Researchers at the Massachusetts Institute of Technology have developed a theory of information transfer in communication networks, which allows optimizing the network capacity and volume of data transmitted. The work is divided into two parts, the first of which was published in the journal IEEE Transactions on Information Theory. Preprints of articles are available in the archives of Cornell University. Summary can be found on the website of the institute. Working of most computer networks today is subject to the problem of reliable transmission of information regardless of the presence of noise. At the same time, the capacity of networks and amount of information transmitted is of secondary importance. Algorithm of the Internet is designed in accordance with this task: each node receiving a message (packet) shall transmit it further to the network to the address of recipient, while contents of packet do not change.
Scientists have proposed a different method of communication – network coding, which should significantly increase the amount of transmitted information in a network.
Network coding works as follows. Contents of different packets A and B merge into a node AB and are transmitted in this way to several other nodes. The node, to which two packets, such as A and AB are coming can, by small calculation, recover B.
Despite the fact that intuitively it seems that the described method over-loads the network by sending redundant information, it actually results in more network capacity. This is also due to the fact that the recipient receives the data in different ways and does not depend on “congestion” on its way.
Authors of the paper analyzed how noise level in the network affects its work, if information in it is not transmitted by classical way, but through network coding, and how to deal with it. In addition, they were also able to calculate the upper and lower limits of network capacity, built on this principle.
Network coding could help speed up the working of various communication networks, especially wireless, sensor networks, peer to peer (P2P) and so on.
Powered by Facebook Comments