The Peak Capacity Building Conundrum

In this interview taken at the Killer App Expo, the Technology Evangelist guys have a conversation with Larry Irving, co-chair of the Internet Innovation Alliance and giver of a rousing keynote address on the morning of May 2nd.

Among the many good points Larry makes about the need for continued investment in networks – something that IIA extols the virtues of regularly – a comment of his about peak vs. non-peak hours caught my attention.

To step back for a second, my grandpa’s been in the power plant building business for 50 years, and we often have discussions about energy related issues. While I often extol the virtues of alternative energy sources like wind and solar, he always comes back to the simple fact that despite the growth of these alternatives, when you’re building a coal-fired power plant you still have to design it so that it can handle peak loads for times when the sun isn’t shining or the wind isn’t blowing.

The same basic principle holds true for various parts of the Internet: when building out capacity, peak loads of simultaneous users are what matter more than cumulative use.

The challenge of the Internet, though, is that the number of people trying to use it is in a constant state of flux.

True, there are usage trends that generally hold on a day to day basis, like the fact Internet traffic is often at its peak during the daytime hours when people are at their desks at work. And overall growth in usage seems to be following a pretty clear trajectory (I’ll work on finding more specific info about this for a later post).

The beauty of the Internet is its distinctly democratic nature, allowing anyone at any moment to go online and access content or run applications. But herein lies the rub: at any moment everyone can go online. Or more precisely, everyone can try to go online.

We’ve seen the results of what can happen when everyone wants to access the same content at the same time. Just look at 9-11 and how a number of major news sites went down, succumbing to the crush of people desperately seeking more information about what was going on.

The really challenging part is that while the Internet as a whole is extremely resilient and adaptable, the individual networks, servers, routers, and the like have physical limitations that can easily overwhelmed.

And while everyone who owns and operates the hardware that enables the Internet are investing to increase capacity to some degree or another, it’s impossible for them to build out sufficient capacity to support the possibility of everyone coming online all at once.

So instead it’s more of an incremental process, where capacity builders try to stay ahead of the overall trends in demand without overextending their investment beyond what current demand for bandwidth will support.

Unfortunately, that means in the near-term we’re likely going to continue having to face the possibility of an overwhelmed Internet, in particular during the times of national crisis when we need it most.