Re-Inventing the Internet

Re-Inventing the Internet Greg White

Greg White
Distinguished Technologist

Feb 8, 2016

The Internet Protocol – IP. It has become practically synonymous with networking, and over the last 40 years the adoption of IP has radically changed the world economy and the way that people interact globally, with CableLabs and the cable industry being a major force in that adoption. The past 15+ years in particular have been marked by the transitioning of the significant majority of industries, services, and human interactions into versions that are mediated by the world wide web, itself built on the foundation of the Internet Protocol. As this transition has taken place, the way that we use the network has radically evolved, from a simple point-to-point (or end-to-end) model with static endpoints, into one of massively distributed computing and storage accessed almost entirely by a huge number of mobile and nomadic client devices (with content and virtualized services being deployed on demand wherever it is most advantageous).

Alongside this usage evolution, network engineers have evolved in the way that they build IP networks, constructing ever more complex systems to try to manage the flow of communication, and have even taken on a change of the IP protocol (from IPv4 to IPv6) in an attempt to solve its most pressing shortcoming, the size of the address space. However, it is becoming more apparent that the address space issue is not the only problem with IP, and that we may be nearing the day when a more fundamental re-invention is needed.

A New Approach to Networking

Information Centric Networking is an emerging networking approach that aims to address many of the shortcomings inherent in the existing Internet Protocol. The field of Information Centric Networking was sparked by work at Stanford University in 1999-2002, which proposed a radical re-thinking of the network stack and a fundamental change in the way information flows in the global data network. Since that time, a number of researchers have developed proposals for the operational model, semantics, and syntax of an Information Centric Network, and today, one technology appears to be leading the pack and to be poised for potential success. This approach to Information Centric Networking, being developed in parallel by two projects: Content-Centric Networking (CCN) and Named Data Networking (NDN), appears to be gaining mindshare in the research community and in industry, and promises to significantly improve network scalability and performance, and reduce cost over a network built on the Internet Protocol. CCN/NDN provides native, and elegant, support for client mobility, multipath connectivity, multicast delivery and in-network caching, many of which are critical for current and future networks, and all of which require inefficient and/or complex managed overlays when implemented in IP. Further, CCN/NDN provides a much richer addressing framework than that which exists in IP (which could eliminate significant sources of routing complexity), and it provides a fundamentally more secure communication model, even for in-the-clear communications.

By moving away from a "host-centric" view of networking functions, where the core protocol (IP) is entirely devoted to delivering data from one specific host to another, to a "content-centric" view, where content objects are identified and routed solely by the use of globally unique names, the CCN/NDN approach provides a more elegantly scalable, faster, and more efficient network infrastructure for the majority of traffic on the Internet today. To get a sense of how big a mind shift this is, consider this: in CCN/NDN devices don’t have addresses at all. A device can retrieve content by requesting it by name, without needing to have a way of identifying a server where that content is stored, or even identifying itself.

At CableLabs, we are experimenting with this new protocol and are deep into investigating the applications that might drive its adoption. Because CCN/NDN is built with mobility, privacy and efficient content distribution in mind from the beginning, we see synergies with the direction that cable networks are going. We see this as enabling a convergence of fixed networking, mobile networking and content distribution – a “Fixed-Mobile-Content Convergence”.

As IP is baked into the networking field (equipment, software stacks, applications, services, engineering knowledge, business models, even national policies), it may seem daunting to consider the use of a non-IP protocol. However, while IP has been a phenomenally successful networking protocol for the last 40 years, as technology and time inevitably marches on it is reasonable to believe that we won’t be utilizing it forever.  And furthermore, while replacing IP with another protocol will certainly bring implementation challenges, it doesn’t follow that doing so necessarily means changing the way applications and users interact with the network. In other words, the web will still look and act like the web, but it will become more streamlined and efficient.

To be clear, CCN/NDN is not a mature protocol ready for immediate deployment. A number of R&D issues are still being worked in academic and industry research groups (including CableLabs), and implementations (both endpoint stacks and network gear) are still largely in the reference implementation or proof-of-concept stage. However, development work is active and confidence is high that the protocol will be ready for deployment in the next 3-5 years.

As part of this investigation, CableLabs has released a publication describing how an existing Content Distribution Network (CDN) evolves to one that leverages the benefits of CCN.

Greg White is a Distinguished Technologist at CableLabs.