Here Come the Petabits
Bandwidth
October 11, 1999
by Robert Buderi
Dave Bishop labors in a Lilliputian land. Wander
the halls around his office at Bell Labs, Lucent?s
(LU) research unit, and you?ll spot the evidence in
blown-up pictures of his group?s creations:
miniature trampolines, microphones and mirrors,
each containing hundreds of moving parts, yet small
enough to fit on a pinhead.
Bishop, who heads Bell?s
Microstructure Physics
Research Department,
oversees the white-hot area
called MEMS, or
microelectromechanical
systems. His powerhouse
group--now more than 20
strong--is hot in pursuit of
the optical switch, a means
of keeping fiber-optic
signals humming along without the need to convert
photons back to electronic form until they reach their
final destinations.
If it pans out, such a switch could improve
transmission capacities by 1,000 times over what?s
typical today. And forget about the traffic
bottlenecks that cause conventional circuits to
overload. Bishop sees the optical switch enabling a
"data mesh": a vast adaptive network offering ways
around virtually any problem.
With huge spoils awaiting the first out of the
blocks--in gaining markets and in setting
standards--the race has drawn a host of entrants,
with almost as many technical approaches as
contestants. However, Bishop believes that Lucent
has the inside track. "Our horse in this race," he
proclaims, "is MEMS."
Bell Labs? run for the optical switch underscores an
important research lesson: Focusing on the right
areas is often more important than hitting a specific
target. Until about 18 months ago, Bishop?s top
priority was to bring the power of fiber to the home.
Given optical fiber?s near-infinite bandwidth, such a
plan promised to trump the efforts of broadband
cable outfits if it proved cost-competitive.
MEMS was the key, obviating the need for
expensive, power-hungry laser transmitters in each
house. For a fraction of the price, movable
micromechanical mirrors would instead convey
signals through patterns of reflected laser light
beamed from neighborhood nodes, simulating lasers
in every home. But Lucent misjudged the
convenience of piggybacking transmissions on cable
TV lines. When AT&T?once counted on to blaze
these fiber trails?began buying cable companies,
notes Bishop, "They very well may have decided the
issue."
All of which might have spelled trouble if Bell Labs
hadn?t already begun shifting its MEMS horse to the
optical switch race, where the payoff was looking
even bigger. Fiber lines typically come together at
major switching stations?dubbed NFL cities
because they correspond roughly with
football-franchise hubs?where signals are rerouted
toward their final destinations. For the electronic
switches performing this task to function, however,
optical signals must be converted to everyday bits.
They?re then assigned a new fiber and reconverted
to optical ones and zeroes for the next phase of their
journey. The big problem: Traffic volume is
growing far faster than the capabilities of electronic
switches to handle it, vastly limiting data flow and
leading to potential bottlenecks.
An optical switch could avoid most, if not all, of the
intermediate conversions from photons to bits and
back, conceivably freeing networks to handle as
much as a petabit of data per second?1,000 times
today?s terabit standard. If each phone call costs 10
cents a minute, Bishop says, "that amount of data is
worth about a billion dollars a minute." Hence the
horse race.
Bishop says MEMS is better than other technical
approaches being pursued because it is "cheap, fast,
small and robust." Computer-controlled
micromirrors could route photons to their
destinations simply by tilting on their axes.
Moreover, MEMS switches can be made almost
exactly like an integrated circuit using inexpensive
"last generation" equipment, potentially lowering the
cost to pennies apiece. "This is a disruptive
technology," he enthuses. |
