Posted: 05/1999

A New Wave in Optical Networking
By Charlotte Wolter

With Paris-based Alcatel and Murray Hill, N.J.-based Lucent Technologies Inc.
announcing optical networking equipment capable of transmitting 40 gigabits per second
(gbps), or OC-768 speed, it could appear that the future of optical networking is to
develop ever-faster lasers to crank up speeds in multiples of today’s top synchronous
optical network (SONET) standard, OC-192 (10gbps).

But some optical networking startups have a different vision. They see an optical
network based on wavelengths or "lambdas," as they often are called, with
multiple lambdas provided by dense wave-division multiplexing (DWDM) technology.

Rather than offering services at different SONET speeds to customers, each wavelength
carries a single SONET service, usually an OC-48 (2.5gbps) devoted to one customer. Each
wavelength can be switched from its origin to destination without SONET rings or overhead.
Scaling the network for new customers involves lighting new wavelengths.

Graph: DWDM Junction Architecture

Sometimes called wavelength routing or wavelength provisioning, this new approach to
optical networks is getting attention from service operators and vendors alike. The
concept of using wavelengths as the basic unit of network services represents an important
first step toward all-optical networks. It’s the Holy Grail of large-scale core networking
technology.

All-optical networks may be five years away, according to analysts’ estimates, but
technologies that begin to use wavelengths are, in some cases, available today. "It
is the concept of delivering wavelengths to the door, not a certain number of bits,"
says networking analyst Frank Dzubeck, president, Communica-tions Network Architects Inc.,
Washington. "This is the first stage of distribution of lambdas."

Two new startups, Sycamore Networks Inc., Tewksbury, Mass., and Monterey Networks Inc.,
Richardson, Texas, are betting their futures that network operators want to begin putting
data on wavelengths and switching data as lightpaths across mesh rather than ring
networks.

And Lucent Technologies Inc., Murray Hill, N.J., jumped on the wavelength bandwagon
with a new management and monitoring product, called WaveWrapper, developed by Bell Labs,
that is the first measurement and monitoring product that deals with transmissions at a
wavelength level and anticipates the advent of all-optical networking.

Sycamore introduced its first product in early March, an intelligent optical transport
system that puts multiple OC-48s on DWDM wavelengths, over an existing OC-192
infrastructure, using the same fiber, same optical amplifiers and same multiplexing. The
company’s second product, which is being tested by Williams Network, Tulsa, Okla., is an
add/drop multiplexer that allows operators to take wavelengths off the DWDM multiplex and
put others on, so wavelengths can be dropped or added without disturbing the other
traffic.

This creates "a straightforward way of moving traffic from one lambda to
another" in a fiber network, says Tom Nolle, president, CIMI Corp., Princeton, N.J.
Operators can create "inside an optical network a virtual lambda network and can mesh
endpoints together without interior nodes." If the same service were done with SONET,
Nolle says, services would have to be concentrated from OC-3s to OC-48s and OC-192s, and
would need very expensive core switches.

Monterey announced its Wavelength Router architecture in February as a way to network
intelligently the point-to-point wavelengths created by DWDM across a backbone network.
The technology links Internet protocol (IP) routers with a mesh, not ring, architecture,
and uses network intelligence to create virtual wavelength paths and restore outages. The
IP routers groom traffic into OC-48 or OC-192 streams, each stream is put on a DWDM
wavelength and the network finds the best path to deliver the entire wavelength.

Michael Zadikian, vice president of marketing, Monterey Networks, says the company’s
technology is based on the concept that most backbone traffic will be at the OC-48 or
OC-192 level. "Once traffic migrates to OC-48 or OC-192, the virtual path will
consume an entire wavelength, so what is [the] point of sorting out cells?" he says.
"You want to get that whole wavelength from San Francisco to New York."

Monterey will not deliver products until mid-1999, but Sycamore’s intelligent optical
transport already has won a contract with Williams Network, which made a $24.5 million
order for products that will be deployed into the Williams Multi-Service Broadband
Network.

Sycamore was founded by Desh Deshpande, who also is chairman; Daniel Smith, president
and CEO; and Eric Swanson, chief scientist. This is the same team that nurtured widely
admired Cascade Communications Corp. before it was acquired by Ascend Communications Inc.,
Alameda, Calif., and made the basis of that company’s core switching unit.

