A STRANGE breed of bee is buzzing around a lab at the Xerox Research Centre
in Grenoble, France. Unlike its hive-dwelling brethren, this creature does not
visit flowers or make honey, instead it sits at a desk and works on a computer,
wanders around the lab to mingle with its co-workers, and occasionally flies off
to conferences on the other side of the world.
This new creature might seem rather alien to the average honey bee, but they
have one important thing in common: both move pollen around. While real bees
move organic specks that fertilise flowers, the new ones鈥攁ctually human
beings鈥攃arry packets of digital information.
鈥淭hink of people as bees,鈥� says Dave Snowdon, the researcher who developed
the idea. 鈥淲e move around from place to place, so why not use this movement to
transfer information around the world just as bees move pollen from flower to
flower?鈥� He calls this concept 鈥減ollen networking鈥� and says it could be a
valuable new approach to distributing digital information.
Advertisement
It鈥檚 not the first time technology has taken its cue from nature. The
aerodynamics of fruit flies helped researchers at the University of California
at Berkeley to design 鈥渞obofly鈥濃�攁 minute flying camera. Xerox researchers
in Palo Alto borrowed a trick from epidemiology to update their databases by
mimicking the way a virus spreads. Foraging theories from biology and
anthropology inspired another team to improve Web searches
(快猫短视频, 11 November 2000, p 38).
Even Velcro was inspired by the way burrs attached themselves to the trousers
of its inventor, Swiss engineer George de Mestral.
鈥淧ollen鈥� is still at the conceptual stage and Xerox doesn鈥檛 think of it in
commercial terms, but Snowdon predicts that it will one day be put into practice
to make cheap and simple data networks. All you need is a swarm of people
carrying personal digital assistants and a collection of fixed nodes to act as
鈥渇lowers鈥�. Whenever someone brushes against a node, their PDA picks up and
deposits grains of digital pollen. In this way messages get dispersed around the
network, riding piggyback on everyday human interaction.
Pollen began with the observation that places such as offices, meeting rooms
and reception areas are full of devices you might want information about:
printers, fax machines, computers, photocopiers, filing cabinets, vending
machines, bookshelves, even the rooms themselves. 鈥淢ost people carry Palm Pilots
or Psion organisers these days, so we thought their screens could be used to
view information on the devices they visit,鈥� Snowdon says. For example, they
could read the fault diagnosis for a broken printer, check a room鈥檚 鈥渕emory鈥� to
see if it was free for a meeting, or examine the contents of a bookshelf. Even
books themselves could carry reviews and a list of previous readers.
Next came the idea of using these office 鈥渘odes鈥� as electronic notice boards
where people could pick up or post messages for each other. From there it was a
short step to the idea of piggybacking data around on people, using their PDAs
as pouches for grains of digital pollen.
To test the idea, Snowdon鈥檚 team built a prototype network in their Grenoble
office. To keep things as cheap and simple as possible, they used coin-sized
computers called iButtons built by the Texas-based company Dallas Semiconductor
as the fixed nodes. An iButton contains up to 64 kilobytes of memory, can run
short Java programs, encrypt and decrypt messages using public key cryptography,
and has an infrared communications channel that is opened simply by touching it
to a compatible gadget.
The team stuck iButtons all over doorways, filing cabinets, printers and
computers in their office to create a network of nodes. To put a message out
onto the network, members first had to deposit it onto a node with a touch of
their PDA. Then the message鈥檚 journey began. The next person to visit that node
would pick up a copy of the message. Every node they visited thereafter got a
copy dumped onto it, ready for other users to pick up and drop elsewhere. If
someone picked up a message targeted at them, it would immediately appear on
their PDA screen. Otherwise they carried it around silently, just as a bee is
oblivious to the pollen stuck to its body.
In this way the grains of digital pollen rode around from node to node. And
when people moved around frequently, the messages leapfrogged around the network
with surprising speed. However, there is a problem with this sort of unchecked
distribution. In a relatively short time a single message can duplicate itself
exponentially and flood the network. To solve this problem Snowdon and his
colleagues introduced two measures.
First, they added an organisational memory they called a hive. When a user
docked their PDA into its cradle, perhaps to download information from their
desktop computer, the pollen grains on it were uploaded to the hive鈥攁
central server that keeps track of the status of messages. If the hive learned
that a message had reached its destination, it would mark it as read and notify
other PDAs that the message should be deleted.
