A growing number of researchers think it’s time to rewire the Internet. A fundamentally new approach could better serve the streaming video, nonstop connectivity, and sizable downloads that users have come to expect, these experts say.
The problem is simple: the Internet was designed to send small packets of data back and forth in a conversational style, says Glenn Edens, who heads networking research at the Palo Alto Research Center (PARC). “Where we are today, the Internet is mostly used for the distribution of content like video, pictures, and e-mails,” says Edens, who is leading an effort at PARC to design and test an alternative way of operating the Internet known as content-centric networking—a project that is attracting increasing support from other researchers and companies.
The Internet is based on routing packets of data between the Internet protocol addresses, which look something like this: 18.104.22.168, belonging to particular computers. When a user loads a video on YouTube, his or her computer sends a message to the IP address of a Google server asking for the data to be sent back to its own address; Google’s server replies with the data in the same way; and the routers that direct traffic across the Internet do nothing more than dumbly pass these messages along.
Eden says that’s a waste of time. “All the infrastructure to do with machine addresses isn’t serving the main need of the Internet—to give people the content they wanted,” he says.
Under the content-centric networking approach, a computer trying to retrieve a video would request the content itself by its unique name, making for faster delivery. Routers would also have the ability to store pieces of content, further speeding the process up.
Under the scheme, when a computer requests a popular YouTube video, a router in the same neighborhood could send it back without having to pass along the request to Google’s server in the Bay Area. “As long as you’re not the first person to request a piece of content, there’s a high probability of finding it,” says Edens.
Edens says this could significantly reduce the cost incurred serving video or other data to large audiences, which should in turn reduce the costs of many services. It could also boost reliability, he says, because content wouldn’t travel as far.
The new way of networking could make for particularly noticeable improvements to the Internet connections of mobile devices, says Edens. When a phone or tablet switches between wireless access points, for example from Wi-Fi to a cellular network, it gets a new IP address, so any data connections are interrupted and must start from scratch. With content-centric networking, the process would be simpler, with no new IP address required, he says.
PARC first released specifications for content-centric networking in 2009, and it maintains an open-source implementation of the idea as well as a website designed to help coordinate research across a growing community spanning corporate and academic research labs around the world.
A recent conference saw contributions to the idea from Toyota, Cisco, and Alcatel-Lucent. “There are now a couple of deployments at a level that aren’t quite production and aren’t quite research,” says Edens. But he adds that an Internet service provider could deploy content-centric networking within its own network and save considerable money.
“The main barrier for deployment is not technical, but economical,” says Pablo Rodriguez, director of research at Telefonica Digital, a division of the world’s fifth-largest wireless provider, Telefonica, based in Spain. Although the content-centric approach could offer significant efficiency improvements, he says, the depth of investment in other solutions to content distribution will likely win out in the short term.
“Since the beginning of the Internet, we have been struggling at handling massive-scale content distribution,” Rodriguez says. Content-distribution networks, such as those operated by Akamai, which help companies deliver popular data more efficiently, are well-developed. “Only time will tell, but I fear that ‘best is enemy of the good’ may apply to content-centric networking for a while,” he says.