Content-centric Networking - How IT Works

How It Works

The existing Internet is a tree of physical equipment to connect streams of packets from any leaf, to any leaf. It is efficient for communication, but not for distribution. The general proposal of content-centric networking recognizes that a great deal of information is produced once, then copied many times. Therefore, it makes sense to distribute the copying and any related activities into the network's tree of equipment. In many cases, substantial storage is already available, and could be used more efficiently if it could recognize particular content and only keep one copy of it. Since the network equipment is tree-shaped, it naturally scales content delivery to the size of the audience, and simultaneously reduces up-stream equipment to just the minimum needed to produce the content. As network service is built out, the content delivery naturally increases at the same time.

Content-centric networking uses a practical data storage cache at each level of the network to dramatically decrease the transmission traffic, and also increase the speed of response. The cache envisioned by CCN is a packet-level cache present at each node in the tree of network equipment not a complete copy of some media file. In that way, the worst case is that everything behaves as it does now: A consumer requests some data and it propagates through the network. However, the second time the data is requested, if it is still in the cache at some level, there are dramatic savings. A number of studies have verified the potential benefits of this approach.

One important issue with content-centric delivery is to assure that the name of the content sufficiently describes the information. For example, the name should include a version, so data can be corrected. It should also include a cryptographic hash, so that the content can be authenticated. The name also needs detailed information about the decoding format. In the case of video, an especially demanding case, it should describe the transcoding format. The general scheme appears able to accommodate all three of the common video distribution systems: Individual transcoding streams, quality layers, and even dynamic recoding. In the case of dynamic recoding, perhaps only a single recoder is required, and then the network caches the new transcoding.

If the content is cryptographically signed, the consumer's equipment can also validate the data easily, and tell upstream devices when it is corrupted. Since the validation complaints can and should be verified at each node, the rejection will propagate only to correct caches with bad data. Hackers will not be able to deny service in the network with lies about hash validation.

A content-centric network is also well suited to provision mobile devices. When content is used, a connection does not need to follow the mobile device. The data needed can be requested anew, and may often be immediately available at new network nodes because it has already been cached.

In this model, the logical place to put commercial copy control and security is not in consumer equipment, but in the neighboring commercial network nodes. If the node agrees that the consumer has a distribution agreement, then restricted content can be delivered. Such delivery contracts require relatively few, cheap CPU cycles from devices already present near the edge of an ISP's net. If there are commercial restrictions, those may need to be included in the content names, as well.