re: article: "Optics Explode" from Telecommunications Magazine
[Again, web site contains graphics, text printed below for posterity]
========from:
telecommagazine.com
The next wave of optical network equipment promises to take the technology
into the last mile
Sam Masud, Senior Technology Editor
There's a bandwidth glut; there's a bandwidth drought; there's a
bandwidth glut. Depending on when you ask the question, all of these
statements are true. Dense wavelength division multiplexing (DWDM)
is pumping up the capacity of the long-haul network while Gigabit
Ethernet is firmly ensconced in the enterprise LAN. And the LAN is
about to get a major boost: Work is underway to release the first draft
of a 10-Gbps Ethernet specification next September and to have a
completed standard by March 2002. Although the year 2005 is still
distant, 3Com predicts that more than one million 10-Gbps Ethernets
will have been shipped by then. But voice-centric SONET rings,
which are expensive to scale, dominate the metropolitan and access
networks that act as the bridge between the enterprise customer and
long-distance service provider. “There are places you could get an
OC-3 turned up in a few hours and there are places you couldn't get
it in six months,” said Mike O'Dell, UUNet's vice president and chief
scientist, expressing the frustration of many service providers. “It
annoys everyone, most of all our sales guys, because we have
installation service level agreements (SLAs) that give customers free
money if we don't hit deadlines. But you can push things only so far.”
The spiraling demand for bandwidth suggests that some service
providers will hasten to adopt optical technology to offer better, faster
services at lower prices. Already some service providers are viewing
SONET more as a protocol than a piece of networking equipment,
prompting Forrester Research to predict that by 2003 the technical
hurdles will be overcome and optical gear will replace SONET for
new equipment purchases. “ISPs and CLECs will be early adopters
of this new optical product, particularly the CLECs that are leasing
fiber because they'll want to squeeze more out it,” said Dana
Cooperson, senior analyst at RHK.
A new breed of vendors is now committed to bringing optical
networking to the last mile. Using passive optical networking
technology that permits fiber networks to be built without costly active
electronics, companies like Quantum Bridge, of North Andover,
Mass., and Sunnyvale, Calif.-based LuxN are offering last-mile
solutions. “A major distinguishing feature of a passive optical network
(PON) is that there is a passive power splitter/combiner that sits in the
outside plant, performing an optical broadcast to multiple locations,
similar to cable TV signal distribution,” said Eugene Park, co-founder
and product marketing director of LuxN. Because the upstream path
from end users returns on the same path, typically a time division
multiplexing (TDM)-based upstream method is used, [thus] limiting the
overall bandwidth links on the PON. One advantage of PON is that
depending on where the optical splitter is located, there could be
substantial savings in the amount of fiber used.”
Michael A. Viren, founder and former chairman of 2nd Century
Communications, a Tampa, Fla.-based CLEC, believes these new
equipment vendors could change the dynamics of the CLEC business.
“The prices of the new edge devices, if they are low enough, could
reconstruct the CAP business again. That's the part I get excited
about—using passive optics to really lower the cost of providing the
last mile with fiber. Today, even CAPs will lease a T1 [from the
RBOC] because it's not worth going into a building if all you can sell
are three T1s. But by using passive optics to really lower the cost of
providing the last mile with fiber, CLECs are going back to the old
CAP model, when you could make money selling a half-dozen T1s
before the RBOCs lowered their prices.”
PON players
LuxN is focused on delivering high-bandwidth services to the
customer premises. LuxN's solution consists of equipment that resides
at the central office (CO) or point of presence (POP) and a small
footprint unit located at the customer premises. “Our product portfolio
consists of what you might think of as an optical DSLAM and an
optical channel service unit. It provides a clear channel stream from
the customer premises to the CO,” said Paul Strudwick, LuxN's vice
president of marketing. The LuxN solution delivers a minimum
connection of OC-3 over a fiber pair. LuxN's ColorValve technology
allows either the carrier or the customer to incrementally upgrade
bandwidth. The company also plans to roll out a multitenant or
multi-application version of the product, with each application or
tenant served by a fiber pair or, if the economics dictates, by a
wavelength.
The Quantum Bridge solution consists of an optical access switch that
resides in the hub location and an optical terminal located at the
customer premises. The company's dynamic wavelength slicing
protocol allows the architecture to fan out each wavelength in a way
that it can serve multiple customers. Depending upon a customer's
needs, service providers can provision 1 Mbps to 100 Mbps of
bandwidth to each remote terminal. The company claims its solution
offers substantial savings when compared to providing transport
services over a TDM network or an ATM network in terms of
equipment cost per customer.
