here is the second half:
SEEING THE LIGHT
Photonic switching may be a long way off or
just around the corner--but it's important,
and it's coming.
By Luc Hatlestad
Paradigm shift. Philosopher Thomas Kuhn's coinage has
become the most overused, banal phrase in the high-tech
industry--and very nearly meaningless. Yet it persists, largely
through the favor of public relations professionals, who promise
in barrages of press releases that this new widget is the one that
will change the face of technology forever. But sometimes
paradigms do shift, and on rare occasions it really is a single
product or technology that brings about the fundamental change.
Technologists and engineers devote their lives to creating these
insanely great things. This quest is the high-tech industry's
lifeblood. And it's this quest that helped create photonic
switching, which really may be a paradigm shift.
The need for speed
It must be noted here that several other paradigm shifts will
likely occur before photonic switching, sometimes known as
optical switching, has its day in the sun. For the moment there
simply are too many questions about its price and performance,
and the networking industry's mainstream researchers are still
tied up with Asynchronous Transfer Mode and Gigabit Ethernet.
But earlier this year, when we asked a group of executives from
small and midsize networking companies what they most feared
(see "Switching to IP"), they told us that, like everyone else in
this business, they feared a revolutionary technology--a
paradigm shift--that might put them out of business. For the
networking industry, photonic switching could create this sort of
upheaval.
It will be at least six months, and perhaps even a few years,
before the first commercial applications of photonic switching
are available. The technology has yet to come up on the radar of
most industry switching analysts. Although large companies like
International Business Machines and Texas Instruments are
committed to developing their own variations of photonic
switching, The Red Herring discovered that the university
research lab still is the best place to look for the technology.
Photonic switching involves storing data on photons--packets of
light--and using the photons to transmit, store, and manipulate
information. In its ideal form, photonic switching will greatly
outperform the current electrical methods. "To this point,
switching has been deployed through purely mechanical motion,
but the disadvantages of this method are its comparative
slowness and the unreliability of its moving parts," says Joseph
Goodman, a professor at Stanford University and a cofounder
of the photonic switching company Optivision
(www.optivision.com).
The idea that there's a speed problem in switching might be
news to those developing ATM and Gigabit Ethernet. These
technologies have only recently been deployed on network
backbones, reaching unheard-of speeds, and since most
corporate desktop users are content with a T1 line, why the rush
to light speed?
The pat answer is the Internet bandwidth bottleneck. As more
people discover the Web and begin using it for games,
videoconferencing, and other bandwidth-hungry applications,
service providers are finding it more difficult to alleviate traffic
problems and keep customers happy. Simon Fok is CEO of
GigaLabs, a switching company that recently announced a
Gigabit Ethernet switch with a 16-Gbps backplane. With speeds
like this in an affordable switch, it's hard to see how the industry
could so desperately need more, but even Mr. Fok is aware that
it does. "Photonics initially has been a niche product, but we're
getting to a point where it's probably the only way to get the
speed you need," he says. "The applications for it are few, but
they're important."
Darkness before the dawn
What's currently keeping photonic switching from escaping its
niche are issues surrounding price and performance. Although
the technology can produce spectacular results in a vacuum, in
the real world it's still contending with the problem of data loss.
"Loss is always a very critical parameter in switching," Mr.
Goodman says. "Early switching technology had very high
losses, and its applications were limited to transmission over
very short distances."
There are myriad ways to pursue optical switching with
acceptable loss. One method involves pumping lasers through a
tunnel, with the light forcing open a shutter at the other end. And
a pet project of IBM's called wave division multiplexing (WDM)
takes data from several sources and switches it arbitrarily to
various destinations. Its secret is that each source has a laser
with a slightly different frequency, and each receiver at the other
end is tunable to that frequency.
The drawback of these approaches, and all others within
photonic switching's realm, is cost. "These switches can cost
several thousand dollars apiece, so only the U.S. government
can afford them," Mr. Goodman says. "The problem is, the cost
won't drop until the market demand increases, and the market
demand won't increase until the cost drops."
This is the catch-22 that will snag photonic switching for the
foreseeable future. "There's a big difference between what is
possible and what is sellable," says Tom Nolle, president of
CIMI, an industry consultancy. "Even if a technology is useful, it
may not be economically feasible. The problem with terabit
trunks, for example, is that they need terabit switches, and we're
a long way from being able to pay for that."
