For those of us who used to dial up to get online, the day broadband
arrived will always be remembered fondly. Widely available high-speed
connections not only made surfing the Web a faster, more pleasant
experience, it ushered in a whole new era of Web content and
applications. Suddenly you could have videoconferencing, voice over
internet protocol (VOIP), music file-sharing, and video blogs.
So imagine what the Web would be like with connections that are 10,000 times faster than those we have today.
"It will totally transform our society," says Connie Chang-Hasnain,
the John R. Whinnery Chair Professor of EECS professor at UC Berkeley.
"Suddenly everybody is a publisher and everybody can receive movies all
the time. Three-dimensional video conferencing could totally transform
many ways we do business, medicine, and education."
What will make such things possible is an optical network that is radically different from the one we have today.
Telecommunications information is carried from point to point over
optical fibers, which have a very high capacity. "One fiber is enough
to carry the entire world's Internet traffic," explains Ming Wu, an
EECS professor at UC Berkeley. "However, the network traffic is very
diverse. Many different kinds of data are being transmitted'some voice,
some text'and each of them is going to a different destination," he
"The Internet was designed nearly four decades ago to support any
applications on any physical layer platforms," says S. J. Ben Yoo, an
ECE Professor and CITRIS Campus Director at UC Davis. "The Internet
traffic continues to grow explosively today. While the modern
multi-wavelength optical networking can now transport such traffic, it
remains an extremely challenging task to route and switch such an
immense amount of traffic to support demanding new Internet
applications," he says.
Routing that data creates bottlenecks, because it must be converted
from light to electricity and back to light before it can be sent to
the next point along the network. It is a bit like having to change
planes several times to get to your destination. Chang-Hasnain, Wu and
Yoo are three CITRIS researchers who are developing technology that
will eliminate those bottlenecks.
"We need a smarter optical network," says Wu. To that end, he is
developing reconfigurable, or tunable, optical Micro Electronic
Mechanical Systems (MEMS). Data passing through the network in the form
of light would be organized by type and destination into different
colors along the spectrum. It is a rough analogy, but think: red for
"traffic" going to New York blue for Chicago green for Los Angeles. The
MEMS, known as a wavelength selective switch, would detect the color,
and a tiny movable mirror on the device would steer it accordingly.
Currently, changes to the optical network'such as adding a node or
redirecting "traffic" to accommodate heavy bandwidth use or avoid a
failed node'must be made manually. "These new tunable devices will
allow the entire process to be controlled electronically or even
spontaneously. When a node senses more traffic is coming this way, it
will allocate more bandwidth to that channel," says Wu.
Other benefits will include networks automatically sensing the
presence of new nodes'no need to log on, as you will be automatically
connected'and instant access to larger bandwidth whenever you need it.
"Much of our research is focused on how to shrink the size of those
smart functions and implement it on a piece of silicon," says Wu.
Prof. Yoo has recently demonstrated a new type of all-optical router
that can switch data at a fraction of a "nano" second, or a billionth
of a second. "Today's routers store packets, segment them into cells,
switch the cells, and reassemble them into packets again. Our new
optical router can pipeline and switch optical packets without storing
them simply by changing the lanes or colors of the packets," says Yoo.
The resulting optical router has been shown to scale to 42 Petabit per
second capacity, or millions of times the switching capacity of today's
large routers found in typical buildings. However, when such routers
try to scale even larger, there are problems.
"Today's highest-capacity (46 terabit per second) router sold today
consumes 1.25 MegaWatts of power, weighs 56 tons, and occupies a
footprint of a tennis court. The fact that our optical router
technologies are being integrated on a semiconductor chip scales this
down to about 100 Watts of power, a half-pound weight, and the size of
a lightbulb, while providing the switching capacity of 46 times the
entire traffic in the United States today. This technology will
completely change the way we run our data centers, healthcare centers,
and entertainment businesses. It will completely transform the
Internet," says Yoo.
Chang-Hasnain hopes to speed up the optical network ironically by
slowing it down. "Traditionally an optical signal propagates at the
speed of light. It can't be slowed down. We're looking into a method of
slowing it down significantly," she says.
Just as memory buffers allow you to watch a video online without
having to download it first, or a roundabout enables cars to pass
through an intersection without stopping, Chang-Hasnain's semiconductor
optoelectronics would bend light as it passed through routers so that
it could sorted without having to be converted to electricity and back
into light again.
"We actually have made huge progress in demonstrating some of these
concepts already," she says. Most recently, her team built a device
that delays a picosecond pulse by 2.5 times its bandwidth at room
Chang-Hasnain is also working on creating nanowire lasers to replace
semiconductor lasers. By reducing the volume of light, nanowire lasers
would decrease the power needed to move it through the optical network,
and the time it takes to do so.
In January, the Defense Advanced Research Projects Agency (DARPA)
awarded researchers at UC Davis and MIT a grant for $9.5 million over
three and a half years to fund work on new high-speed devices for
optical networking. Yoo's research is aimed at designing and building
thumbnail-sized chips that can encode data at rates 10,000 times faster
than current ones. "This brings an exciting opportunity for us to
integrate functional optical signal processing units onto a single
piece of semiconductor chip. Ultrawide-band (100 THz) signal
processors, synthesizers, and arbitrary waveform generation for a next
generation optical networking can be enabled by using this optical
"When we first started back in 2001, you could count maybe a handful
of people, mostly physicists, who were working in this field," says
Chang-Hasnain. "Last year, we co-chaired with the
Optical Society of America (OSA) the
first slow and fast light conference, we anticipated maybe 50 people.
It sold out, about 120 seats, and we had to cut registration one day
before it closed."
Clearly, it is an exciting time in an exciting field, in which results are expected sooner than later.
"In about five years, our network will look very different from today's network'a lot smarter and a lot more flexible," says Wu