As with Cascade, Sycamore builds its products based on off-the-shelf components. What
the company has added to standard DWDM components is software intelligence that reduces
the amount of monitoring needed to set up the network. New wavelengths are added with
cards that tune themselves automatically to the other elements in the network and balance
optical power.

Besides the network intelligence, one of the attractions of the wavelength routing
technology that Sycamore introduced is cost. The use of wavelength-based services allows a
network to scale very cost-effectively. A new customer can be turned on just by adding
line cards at the source and destination. A service provider’s existing network, such as
an OC-192, remains in operation because the DWDM overlay uses different wavelengths than
the existing network.

Rod Alferness, chief technical officer, Optical Networking Group, Lucent Technologies,
says the concept behind the WaveWrapper technology is that wavelength-based systems have
to evolve beyond just "fat pipes." "You have to use wavelengths as a means
of managing, monitoring and controlling the network," he says, "especially
because on those optical channels we expect to carry different formats, such as ATM
(asynchronous transfer mode), SONET and IP."

WaveWrapper data includes information on the origin and destination of each channel,
the strength of the signal and whether it has passed certain points in the network. This
information can be used to provide protection so outages can be detected rapidly and
signals restored, all without using more elaborate and costly SONET protection features.
Lucent believes that wrapper technology can be a cheaper alternative to SONET protection
in wavelength-based networks.

The Lucent WaveWrapper system adds a small amount of data to the optical transmission
on each wavelength (also called "channel") to help track that transmission and
provide alerts when there are breaks in transmission. The extra bits are outside the frame
of the signal and don’t interfere with the transmission’s payload.

Lucent has proposed WaveWrapper as a standard to committees of the International
Telecommunications Union (ITU), which already recognizes wrapper technology as a standard
for undersea systems, the American National Standards Institute’s (ANSI) T1 –
Telecommunications Standards Committee, and to the Optical Internetworking Forum, a
leading industry standards body, where it is under discussion along with several other
approaches.

Perhaps more important in reducing cost is the fact that wavelength-based networks do
not use expensive SONET multiplexing to switch services. It is very expensive to separate
an OC-3 from an OC-48 or OC-192 multiplex, because the whole multiplex must be broken down
to switch out the smaller bandwidth. With wavelength-based technology, one wavelength in a
DWDM multiplex can be switched and rerouted without touching the other wavelengths.

The accelerating growth of data traffic is creating demand for OC-48s from individual
customers, such as Internet service providers (ISPs), Dzubeck says. AT&T Corp.
actually delivered its first OC-48 to a customer in early March, he says, and there is a
backlog of orders. Peeling off an OC-48 from an OC-192 mux, as AT&T would have had to
do for the first customer, would be an expensive proposition, but delivering an OC-48 on
one wavelength would be much cheaper.

Also, if a SONET network must be upgraded in speed, all the nodes of the network must
be changed for the higher speed–a so-called "forklift upgrade" that can be very
costly. And the return for the investment is limited. If an operator upgrades from OC-48
to OC-192, the network has been scaled by just a factor of four, whereas even the most
basic and economical DWDM system has 16 wavelengths, allowing an operator to scale at a
minimum by a factor of 16.

Thirty-two wavelength systems are commonplace and systems with 80 or more wavelengths
are beginning to appear.

Assuming there is this demand for OC-48s, a network of OC-48 wavelengths can be much
less complex than a SONET network with comparable bandwidth, and it is significantly
cheaper to scale for new customers.

Neither of these companies yet can switch wavelengths entirely in the optical domain, a
technology analysts say could be as much as five years in the future. Transmissions use
SONET speeds and must be brought down to an electrical signal for an add/drop multiplexer
to switch to another wavelength or to route the service. "It is true that in the
interim we have to stay with the SONET numbers because there is no way to do photonic
fiber-to-fiber connections," Nolle says.

Nolle adds that, despite the interest in what Sycamore has done so far, "the point
is the endgame. What [the industry] is really trying to do is eliminate electronic
handling in the care. That is the long-term objective, and it is doable in the long
run." Nolle estimates that technology for true all-optical switching won’t be
available until 2005.

Charlotte Wolter is infrastructure editor for PHONE+ magazine.