IE 11 is not supported. For an optimal experience visit our site on another browser.

Scientists abuzz over more efficient Web servers

An intricate honeybee dance has generated a big buzz among scientists by serving as the model for an Internet server system that adapts in response to changing user demand.
/ Source: msnbc.com contributor

An intricate honeybee dance has generated a big buzz among scientists by serving as the model for an Internet server system that adapts in response to changing user demand.

By studying the famous honeybee waggle dance that communicates the location of top-notch nectar, researchers have designed a more efficient server system that also benefits Web surfers by cutting down on frustrating delays in accessing newly popular sites. Initial tests by collaborators at the Georgia Institute of Technology and the United Kingdom’s University of Oxford showed that the bee-like way of homing in on sweet spots improved a Web-hosting company’s revenue by up to 20 percent.

“It is amazing how inspiring the natural world can be,” said Craig Tovey, co-director of Georgia Tech’s Center for Biologically Inspired Design.

Tovey’s first honeybee revelation arrived in the late ’80s when a colleague told him about a National Public Radio spot featuring bee researcher Thomas Seeley, now at Cornell University. Tovey reasoned that the bees’ unusual communication system, with the waggle dance as its focus, might help him learn more about controlling groups of robots.

Instead, he found himself collaborating with Seeley to test how a honeybee colony allocated its foragers among different nectar sources at a biological station in New York’s Adirondack State Park. “It was one of the most wonderful, exciting weeks of my life,” Tovey said of the group’s “BYOB” outing, short for “Bring Your Own Bees.” (Seeley brought 3,000.)

The waggle dance
On a basic level, the honeybee’s dilemma is a tale of two flower patches. If one patch is yielding better nectar than the other, how can the hive use its workforce most efficiently to retrieve the best supply at the moment? The solution, which earned Austrian zoologist Karl von Frisch a Nobel Prize, is a communication system called the waggle dance.

“It’s basically a statement that says, ‘This is the kind of flower, and it’s in that direction and here’s how far away it is,’ ” Tovey said.

When a scout bee has discovered an attractive nectar source, it dances on the floor of the hive, waggling its body rapidly from side to side. Each element of the dance — including motion, scent and possibly even sound — gives eager followers clues to the nectar’s whereabouts.

The number of turns, for example, indicates distance, while the angle of the bee’s principal dance lines signals the proper flying direction relative to the sun. For particularly good sources, the bee may repeat the whole sequence several times so more workers can get in on the act. The moves are so telling, in fact, that researchers can learn the location of a flower patch simply by watching a dancing bee.

Tovey said his collaboration with Seeley demonstrated that the communication provides a “beautiful” feedback loop to prevent one flower patch from being abandoned while another is depleted. For a superior patch, more bees will shake it on the dance floor and recruit workers to join them. As the nectar level drops from all the hubbub, the bees take longer to fill up, delaying their repeat performances back at the hive. The drop-off in dance routines gives scouts returning from alternative sources a better chance to create their own dance fever and transfer worker allegiances. With the shifting allocations, the system continually equalizes itself and offers a steady stream of nectar.

Real-world applications
A chance to apply that lesson to other applications arrived a few years ago when Oxford University computer science graduate student Sunil Nakrani walked through Tovey’s office door with a vexing problem related to Internet servers.

Within Web-hosting businesses, the Internet servers needed to power Web sites are often divvied up according to specific clients or sites. Although the servers are optimized for normal conditions, Internet surfers are notoriously fickle. Sudden and unpredictable spikes in demand can overwhelm a site’s allotted servers as they reach capacity while others are sitting idly by, creating the possibility of lock-outs or long queues for frustrated customers — and a missed opportunity once they leave in exasperation.

“He just started to explain his problem and I immediately saw a superficial resemblance between what he was describing and honeybee communication,” Tovey said. But superficial resemblances, he said, often break down when you delve into the details. “The amazing thing was that when he was still describing the problem after 15 minutes, I was still holding the bee analogy in my head.”

Tovey realized that on a fundamental level, both the servers and the bees faced similar barriers to efficiency in an unpredictable, ever-changing environment. “So I said, ‘Sunil, let’s try imitating what the bees do.’ ”

No, the network of Internet servers did not gyrate in an electronic version of a line dance. As a stand-in for the dance floor, Tovey and his colleagues used what they called an advertisement board, which sent messages to communicate the location of hot Web sites. When one server received a user request to help out at a specific site, an internal ad popped up on the board to attract other servers in the hosting center. As in the hive, ads for locations in demand and offering better income potential lasted longer. And the longer the ads aired, the more they increased the chance that other servers would be recruited to help power the site du jour. Tovey said revenue, page hits or other parameters measuring a site’s popularity could all do nicely as online nectar substitutes.

Bee-inspired ad system
For an Internet hosting company whose income depends upon completed transactions, Tovey and his colleagues used the bee-inspired advertisement system to increase revenues by 4 percent to 20 percent, according to their study in the December issue of the online journal Bioinspiration and Biomimetics.

Jennifer Fewell, co-director of the Center for Social Dynamics and Complexity at Arizona State University in Tempe, said the waggle dance “works beautifully” as a distributed system involving rather uncomplicated individuals. As long as the bees know where to get their information — the hive’s dance floor — they don’t need much in the way of advance instructions.

When told about Tovey’s new research, Fewell said the dance struck her as a great model for solving the Internet server problem. “If you can do something with a simple rule set, that’s usually the best way to do it,” she said.

Tovey said the honeybee communication system would be suboptimal in a world that lacked variability. The return for the bees’ sacrifice of perfect efficiency is an uncanny ability to rapidly shift foraging strategies to home in on the sweet spots of the moment. Similarly with the Internet, “it wouldn’t be such a difficult problem if you knew in advance what the traffic was going to be like,” he said.

The researchers found that their honeybee analogy could be extended even further to help Internet centers save on electricity costs. On a cloudy day, for example, only a fraction of the hive’s workers forage for nectar while the rest stay in reserve. The homebodies aren’t expending much energy during their downtime, almost like a computer in sleep mode.

“We imitated that aspect of what honeybees do, and we’re trying that out on the Web center’s hosting problem,” Tovey said. So far, the method has reduced energy prices by 15 percent to 20 percent, with only a slight dip in revenue. Oxford and Georgia Tech have taken out a provisional patent on the energy-saving application, though Tovey said his team is still refining the methodology.

The bees, it seems, are showing once again that it really pays to mimic Mother Nature.