Animals Wildlife How Honeybees Make the Internet Work By Christine Lepisto Writer St. Olaf College University of Minnesota Christine Lepisto is a chemist and writer from Berlin. A former Treehugger staff writer, she now runs a chemical safety consulting business. our editorial process Christine Lepisto Updated October 11, 2018 Promo image. Yumiko Sakurai Share Twitter Pinterest Email Animals Wildlife Pets Animal Rights Endangered Species Computer engineers study the mathematics of how to optimize complex systems. In one example, they face a logistics challenge known as the "travelling salesman problem:" how can a hypothetical salesperson visit every city on their route in the shortest distance? The algorithms developed to answer these sorts of questions are useful in many situations, such as reducing the costs of and pollution from a fleet of delivery trucks. But when engineers tried to optimize traffic on the internet, they found their methods wanting. Demand quickly rises and falls - for example, an oncoming hurricane drives traffic to a weather website, or a sports team's pageviews peak when there is a big play at a game - so the resources cannot be allocated systematically but must be continuously reorganized in response to a changing situation. Honeybees don't study mathematics, but the demands of evolution reward those colonies that succeed in optimizing their resources. Fortunately, in the strange tale of how honeybees make the internet work. the scientists were smart enough to see that the honeybees knew better than they did. Can systems engineers offer consulting services to the honeybees? It all started when systems engineer John Hagood Vande Vate heard a story on NPR about honeybees. Cornell honey bee researcher Tom Seeley described how foraging honeybees returning with nectar can guess whether the harvest is plentiful by how long it takes them to find a hive bee available to take the nectar into storage. If the hive bees are scarce, the foraging bees will preserve their energies by being picky about harvesting in the easiest places. But if the hive bees are needing more nectar, a bee that has succeeded in finding a good source of nectar will perform a lively "waggle dance" to get others to follow to their treasure trove. Over lunch that day, the system engineer shared the tale with his colleagues John J. Bartholdi III and Craig A. Toveyat at Georgia Tech, and they wondered together if they could use their knowledge to make the bees even more successful. If only the bees could hire them! A collaboration was born. Using funding designed to support basic research with no foreseeable applications, the Georgia tech systems engineers teamed up with the Cornell bee guys, and they came up with a mathematical model that described how the bees distributed themselves among the resources - patches of flowers that varied based on the time of day, weather and seasons. Strangely though, the model describing the bees' foraging was not "optimal" - a term that is defined very specifically in the context of systems engineering. But further study indicated that the bees' model led to highly efficient collection of nectar across a wide range of conditions. The Georgia Tech team realized they were on to something: "the Honeybee algorithm" could beat out the traditional mathematical solutions. It would be some years more before the scientists would have proof that the honeybees' behavior actually performs more profitably than the optimization algorithms in cases where the conditions are highly variable. The "Honeybee algorithm" works on the internet At this point the research hit a dead end. Attempts to apply the honeybee algorithm to various situations such as explaining how ant colonies organize or optimizing highway traffic didn't quite fit. A fortuitous meeting changed that. One day Sunil Nakrani walked into Tovey's office, looking for some mentoring on a systems engineering problem related to web hosting and variable internet traffic. Nakrani didn't know about Tovey's excursions into honeybee research, but Tovey saw very quickly that the problem Nakrani described was "just like the honey bee forager allocation problem!” It turns out that shared web hosting servers can only run one application at a time (for security reasons) and each time a server switches applications, time (and money) is lost. The best server allocation algorithm must allocate resources to optimize profit even as the sources of traffic (=revenue) can become highly unpredictable. When Nakrani defended his dissertation on an algorithm in which the servers do their own "waggle dance" to communicate that they are involved in a profitable client, he was surprised that instead of questions about his methods and conclusions, he faced the panels' question, "Have you patented this?" In defense of bio-mimicry and of basic scientific research At this year's annual meeting of the American Association for the Advancement of Science in Austin, Texas, Tovey hopes to inspire others with his "awe and affection for nature's solutions" as he shares the story of how curiosity led to learning from honeybees how to make the $50 billion - and growing - web hosting industry work. Tovey's tale defends the need for funding that allows scientists to follow a wild hunch, or study a crazy notion, even if it seems there is little use for the knowledge at the time. And it makes a strong case for biomimicry - sometimes we can learn more by looking at the way nature solves a problem than we can by using our human logic to solve the problem ourselves. Because in the final analysis, the "honeybee algorithm" beat out the best algorithms in tests and even outperformed a hypothetical "omniscient algorithm" that could predict future traffic in advance when the conditions were highly variable - a not uncommon case on the internet. By virtue of trial and error, the bees are smarter than our best mathematicians. And luckily, Nakrani's answer to the dissertation panels' question had to be "No, we haven't patented this." Because the work was inspired by the quest for knowledge rather than for personal gain, the "honeybee algorithm" and its applications had been published and was no longer eligible for patent protection. So each and every one of us benefits from cheaper, faster web servers that work efficiently because they learned from the honeybees.