analysts might have suggested growth of 100 percent, or more, each year.

For a couple of reasons, those forecasts proved too high. There is a "law of large numbers" effect, for starters. Any new development or trend that is very popular, but starting from a low installed base, will show "big growth" numbers on a percentage basis.

Just as certainly, growth slows over time as the installed base becomes very large. So it is that annual year over year bandwidth growth now is more on the order of 40 percent than 100 percent. That's still a big percentage, meaning that, in the aggregate, bandwidth consumption doubles about every two and a half years or so.

But user behavior is highly responsive to pricing signals. People already have learned how to offload much traffic from their mobile devices to fixed networks (untethered use), since the per-biut prices of mobile bandwidth are significantly higher than rates for fixed network bandwidth.

There are other ways to shape demand, though. A higher use of unicast video (people watching YouTube video, for example) creates more demand, compared to any multicast method. That is why point to multipoint delivery of audio and video traditionally has been in "broadcast" mode.

So, over time, suppliers or users can shape demand by consuming more video in multicast mode than in unicast mode. One might argue that consumption is shifting to on-demand modes, but store and forward is a very efficient way of meeting much of that demand.

In other words, suppliers can broadcast content, but users can use DVRs to capture and store that content for later on-demand, or "near on demand" viewing. Users can choose to time shift their consumption as well, delaying "watching video" sessions from full mobile scenarios to offloaded sessions on their home Wi-Fi (untethered) connections.

Over time, one should not discount the amount of behavior change possible when price signals are sent about consumption modes. In other words, people will behave rationally if they know it makes a price difference when consuming video on a mobile, using the mobile network, compared to consuming that same video when at home, with demand shifted to the fixed network.

The point is that there is nothing "inexorable" about bandwidth demand. It can be shaped by suppliers or users, using price mechanisms, and will be shaped in such ways. Analysys Mason projects that mobile data in Western Europe will grow at a compound annual growth rate of just 29 per cent from 2012 to 2017, for example.

At a global level, Analysys Mason predicts that mobile data will grow by a multiple of 5.5, equivalent to 41 per cent CAGR. Over time, it is reasonable to expect continued growth, but the rates further are susceptible to changes in user behavior in response to pricing signals. 

In other words, there is nothing inexorably fixed about bandwidth consumption or rates of growth. Like any other product, higher prices will lead to less consumption, lower prices will lead to higher consumption. 

No comments:

Costs of Creating Machine Learning Models is Up Sharply

With the caveat that we must be careful about making linear extrapolations into the future, training costs of state-of-the-art AI models hav...