Bandwidth solutions like these are appealing to Sameer Siddiqui,
senior network architect in Sun Microsystems' MIS shop. The
company has been lab testing products from a start-up that would
allow it to replace the tens of T3 lines it uses to move traffic between
Gigabit Ethernet switches at its campus sites in the Bay Area with an
optical network. Although Sun is keeping its options open, the
product in trials would provide a managed wavelength to transparently
support any transport protocol such as Gigabit Ethernet, ATM or
Fibre Channel. Siddiqui plans to do a pilot between the company's
Menlo Park and Newark sites and later extend the solution to other
sites. “I think of this product as an optical modem,” Siddiqui said.
“We considered getting SONET service last year but it turned out to
be pretty pricey and the deployment time was way too long. Besides,
we haven't had such a great experience provisioning SONET.”
Although LuxN product literature proclaims that “The Optical
Renaissance has Begun,” such a renaissance presupposes the
deployment of fiber to or close to the customer. Estimates vary on
how close customers are to fiber today or will be in the near future.
Vertical Systems Group estimates that 76 percent of midsize
businesses are within one mile of fiber. The Federal Communications
Commission notes that between 1995 and 1998 the amount of fiber
miles deployed by CAPs grew more than 500 percent.
AT&T, having spent over $100 billion to acquire cable television
companies Tele-Communications and MediaOne, is only too mindful
of getting the fiber closer to the subscriber in order to deliver a richer
set of voice, data and video services than those currently supported
by the hybrid fiber coax network. AT&T's LightWire system, which is
currently being piloted in the Salt Lake City area, extends the fiber
closer to subscribers. According to AT&T, it supports all current
services (video, data, telephony) and provides an envolution path to
very aggressive bandwidth. LightWire, the company claims, expands
the total bandwidth of the coax plant to 1 GHz, triples the upstream
bandwidth and, over time, will increase upstream bandwidth by a
factor of 10. Additionally, it improves network reliability, reduces the
number of active components in the network by 60
percent—consequently cutting power consumption in half—and
supports all industry standards while future proofing the network for
emerging service demand.
Minneapolis-based Optical Solutions already has a product in service
for delivering data, voice and cable TV over fiber called FiberPath
The fiber-to-the-home solution requires no electronics between the
headend and the subscriber, who might be five miles or more from the
headend.
From a splice point in the field, the company's solution requires a fiber
pair to each home, with each splice point serving up to 32 homes.
FiberPath supports up to six phone lines, both NTSC (135 channels)
and digital CATV, and data services that the service provider could
upgrade in 64-kbps increments to a maximum of 10 Mbps. The
system could also support meter reading and security services. Under
FiberPath, all services are accessed by the subscriber from a network
interface device that Optical Solutions calls a universal demarcation
point (UDP). “The UDP, which is AC-powered with battery backup,
resides on the side of a home. It takes in the fiber signals and transfers
them through the appropriate home wiring,” said Bob Lund, chief
technology officer for Optical Solutions. “We have plans to deliver
OC-3 service through the same infrastructure without touching a
thing,” Lund said.
The FiberPath solution is already being used by All West
Communications of Kamas, Utah; FutureWay Communications of
Toronto; and Rye Telephone of Colorado City, Colo. Rye Telephone
has purchased FiberPath to serve a 4000-home development with
each home connected by a six-fiber drop. “The decision is pretty
straightforward when you compare networks,” said Jeff Starcer, chief
engineer and plant manager for Rye Telephone. “A totally passive
optical network to the home will carry our company and customers
well into the future.”
Metro strategies
While these players are focused on a last-mile solution, other optical
equipment vendors such as Sycamore Networks, Kestrel Solutions
and Chromatis Networks are seeing opportunities in the metropolitan
network. Millennium Optical Networks, a Manhattan-based CLEC,
has deployed Sycamore's SN 8000, a family of intelligent optical
add/drop multiplexers that adapt, switch and route multiprotocol
wavelengths so that a user can come into the network at any point and
exit at any other without SONET termination gear. Millennium has
deployed the SN 8000 in carrier colocation facilities in New York
City to provide high-bandwidth carrier-to-carrier services on an
as-needed basis. “We're starting out with the carriers because they
are the high-volume customers. Their criteria for purchase is price,
performance and [time to] installation,” said Peter Tierney, president
of Millennium, which currently is using leased fiber facilities but also
plans to lay its own fiber. Although Millennium currently offers only
OC-48 service, the company expects to soon have Sycamore's new
line cards that support OC-3/OC-12 services. “With the SN 8000,
my cost for provisioning a channel is lower. Speed to provision is a
problem in the carrier market today because it takes six to eight
weeks to deliver an OC-12, and even at the OC-3 level it could be a
long, drawn out time. An OC-48 could take six weeks to six months,”
Tierney said.
