Jacques Latour, IT director at the Canadian Internet Registration Authority, is something of a realist when it comes to a research project called Named Data Networking (NDN), which would overhaul our Internet architecture with a new, more efficient model.
“You would need a lot of beer,” Latour told me jokingly when asked what it would take to implement the new model.
The new architecture would replace today’s Internet address-based system (TCP/IP) with one based on unique, authenticated data packets that the network finds by name instead of a server location. The system would save on a lot of signal travelling by enabling devices to talk to each other directly and share the closest piece of data.
In other words, leave it up to the network to find the closest available location for the data your device requests, which could be stored in a nearby cache (with another user or a network node) or on an original server.
In a recent interview with The Wire Report, Latour said the NDN model could be possible within about 40 years, though that would require creating new standards, defining new protocols, getting the Internet community and global governing bodies to agree to it, ensuring vendors adopt it, and getting support from Internet service providers.
“Even then, I don’t know if it’s feasible. It’s assigning a name to a chunk of data, and in 40 years we’re going to have so many chunks of data, it might not be the right way to [be] naming stuff. It might be flawed,” he said.
He said although it’s an interesting academic exercise, he doesn’t see it as the future of the Internet.
“If you start from scratch, it’s a good idea. When you already have a billion users connected, to migrate to that is not an easy challenge,” Latour said.
He noted that other research projects are also looking at superior Internet code and architecture, such as the OpenDaylight project for an industry-supported open source networking platform called Software-Defined Networking.
Lixia Zhang, a professor of computer science at the University of California at Los Angeles (UCLA), one of the researchers leading the NDN project, told me the proposed system would create a lot of room on the Internet for new new networking applications.
“Today, devices next to each other often have to communicate via a centralized server hundreds of miles away. The NDN architecture will enable nearby devices to talk to each other directly,” Zhang told me the other day. “We hope it gets deployed on the Internet.”
The NDN model offers a more efficient approach to the Internet as traffic grows exponentially and more connected devices come online, including automatic devices engaged in machine-to-machine communications. Global data centre traffic will rise fourfold between 2012 and 2016, reaching a total of 6.6 zettabytes annually, Cisco Systems Inc. forecasted in a report released last October.
Zhang said the existing TCP/IP design of the Internet didn’t really have security built into it, whereas this new architecture has embedded security, with every data packet carrying a signature that authenticates it.
There appears to be global interest in the NDN project, which has about 15 participating researchers. The Institute of Electrical and Electronics Engineers’ InfoCom conference last year in Orlando, Fla., held two workshops on named-oriented networking, Zhang pointed out, while the Association for Computing Machinery’s SigComm conference in Hong Kong in August will hold another workshop, and Cisco has a team building an NDN router prototype.
“NDN aims to solve real problems. When NDN can help people solve real problems, people will want to deploy it. I don’t believe in government push to deploy new technology. A good new architecture must be able to solve problems that TCP/IP cannot,” Zhang said. “That problem, number one, is security. We’ve been working so hard to address security on the Internet.”