PARC hosts summit on content-centric nets

PALO ALTO, Calif. –  PARC, a Xerox subsidiary, will gather as many as 100 researchers in September for the first event focused on content-centric networking, a new direction for organizing Internet traffic. The approach promises greater security and faster connections to popular content but will require new protocols and changes in router chip and systems designs.

Content-centric networking represents a shift from today's focus on using network addresses to find content. Instead it proposes a protocol that specifically defines and tracks content. Backers say it represents an evolutionary change similar to IP forwarding.

"We think it's definitely a concept that will change how people design high performance hardware," said Jim Thornton, a principal engineer who leads a dozen researchers at PARC working in the area.

The idea is largely the brainchild of Van Jacobsen, an Internet pioneer who helped develop multicasting and trace route capabilities for Internet Protocol. About four years ago, Jacobsen brought the concept to PARC which has been gestating it ever since.

PARC won a National Science Foundation grant to develop the concept working with a handful of universities as part of the Named Data Network project started in 2010.

"They aim to provide an idea of the cost and complexity trade offs" of the concept, said Thornton. "The sense we have is this is doable, it won't kill us and forwarding hardware has always stepped up to the challenges new application demands," he said.

In a separate project in Europe, Alcatel-Lucent, Orange and French universities are working on similar ideas. "We know there are other companies beginning to look at this too, [but] in the router hardware community it's still early in the life cycle of these ideas," said Thornton.

PARC recently released open source code for testing content-centric networks. One goal of the September meeting is to gather researchers to share their work with the software. Their findings begin to shape the kind of content ID and router frameworks needed to make the concept work.

"We are trying to define and validate what you could think of as an Internet Protocol for content," said Thornton. "It's this kind of work with people experimenting with apps and use cases that will push toward the design of a core protocol," he said.

Content-centric networks hold the promise of pervasive caching of popular Web content—often far out to the network's edge--based on actual demand. Today caching is limited to content-distribution networks (CDNs) set up for content and network owners, typically set up near the network core.

The approach could ease the job of getting popular videos quickly to widely dispersed users.  It would not eliminate CDNs, but it could open up new business models for how carriers, content owners and CDNs create new caching capabilities.

The new protocols also would enable new levels of security and privacy. Content packets could carry digital signatures which could authenticate authorized users and verify no one has tampered with the content—services of interest to a wide range of providers from banks to Hollywood studios.

To enable content-centric nets, data packets will need to encode new information identifying its data type and where it fits in a content hierarchy. Router chips would need to inspect packets for that information, add new state information based on it and store the results.

Additional logic could help track optimal paths to the nearest similar data, stored in local tables. Systems may be able to use existing buffer memory and statistical multiplexing techniques for holding and maintaining the look up tables, Thornton said.