October 7, 1999
SMARTMONEY.COM: Bright Lights, Big
Network
By TIERNAN RAY
Smartmoney.com
(This story was originally published Wednesday.)
NEW YORK -- Call it the grandma rule. Every time a hideously complex
technology is unveiled by the tech world, grandmothers get dragged into
the discussion to excuse our general inability as a species to comprehend
jargon-filled technical information.
The latest example: optical networking, a swirling mass of confusion that is
suddenly more critical to understand than ever before in the wake of Cisco
Systems' (CSCO) purchase of optical-switch maker Cerent for $7.8 billion
a month ago. I'm sure there are plenty of savvy octogenarians who surf the
Web. Probably, though, most are as confused as the rest of us about this
optical-networking business.
A week ago Tuesday, as Nortel Networks (NT) rolled out yet another
new optical-product offering, a reporter once again invoked the grandma
rule, asking Nortel's optical-networking Veep Clarence Chandran to
explain the fine details of a technology that, judging from the Q&A, had
quite clearly baffled adults of all ages listening to the call. Chandran's
response? "My father is 75, and he uses email to communicate with his
grandkids. This technology will speed up the Internet, giving him instant
access to his grandkids."
Wow. And you thought the Internet was all about sex, shopping and stock
quotes. Seriously, Chandran's comment reminds us the uses of any new
technology can be obscure at first - like trying to explain the car to a
horse-and-buggy culture. But there are a couple of good reasons why
users of the Internet and technology investors should care about optical
networking - and rapid access to the grandkids, not to be too hard-hearted
about it, is pretty far down on the list.
One of the primary benefits of optical networking, of course, will be speed,
but more important is a new level of control it will give to the phone
companies. While consumers increasingly have access to high-bandwidth
connections to the Internet in the form of cable modems, business
customers who buy high-speed phone lines, such as a T1, have to wait
days, weeks or months to get connections installed. That's not just a
bummer for business customers, it hinders the rollout of much-needed
bandwidth for Internet sites. And that makes the entire experience of
surfing the Web less than what it should be. Optical networking could help
by giving the phone companies a very flexible kind of network, which
would in turn allow them to turn on new bandwidth in seconds as needed.
If bandwidth is easy to provide, it stands to reason the uses of banwidth
could become more diverse and more interesting, leading to new
applications of the Internet.
The other point to note about optical networking is that the technology will
allow for unprecedented control over the allocation of bandwidth. You can
imagine all sorts of outcomes there. An online brokerage could, for
example, create a kind of private lane on the Internet for its best
customers, giving them faster - perhaps even real-time - connections with
which to make trades. And forget about cable modems: A phone company
could sell you not just the so-called "last mile" to your home, but also for a
small fee a level of service that is better than what your neighbor has.
If you want to understand how all this can happen - the technology of
optical networking, in other words - all you have to do is keep in mind the
simple fact that there are really two stories about optical networking. One
story is about classical optics, the science of light. The other is about what
is happening to the r outer market owned by Cisco. Thankfully, only two
buzz-acronyms need be memorized to understand each story. The first
buzz-acronym is DWDM, for dense wavelength division multiplexing. Yes,
a mouthful. Big thinker George Guilder (who, incidentally, started much of
the optical networking craze early in the decade by pushing the idea of an
all-optical network), once remarked that PCMCIA, the name given to
those little slots on the side of a laptop computer, stands for "people can't
memorize computer indus try acronyms." It is perhaps his most eloquent
argument.
The first story, the story of classical optics, is about sending light down a
strand of glass fiber, and it's a tale that's been around at least since the late
'70s, when separate developments at AT&T's (T) Bell Labs and Corning
(GLW) made it commercially viable for the first time to flick a laser on and
off in succession in order to create signals that convey information. In the
'80s, as lasers inside fiber-optic cables got buried in the ground, phone
companies started to dump massive amounts of traffic rather
indiscriminately into these fast arteries of their networks.
The explosive growth of the Internet in the '90s increased the need for
capacity in the existing fiber cabling. Starting a few years ago, DWDM was
introduced to solve that need by putting a prism in front of the laser. That
made the laser flick on and off in different colors, multiplying the signal and
thereby multiplying the bandwidth. For example, Nortel says that next year,
its optical switch, which it calls the OPTera Connect, will be able to split a
laser into 160 different colors, each of which sends bits of data at a rate of
10 billion per second, so that the entire device can transmit 1.6 trillion bits
in a second.
