F R I D A Y , F E B R U A R Y 2 4 , 2 0 0 6 A g e n d a
Distributed Decision-Making in the Biomimicry, Private, and Public Sectors
Dr. SUNIL NAKRANI
Computing Laboratory, University of Oxford, UK
Dr. CRAIG A. TOVEY
Professor, Georgia Tech Center for Biologically Inspired Design (CBID)
Dr. JEANNETTE YEN
Professor, Director of Georgia Tech Center for Biologically Inspired Design (CBID)
Internet server colonies co-host web services on servers that are leased by its customers on a pay-per-use Service Level Agreement (SLA). A hosting center thus must allocate its servers among host customers to maximize revenue. Optimizing the allocation is challenging because of the limited number of servers, costs of migrating from one service to another, and unpredictability of future HTTP request loads.
The hosting center allocation problem is remarkably similar to a problem in the natural world. Honey bee colonies must gather enough nectar during the summer to survive the winter. The colony must allocate its forager bees among flower patches to maximise nectar influx. We expose many close parallels between server allocation between the problems. Based on this homology, we develop a honey bee imitative algorithm which dynamically allocates servers to satisfy the unpredictable HTTP request loads.
We use HTTP trace data from a commercial service provider and synthetically generated HTTP requests to evaluate algorithm performance. The experimental results show that our algorithm performs very well, indeed better than a conventional algorithm and an omniscient static algorithm, for highly variable request loads. In contrast, it is outperformed for some low variability access patterns. This suggests that real honey bee colony forager allocation, which is suboptimal for static food sources, possesses a counterbalancing responsiveness to food source variability. Additional experimental tests find that the honey bee algorithm is more resilient than the conventional.
Honey bee foraging also has a similarity to highway traffic. When commuters separately choose which route to take, they equalize travel time much as forager bees separately choose flower patches to equalize collection time. The resulting distribution of vehicles on the highways follows the same pattern as the distribution of honey bees among flower patches. By analogy with the biological system, this suggests some conjectures regarding feedback loops, information, and traffic dynamics.
More on Georgia Tech's Center for Biologically Inspired Design:
http://www.gatech.edu/news-room/release.php?id=684
The Economist on the topic and our speakers:
http://www.hawaiibeekeepers.org/news-bees-internet.php
"HONEYBEES want to make honey, while internet host providers want to make money. And the rhyme is not the only thing to link honeybees and internet host providers, who operate vast warehouses full of server computers on behalf of their customers. Sunil Nakrani of Oxford University and Craig Tovey of the Georgia Institute of Technology, in Atlanta, have recently developed what they believe is an efficient way of organising internet servers, by mimicking the behaviour of honeybee colonies."
B I O M I M C R Y is a new science that studies nature's models and then imitates or takes inspiration from these designs and processes to solve human problems. Biomimicry uses an ecological standard to judge the "rightness" of our innovations. After 3.8 billion years of evolution, nature has learned: what works, what is appropriate, what lasts. Biomimicry is a new way of viewing and valuing nature. It introduces an era based not on what we can extract from the natural world, but on what we can learn from it.