Two researchers may have found a way to greatly reduce traffic bottlenecks that could take place in ad hoc networks. Such work may be essential for the future development of sensor networks, they say.
Keren Censor-Hillel, a postdoctoral research student at the Massachusetts Institute of Technology's Computer Science and Artificial Intelligence Laboratory, and Hadas Shachnai, an associate professor of computer science at the Technion Israel Institute of Technology, presented their work last week at the 2011 Symposium on Discrete Algorithms, held by the Association of Computing Machinery and the Society for Industrial and Applied Mathematics.
Ad hoc-styled networks are predicted to be used widely in the decades to come. As the cost and power requirements of processors continue to decline, they can be used in large numbers in low-power sensors that could monitor anything from potential volcano activity to highway traffic.
By their very nature, ad hoc networks are not managed by any one controlling device, such as a router. Instead, each end node acts as a router, forwarding any data it gets or generates to its nearest neighbors, and eventually the data will make its way back to a collection point for the network.
Typical designs for ad hoc networks have each end node randomly select another node to pass its data onto. This approach is designed to ensure that traffic is distributed evenly across all the nodes. If one node goes down, another can take its place. The problem with this approach, however, is that bottlenecks could be created when only a few end nodes are tasked to convey all the traffic.
Spreading it around
The algorithm Censor-Hillel and Shachnai developed spreads out the traffic in a way that prevents such bottlenecks. In their approach, a node selects another node at random every other time data needs to be conveyed. In the alternate rounds, the node sends traffic not at random but to a node it has not communicated with recently.
Alessandro Panconesi, a professor of computer science at Sapienza University of Rome and an expert on network analysis, concluded that the algorithm is "an interesting contribution."
"Essentially, a node in this network can wake up and start operating by using this algorithm, and if every node in the network does the same, then essentially you give communication capability to the entire network," Panconesi said.
He cautioned, however, that this algorithm in its current form is still too elaborate for simple computational devices. Because devices on ad hoc networks tend to have limited computational power and battery life, they require very simple networking protocols.
"The algorithm is very expensive in terms of the information that it needs to exchange," he said, adding that a simplification of this algorithm would be feasible.