Network Coding May Make Networks More Efficient

Three researchers at MIT, CalTech and the University of Applied Sciences in Munich, Germany, have published a few details about problem areas that surround a new approach of writing code that may enable faster data transfers than we are used to today.

Two related papers that are published by IEEE Transactions on Information Theory discuss the possible combination of traditional coding as well as network coding to maximize the capacity of a network.

Network coding is referred to as a process in which each node scrambles data into hybrid data packets that are sent down not just on one path, but on multiple paths at the same time. The idea is that data packets that hit congested pathways and may be halted due to failed links, would still arrive via another way. Even in a worst case scenario, the arriving packets could be put together to the original message in most cases. Of course, the downside of this process is possible network congestion due to the increased traffic volume. The challenge is to combine traditional error-correction coding with network coding to maximize the network capacity and receive the greatest possible benefit without causing a network to break down.

In some cases, the noise arriving at a link may be "noisy enough" to require encoding for error correction, which is what traditional coding would do. "Suppose that I'm a node in a network, and I see a communication coming in, and it is corrupted by noise," MIT's Muriel Médard said. "I could try to remove the noise, but by doing that, I'm in effect making a decision right now that maybe would have been better taken by someone downstream from me who might have had more observations of the same source."

However, if there is a decision made not to run error correction, network bandwidth could be wasted. "If the node takes all the signal it has and does not whittle down his representation, then it might be using a lot of energy to transmit noise," the researcher said. "The question is, how much of the noise do I remove, and how much do I leave in?"

The scientists describes in one paper that network coding and error-correcting coding can be handled separately in a wired network without impacting that network's overall capacity. They also describe a method of how to calculate the resulting capacity of a network. However, that is a different scenario in wireless networks, which are explained in a second paper.

The difference is that, in most wired networks, the "noise" that can be or may not be removed at a node is unrelated to the signals traveling over other links. In wireless networks noise, the noise likely is related to the signals on other links, since neighboring base stations can interfere with each other.

Still, the researchers said they have found a way to calculate the network capacity when traditional and network coding techniques are involved, which would enable network operators to compare real world environments with theory and make decisions whether an investment in network coding can pay off, MIT said.