Fruit flies show how to simplify computer nets
Share with others:
And you thought the fruit fly was just a pesky little insect that likes rotten apples.
Actually the fly, or at least its nervous system, has inspired a better way to organize and operate computer networks, especially wireless sensor networks.
Ziv Bar-Joseph, a computational biologist at Carnegie Mellon University, was studying the fly when it struck him that its nervous system of hairlike structures that allow it to feel and see was not only doing what computer networks try to do but did it more simply.
A study by Dr. Bar-Joseph and five co-authors, published today in the journal Science, reveals that the fly's nervous system serves as an efficient model for organizing numerous cells to operate in unison to accomplish prescribed tasks. The group used that knowledge to write a computer algorithm, or program, to better operate computer networks.
When the fly's nervous system develops, a minimal number of cells emerge as leaders that message neighboring cells not to become leaders. In time, each cell becomes attached to a leader that helps control the entire system without knowing how many cells it controls or what the system is working to accomplish.
Computer networks use processors rather than cells. Inspired by the fruit fly, the team's algorithm or program allows a minimal number of processors in a wireless computer network to emerge as leaders to operate the system. The smaller number of leaders using simple one-bit messages makes the entire system work more simply and with less energy.
The system has numerous potential applications including searching the Web, controlling airplanes in flight, or doing environmental monitoring, where sensors are dispersed in a lake or waterway. CMU is already using the system to control swarms of tiny robots.
"It's such a simple and intuitive solution. I can't believe we did not think of this 25 years ago," study co-author Noga Alon stated in a news release. He is a mathematician and computer scientist at Tel Aviv University and the Institute for Advanced Study in Princeton, N.J.
Dr. Bar-Joseph, of CMU's Lane Center for Computational Biology in the School of Computer Science, said scientists had long used computational and mathematical models to analyze biological systems. "Here we've reversed the strategy, studying a biological system to solve a long-standing computer science problem."
When the fly's nervous system is in development, a small set of cells emerges as leaders that work to control the entire network in solving problems. The study explains how those leader cells emerge, then take control of the system with little input or knowledge.
Until now, computer networks used probability to determine which processors would serve as leaders, usually based on how many connections each had with other processors. The chance of any processor becoming a leader increases based on the number of its connections. Such systems work quickly but require lots of messaging and energy to operate.
The processors must also know in advance how they are connected in the network, creating problems in wireless networks where sensors might be distributed randomly and might not be within communication range of each other, the release stated.
By using the new system, Dr. Bar-Joseph said, sensors dropped from a plane to study volcanic activity can be organized by a minimal number of processors without requiring lots of energy or complicated messaging across the network.
As such, the fly's nervous system produced "a fast solution" that could be used in many network applications.
His research probably won't end with the fruit fly.
"Now that we understand the biology, I think we can take more and more ideas from biology and use them in computer networks," Dr. Bar-Joseph said.
First Published January 14, 2011 12:00 am