Bumblebee boogie analysis in webserver boost
Drone dancefloor plan is the proverbial insect limb joint
Researchers in Georgia, USA, reckon they can increase the efficiency of web servers by mimicking the methods used by honey bees to collect nectar.
It seems that the new hive-mind technology was the brainchild of Professor Craig Tovey of the Georgia Institute of Technology (GIT).
“I studied bees for years, waiting for the right application,” Tovey said. “You have to look for a close analogy between two systems — never a superficial one... this definitely fit the bill.”
By lucky hap, Tovey's years of avid bee-watching paid off during a discussion with Sunil Nakrani of the Oxford computer-science department. Tovey and Nakrani realised that the task of serving web pages was uncannily similar to that of sucking sugar out of flowers.
More specifically, it was the bees' internal organisation that held useful lessons for servers. Bees have no central command or leadership (the "queen" is mainly concerned with laying eggs rather than directing the nectar harvest). Rather, the cunning insect workers pass information to each other by doing complicated little dances in the hive.
Bees coming back from a rich patch of nectar let their buddies know about it, and more bees pile in to the relevant flowers until all the sticky goodness is gone. The last few bees coming back tell everyone that the patch is worked out, and resources go elsewhere.
Tovey and Nakrani set to work at GIT to develop a "dancefloor" for servers, assigning resources to sites in the same way as bees deal with flowers. In their research, published in Bioinspiration and Biomimetics, they claim that "the honeybee method typically improves service by 4 percent to 25 percent in tests based on real Internet traffic."
The main question raised by all this, of course, is what happens to biomimetics researchers who study the wrong kind of animals and never develop a useful application? Presumably there are people out there furiously boning up on weasels, fruitbats, plankton etc, hoping ever more desperately as the years go by that some kind of robotics or computing analogy will present itself, only to eventually retire and die disappointed.
Perhaps we need some kind of bee-style way of assigning biomimetics researchers more efficiently. ®
Sponsored: IBM FlashSystem V9000 product guide