Boston is a famously tough place to drive, largely because the city’s so old: Its roads were laid down hundreds of years ago, with no consideration for cars, and over time highways and exit ramps have been awkwardly grafted onto the existing infrastructure. As a result, traffic snarls and drivers get lost easily. The roadway system is far from the one we’d design today if we were starting from scratch.
The infrastructure that makes the Internet work is a lot like that, too — a system of kludges to meet the demands of innovations like smartphones, 5G networks, and the “Internet of Things.”
“We see new apps and devices coming into play, new cases like cloud computing, which showed up almost out of nowhere,” says Aditya Akella, a computer scientist at the University of Wisconsin. “We design something quick and deploy it to make sure a new population of users and devices sees sufficient performance. It’s a very reactive approach.”
Akella is trying to change that, with an ambitious project to “future-proof” the Internet. His goal is to create network infrastructure that can adapt to whatever we need from it in the future, while remaining fast, secure, and flexible, with a low failure rate. “Future-proofing asks whether our networks can continue to offer these good properties in the face of disruptive applications,” he says.
Akella cites geolocation technology on cellphones as one example of how quick fixes have created problems down the road. Long before smartphones we had desktop computers, whose geolocation is designated by an IP address. Content distribution networks decided to use the same IP address-based system to geolocate cellphones, but it doesn’t work as well, for the obvious reason that phones move around and because of the way cellular providers allocate IP addresses to phones. “Cellular geolocation is horribly broken,” Akella says. “The typical geolocation techniques you’d use for wired users didn’t apply to cellular.”
Fixing these kinds of problems isn’t easy. Part of the solution comes from hardware and software manufacturers rolling out better products. Akella’s lab developed software that speeds data transfers, and Microsoft has incorporated it into its Microsoft 8 package since 2012. Another strategy is to try to get the infrastructure for new technology right from the start. Akella is currently running a project called Cloud Lab that lets engineers around the world experiment with different ways of building cloud computing infrastructure.
But Akella says there’s only so much that individual companies can do to improve things. After all, it took the monumental Big Dig to improve Boston’s highway system, and something similar my be needed for the internet.
“Working with the government to force certain things may be crucial,” Akella says. “There are certain changes that are very hard for established players and incumbents to make.”Kevin Hartnett is a writer in South Carolina. He can be reached at email@example.com.