Content-Centered Internet Architecture Gets a Boost
Research to improve the underlying Internet architecture got a boost in August with a $7.9 million grant from the National Science Foundation. Named Data Networking (NDN) aims to enable direct access to network content by name, without reference to specific source and destination machines, as today's TCP/IP transport suite requires.
"NDN is the next step in architectural evolution to achieve the goal for delivering data in the most efficient, robust, and secure way," said Lixia Zhang, computer science professor at University of California, Los Angeles (UCLA) and recipient of the IEEE's 2009 Internet Award for her contributions toward developing the Internet's architecture.
Zhang will lead the NDN project, which includes researchers from eight other universities. Additionally, Van Jacobson, a research fellow at the Palo Alto Research Center (PARC), will be the NDN chief architect. Jacobson launched PARC's content-centric networking (CCN) research, which is contributing early protocol specification and open source as a base for the NDN project.
A Conversation That Became Ubiquitous
Early Internet research focused on interconnecting a relatively small number of large computers, mainly for resource sharing. This focus led to the development of IP and a domain name system to translate an HTTP query into a call against a specific host machine. The combination of content identification with host location worked well for the conversational traffic that characterized early Internet applications.
But the use of the Internet has fundamentally shifted, said Jacobson, who wrote the original TCP/IP header compression and is well-recognized for his contributions to the current Internet’s technological foundations. Today's users are mostly interested in information retrieval and distribution, and the machines they use are ubiquitous and often mobile.
This shift has created a disconnect between the way the Internet is used and the way it's modeled. The disconnect is complicated by the emergence of special techniques, such as network address translators, to make many physical machines appear as one. These techniques have added to the programming overhead and complexity of data-intense applications and mobile computing.
NDN would resolve this disconnect by using data names, rather than IP addresses, as the central communication construct. Each piece of content can be encoded using public key cryptography (PKC). In the most basic case, NDN uses PKC only to authenticate each piece of content and confirm that it has been transmitted accurately and fully. In more advanced cases, applications can also use PKC to encrypt data so that it can be hidden from third parties.
Building trust directly into the infrastructure lets architects and developers rethink the way they develop applications and distribute content. Applications would have to concern themselves with only the data’s accuracy, rather than the path it follows.
Securing the data directly means applications can behave the same whether the data comes from a router or a cache somewhere in the network. This will become more important as the cost of memory continues drop at a faster rate than network bandwidth, Jacobson said. "We have always treated moving bits across a wire as fundamentally different than putting bits into and out of memory or on and off a disk. This architecture unifies storage and communications."
A Lighter, More Secure Load
NDN could operate as an IP complement or replacement in a networking stack.
Programmers using NDN will find it easier to write highly scalable applications, said Tarek Abdelzaher, computer science professor who leads the University of Illinois' participation in the project. The architecture reduces middleware requirements because the network and applications see the world the same way. For instance, Jacobson wrote a highly scalable Twitter clone on top of NDN in about 20 lines of code.
NDN will also improve scalability by making it easy for special routers to cache frequently requested content, thus pushing popular content out closer to users. Abdelzaher said that inherently publish/subscribe content-sharing applications, such as Twitter, could see a 10-fold reduction in the computer and networking resources required to achieve the same performance. Today, when a site becomes popular, it increases the traffic load on not only the server but also all the network connections between the server and the client. With NDN, hybrid router/caching appliances could reduce the load on the network backbone by automatically distributing popular content across the network.
Point-to-point applications, such as a voice chat client, wouldn't see the same kind of networking efficiencies, but they might be easier to write.
NDN also includes a more efficient data-repair algorithm. When a traditional IP network loses a packet, it must retransmit from the source, burdening all the routers and links along the networking path. NDN allows a packet to be retransmitted from the closest good copy. Jacobson said that this could help in wireless networks, where most of the loss occurs at the network’s edge.
Finally, security is built into NDN at the content level, which provides an implicit chain of trust in all communications. Any network is only as strong as its weakest link. Consequently, traditional IP security techniques had to protect each host with its large collection of software. NDN protects the data itself, regardless of whether any intervening hosts have been compromised.
Evolving the Name Space
NDN represents a new networking paradigm and requires significant research before it's widely adopted. Zhang said that getting the naming system right will be one of the biggest challenges. "Fortunately, we don't have to wait until we understand this name-space design completely before we get started," she said. "As long as the name space interpretation is not hardened into the infrastructure, we have the flexibility to evolve the name space in parallel with the effort of the infrastructure design and deployment, which was the strategy that worked for IP."
Jacobson expects network carriers, telcos, and cable companies to be early commercial adopters. In the long haul, he sees NDN changing the way data is shared across personal computing devices. Instead of transferring movies, songs, and calendar information to each new device we use, we can have our settings and content follow us. "Users won't have to manage the plumbing to get what they want," he said.
For more information, see www.named-data.net. For related code, see www.ccnx.org.
George Lawton is a freelance journalist based in Guernevilla, California. Contact him via his website http://glawton.com.