The hive can help in other ways too. It can learn a certain amount about the
routes its bees take and use this information to optimise the flow of messages,
by sending messages via PDAs that are more likely to reach the right
destination.
As the second measure, the team decided to give each message a fixed
lifespan, expressed in terms of the number of nodes it could visit, or the
number of hops it could make, before it was automatically deleted. Each grain of
pollen had a counter attached to it, starting at zero. Every time it hopped from
one node to another it clicked the counter up until it reached the maximum
number鈥攂etween six and ten. Then it self-destructed. 鈥淭his controlled the
flood of messages very well,鈥� Snowdon says.
Snowdon is quick to point out that pollen won鈥檛 replace existing cable,
fibre-optic and wireless networks, but he says it will complement these in new
and useful ways.
鈥淧ollen grains could be a software upgrade, an anti-virus
update, or a bug fix for your PDA or some other electronic device,鈥� he says. You
might not yet know that your PDA is buggy, or is susceptible to a new virus. But
its manufacturer could broadcast the fixes as pollen and let them reach as many
of its devices as possible, says Snowdon. When the fix arrived it would install
itself automatically. The manufacturer would never need to know where the
devices were, or to whom they belonged. 鈥淥f course鈥�, he says, 鈥測ou could
download the fix yourself once you encountered the problem, but wouldn鈥檛 it be
better if the bug never reared its head in the first place because it had
already been surreptitiously mended in advance?鈥�
Snowdon admits that pollen networks have some drawbacks. Information travels
through them at an unpredictable rate, usually much more slowly than in
conventional digital networks, so it鈥檚 no good for urgent, time-sensitive
communication. Also, because of the decentralised nature of the network, it鈥檚
difficult to acknowledge that a message has been received. What鈥檚 worse, the
network cannot guarantee that a message will arrive, just as a flower has no
absolute guarantee it will be pollinated. So pollen is not for mission-critical
or sensitive information.
There are practical problems to iron out too. 鈥淭he technology worked, but we
decided it was simply too much hassle for users to have to touch their PDAs to
the iButtons,鈥� he says. 鈥淚n some cases the communication between the two would
take up to 30 seconds. This just seemed unacceptable.鈥�
Snowdon hopes that with the advent of short-range radio communication systems
like Bluetooth, which wirelessly connects devices within 10 metres of each
other, pollen networks will become more user-friendly. 鈥淲ith technologies like
Bluetooth, the user would become completely oblivious to the process, and that
seems to be a much more practical solution,鈥� he says. He plans to incorporate
Bluetooth into the next evolutionary stage of pollen.
Despite these drawbacks, pollen does have important advantages. First, since
it uses people to shunt data around it needs no physical infrastructure, which
means very low set-up costs. Secondly, it鈥檚 easy and cheap to expand the
network鈥攋ust add a new node. That makes it attractive as a system for
non-urgent, non-critical information, freeing up space on hard-pressed
high-speed networks. In its pre-Bluetooth form it could also be handy in
settings where wireless networks can鈥檛 be used because their signals interfere
with vital electronic equipment, such as in hospitals, aeroplanes and
spacecraft.
So where next? Mathematical modelling suggests that the system will scale up
well, as long as there are enough bees. For example, in a network with a
thousand or more nodes, you need 0.6 bees per node for a message to achieve 90
per cent coverage in 10 hops. Increase the number of bees to 1.4 per node and
you achieve the same with 7 hops.
The team also found that, in a network with equal numbers of bees and nodes,
two-thirds of the nodes would be pollinated in a time proportional to the
logarithm of the number of nodes. Snowdon says this is an encouraging result. It
means that in practice, as long as the number of nodes and PDAs is equal, the
network can grow freely with only a small impact on the time it takes for a
message to reach the majority of the nodes.
It鈥檚 still early days, but in our shrinking world of frequent fliers and
high-tech workers, pollen networks seem like an ideal, low-cost means of
communication through an ever-expanding medium鈥攗s. Soon you won鈥檛 need an
excuse to stroll over to the coffee machine or have a chat with your mates on
the other side of the office. Just tell the boss you鈥檙e as busy as a bee.