Sunday, June 3, 2007

Network Comms - Background

Excellent background piece in the latest issue of Scientific American on network communications.

"The military has recognized the robustness of network coding and is now funding research into its use in mobile ad hoc networks, which can form on the fly. Such networks are valuable in highly changeable environments, such as on the battlefield, where reliable communications are essential and establishing and maintaining an infrastructure of fiber-optic cables or cell towers is difficult. In an ad hoc network, every soldier's radio becomes a node in a communications system, and each node seeks out and establishes connections to neighboring nodes; together these connections establish a network's links. Every node can both send and receive messages and serve as an intermediary to pass along messages intended for other receivers. This technique extends communications capabilities far beyond the transmission range of a single node. It also allows enormous flexibility, because the network travels with the users, constantly reconfiguring and reestablishing connections as needed.

By changing how networks function, network coding may influence society in ways we cannot yet imagine. In the meantime, though, those of us who are studying it are considering the obstacles to implementation. Transitioning from our router-based system to a network-coded one will actually be one of the more minor hurdles. That conversion can be handled by a gradual change rather than a sudden overhaul; some routers could just be reprogrammed, and others not built to perform coding operations would be replaced little by little.

A bigger challenge will be coping with issues beyond replacing routers with coders. For instance, mixing information is a good strategy when the receiving node will gather enough evidence to recover what it desires from the mixture. This condition is always met in multicast networks but may not be the case in general. Moreover, in some circumstances, such as when multiple multicasts are transmitted, mixing information can make it difficult or impossible for users to extract the proper output. How, then, can nodes decide which information can and cannot be mixed when multiple connections share the same network? In what ways must network coding in wireless networks differ from its use in wired ones? What are the security advantages and implications of network coding? How will people be charged for communications services when one person's data are necessarily mixed with those of other users? In collaborations that span the globe, we and others are pondering how to unravel such knots even as we strive to enhance the capabilities of the communications networks that have become such an integral part of so many lives."

-Mike

No comments: