Bees Help Servers Grow Smarter

Ah, honeybees. Who doesn’t love honeybees? Gardeners adore them, and especially in this rough year—with mysterious cases of Colony Collapse Disorder and rumoured rogue viruses decimating their populations—more than the usual crowd of us nature lovers have felt pangs of sorrow on their behalf.

But there’s at least one happy bee story this year, and it’s got nothing to do with Jerry Seinfeld. Even as we’ve come to rely on bees for all sorts of natural, organic goodness, these fuzzy arthropods are also inspiring new ways to achieve energy efficiency. 

Computer scientists at Georgia Institute of Technology and the University of Oxford have been applying the behavior of bees in a colony to a very different type of colony—the collection of data servers that manage network resources for the internet. And the outcome, published in a paper this month, shows how a bee-inspired server configuration can save Web-hosting services 15-20 percent on their energy consumption. That’s not so shabby, especially when you consider that most improvement in energy consumption at data centers comes in the form of pricey equipment upgrades, for instance introducing new power supplies and DC-to-DC converters that are more than 90 percent efficient. (True, true, they pay themselves back through savings on energy bills—but sometimes that argument doesn’t go far in boardrooms.) And in data centers where servers can be clustered by the hundreds, that energy savings really adds up. For every watt spent to power a server in a typical industry data center, another watt is expended on cooling systems for the facility. So making a server more efficient also means that lots of energy can be saved in running the buildings that house them.

That’s all well and good, but let’s get back to what this has to do with bees. A beehive has a limited number of workers to fetch nectar from flowers and return it to a hive, and likewise there are only so many servers that can respond to page requests from Web browsers. Bees conserve the energy they all use scouting for nectar by communicating to each other through their famous waggle dances. When a bee finds a nectar-filled flower patch, it returns to the hive and performs a dance in front of its fellow bees, where each wag and rotation indicates something about the location of that patch. Other workers interpret those wags and fly off to collect the nectar.

With a Web server, predicting demand for any given Web sites at different points in time can be difficult, if not impossible. The servers are optimized for average conditions, but that means they’re poorly configured to handle greater demand when, say, a YouTube video suddenly becomes all the rage. To address that, these two computer scientists tried to translate the bee behavior into a similar code for servers. In a nutshell, a bee-inspired server under high demand will report—aka waggle-dance—the conditions of its duress to the other servers on its network in an attempt to draw idle servers to serving that site. For more details, see here and here.