When you increase the capacity of a simple device in this way, you also
increase the complexity. Suddenly, a fiber-optic cable is no longer just a
dumb pipe sending lights that flick on and off. Each fiber can now be a
network of a sort unto itself, with colors communicating with one another.
The whole story of optical networking as "optics," then, is about how
DWDM created complexity and what's being done to deal with that
complexity.
What's being done is that Nortel, Corning, Lucent Technologies (LU) and
others are building all sorts of lasers and what are called waveguides to
make the various colors in fiber talk to one another. Nortel et al. will excel
in this realm based not so much on their understanding of networking, as on
the principles of physics. They must understand how to manipulate laser
signals and how to mix the signals with just the right kind of interference to
achieve communication rather than garbage.
You would imagine that as fiber-optic cabling gets more complex in this
way, something has to happen to the routers that Cisco has classically sold
to Internet service providers. That is because it is the role of the router to
decide where and how to send bits in a network, which, after all, is made
up of fiber-optic cabling.
So the second story of optical networking is how to make a better router
to take advantage of the complexity that DWDM has introduced and all
the breakthroughs that Nortel and Corning are coming up with in the
physics of optics. Here the key buzz-acronym is MPLS, for multiprotocol
label switching. Don't worry, that's it for buzz-acronyms.
As DWDM lets optical fiber transmit more data at faster and faster
speeds, routers need to get faster to keep up with those speeds. That is
why Cisco and upstart competitors such as Juniper Networks (JNPR),
which went public this past spring, now sell routers that advertise the ability
to transmit 10 billion bits of data per second, just like Nortel's optical
switches.
Private companies such as Avici Systems, of Billerica, Mass., in which
Nortel holds a 16% stake, promise to send speeds at even faster rates of
40 billion bits per second. Nortel broke off its development arrangements
with Avici, but it plans to hit that 40-billion-bits-per-second rate next y ear
with routers it owns from its acquisition of former Cisco competitor Bay
Networks.
But again, it's the control, not the speed, that's really important. What's
crucial is that the new routers made by Cisco et al. will for the first time
ever be controlling the way that the underlying fiber-optic cabling sends
bits. Traditionally, a router processes packets of data as electrical signals
and then dumps those packets into the fiber optic cabling beneath rather
indiscriminately. Routers hover etherea lly above DWDM on a separate
plane, in a sense, the two never communicating directly.
That must change. With ever more complex fiber optics thanks to DWDM
and faster routers that can keep up with that complexity, an Internet router
ultimately needs to tell a fiber optic cable which of the many colors or
wavelengths should send a bit. If a piece of DWDM equipment has a
failure with one of its wavel engths, for example, the router should know
about that, and it should help to decide which wavelengths ge t the bits
instead. MPLS, a method for routers to communicate with one another and
with other equipment, may well be the technology that gives the router that
kind of control.
As you can imagine, at some point both stories of optics come together.
The routers get faster and learn more about the underlying fiber optics, and
eventually, maybe, the two pieces of equipment are fused into one piece of
equipment. That would allow the Internet service providers, Internet
content providers and e-commerce sit es to deliver tiered services based
on speed with pinpoint accuracy. It gives the phrase "being bumped to the
head of the line" a whole new dimension.
Nortel certainly hopes this kind of technology convergence is the future.
This week it announced its big, fast router, called the OPTera Packet
Core, which can talk to those underlying DWDM machines. But the
Packet Core isn't even in alpha testing with customers. We'll probably see
it rolled out some time early next year, says Nortel. (Nortel says the
Packet Core and the Optera Connect will be presided over by a "common
control" system that uses MPLS, kind of like the Freudian superego.)
Meantime, skeptics insist that the two stories may not come together for
some time. That's because if you buy a single piece of equipment from
Nortel with routing and optics, you might get decent DWDM equipment,
but the MPLS capabilities of the routing wouldn't be as good as, say, what
Cisco could offer. The reverse is true of Cisco's knowledge of optical
physics, some say, the Cerent purchase notwithstanding.
In the meantime, both companies will probably sell to their strengths, where
they are best able to explain to their customers at the phone companies -
and their customers' grandmothers - what all this stuff is about.
For more information and analysis of companies and mutual funds, visit
SmartMoney.com at smartmoney.com
|