Even evangelists like Mr. Goodman agree that the price of
photonic switching will remain its biggest question mark for some
time. "Although increasing the usable bandwidth of optical fiber
will bring down overall costs with time, the arrival of optical
switching won't be as early as has been predicted because the
cost of electronic switching is also dropping," he says.
Calling on the telcos
Given that massively monied organizations like the U.S.
government are a limited market, Optivision has turned its
attention to the next-most-logical consumers: the telcos.
Although the company's government contracts require that its
officials zealously guard information about ongoing research
endeavors, Mr. Goodman did say that Optivision's expansion
toward the consumer arena would begin with a partnership with
"a major telco." (He refused to identify the telco in question, but
Optivision observers say the company has a close working
relationship with American Telephone & Telegraph.)
Although cryptic in his description, Mr. Goodman says the target
application for the project will enable the telco to switch all the
channels in one fiber to all the channels in another fiber, a
process known as multiplexing. "We think this is where the first
commercial application will occur," Mr. Goodman says. "It's
probably needed right now, but it won't be deployed for a year
or two because development is ongoing."
He says it will be midyear or possibly 1998 before privately
owned Optivision makes public its photonic development plans.
"We've found a role that not everyone is aware of yet," Mr.
Goodman says. "We've kept quiet because we worry about the
ability of a small company to compete with a Lucent if the word
gets out. But on the other hand, a good market to us may not be
a good market to them."
Although details remain sketchy, observers say the telcos may
be just the right target. They add that it's incumbent upon the
telcos to take advantage of the technology before ISPs or some
other industry does. "The technology is tied far more closely to
the telcos and the cable companies than to the electronic
switching vendors," says Jayshree Ullal, a vice president of the
workgroup business unit at Cisco Systems. "But if telcos don't
jump on optical switching as a service and it takes off, it then
becomes a way for corporations not to rely on leased-line
providers for their connections."
Patience, patience
But given the relative infancy of ATM and Gigabit Ethernet and
the lack of urgency felt by average corporate users to upgrade
their network connections, Ms. Ullal still thinks photonic
switching's day is a ways off. "Every year recently has been
heralded as the year of optical fiber, so this year is no different,"
she says. "But switching beyond gigabit speeds is still undefined.
Most desktops are still at 10-Mbps speeds, and it probably will
be five years before we see massive amounts of gigabit switching
out to the desktop, so ideas of photonic switching are still chiefly
ideas at this point."
Photonic switching may be appropriate for more than linking up
the Internet. Ms. Ullal says businesses like banks could use fiber
to link up their branches, or the wires could be used as a
high-speed (and, no doubt, high-premium) connection between
banks and telcos or ISPs. Mr. Nolle sees a possibility of
embedding photonic functionality into silicon to alleviate the
heat-emission problems currently faced by silicon-chip makers
that are migrating to smaller manufacturing processes. If chips
can be developed to send light signals instead of electronic
impulses, the need for transistors, and thus the heat problems,
will be eliminated.
Mr. Goodman and his cohorts don't deny the potential of
photonics in these areas. In fact, Optivision does research in at
least four variations of the technology. But for photonics' big
splash, the company is still betting on the WAN switching arena.
The advantages are photonics' speed, its potential accuracy, and
the fact that because it switches light particles, it works with any
pre‰xisting technology. "It's best applied over long-distance
transmission networks, or on networks with a large aggregation
of data, or in local networks that connect very high-performance
computers," Mr. Goodman says. "It's primarily focused on
gigabit speeds and beyond, but at least in the case of WDM, its
time is not some undetermined point in the future. It's being
deployed now."
With ATM and Gigabit Ethernet still finding (or escaping from)
their niches, there may be a paradigm shift or two in the
networking arena itself before photonic switching becomes
widely deployed. But by now the momentum of the industry as a
whole and of the Internet in particular is undeniable. More
people than ever are hooking up and dialing in with increasingly
powerful computers and running applications that demand more
bandwidth than ever. Many solutions are out there for alleviating
what is sure to become a monumental need for bandwidth, but
it's going to be tough to top the speed of light.
Does anyone notice a lot of correlation?
|