If DWDM technology is being used to provide virtually inexhaustible
capacity in the WAN, then metro networks, where linked SONET
rings now dominate, will also have to be upgraded as part of a
solution to eliminate the bandwidth bottleneck. Although the metro
DWDM is a recent phenomenon, metro systems from DWDM
vendors that have established themselves in the long-haul market are
beginning to be accepted by service providers such as America
Online, Cable & Wireless and Metromedia Fiber Network. AOL
selected Ericsson's ERION system last summer for a 140-km
network in Washington, D.C. Cable & Wireless has two metro
DWDM networks up and running and plans to install Ciena's
MetroWave product in about two dozen markets. “Metro DWDM is
lower in cost than putting in more fiber and allows quicker time to
market for services,” said Wesley Ford, transmission systems planning
director of Cable & Wireless, which typically installs the DWDM
equipment in carrier hotels and RBOC colocation facilities, since
about 80 percent of the company's market consists of services sold to
other carriers and ISPs. Currently, Cable & Wireless is running six
2.5-Gbps channels but expects to increase capacity by as much as
nine to 12 channels in some markets. Plans also call for increasing
channel speed to 10 Gbps during the first half of next year in some of
those markets. “Right now we are getting requests for (155 Mbps)
OC-3 and (622 Mbps) OC-12 services, and some inquiries for
OC-48, but we expect to start seeing firm orders for OC-48 within
the next 12 months,” Ford said.
White Plains, N.Y.-based Metromedia Fiber Network has targeted
certain tier one markets for building a fiber-rich local loop. According
to Ron French, Metromedia's vice president of commercial
marketing, Metromedia is putting in 864 strands of fiber in each cable
and laying multiple conduits within each build to support additional
fiber. Although the company's primary business is leasing dark fiber
(in October it announced a $500-million dark-fiber deal with Bell
Atlantic and a similar $300-million dark-fiber agreement with WinStar
Communications), Metromedia also offers managed optical services
for enterprise customers using Nortel's 32-channel OPTera metro
DWDM technology and Cisco Systems' Catalyst switches for a
managed Gigabit Ethernet service. “If an enterprise customer wants to
run multiple protocols, such as Gigabit Ethernet or (IBM) ESCON,
we can support those protocols. For the user, adding a service
becomes a plug-and-play capability,” French said.
Although vendors often pitch these DWDM systems to the metro or
interoffice market, UUNet's O'Dell pointed out that the CO is no
longer the only place where the demand for local bandwidth is
increasing. “COs are historical artifacts where you used to have a lot
of telephony hardware owned by the phone company. Now there are
a lot of places where there's a whole pile of communications
hardware. Sometimes these are called POPs, cable headends or
whatever. If a customer has an OC-48 from me, that [customer] might
as well be a CO because I've got the same tactical and strategic
issues of getting bandwidth to the customer. The metropolitan network
capacity has been burned through at a rate that is nothing short of
terrifying to traditional network planners, and that's one of the reasons
customers want to colocate their Web farms in our infrastructure:
They want to be able to provision additional capacity quickly,” said
O'Dell. UUNet is building a dedicated fiber network and will use
DWDM technology for fiber gain. In some places UUNet will own
the fiber, in other places it will be a managed service. However,
O'Dell declined to disclose whether it would own the DWDM
equipment.
While the advantages DWDM offers by delivering a forecast-tolerant
network are obvious, the systems are not going to be cheap, even in
their metro versions. This is partly because DWDM systems require a
pricey optoelectronic conversion at each node to add or drop a
channel, a capability likely to be used more often in the metro area
given the dynamic pattern of the traffic in this network segment.
Equally or perhaps even more importantly, some metro DWDM might
consume an entire wavelength (a situation referred to as burning a
wavelength) instead of packing multiple services until the channel is
full.
According to Ted Rado, marketing director for Mountain View,
Calif., start-up Alidian, this is because many metro DWDM vendors'
products are made up of transponders, which take in an optical signal
of a specific wavelength (typically 1310 nm), convert it to an electrical
signal, and back again to an optical signal of a different wavelength
(1550 nm) for long-haul transport. “They take in an OC-3 ATM
service from a piece of customer premises equipment, for example,
and transpond it onto a wavelength that their system can transport.
The disadvantage of this system is that it takes many wavelengths to
transport many services. Each wavelength can take in a service, but
when you want to add another, it takes another wavelength,” Rado
said. Alidian is expected to announce what it calls an Optical Service
Network solution that will not only ensure that wavelengths are
properly packed, but also are service aware. For example, the
solution recognizes whether certain traffic has an associated ATM
QoS or multiprotocol label switching (MPLS) tag. An Alidian
network will also be able to switch individual services on and off at
wavelength add/drop points.
Other vendors also are seeing shortcomings in metro DWDM
products and are promising to deliver more granular systems for the
metro and access markets. Qeyton Systems, a Swedish start-up that
will open a U.S. office, will soon unveil a DWDM system that will go
into beta trial with a large Swedish ISP as well as a U.S. ISP. The
Qeyton solution is intended for the interoffice ring as well as lower
bandwidth requirements. Qeyton's hub node will offer per-channel
upgrade to a maximum of 48 channels (the initial product will support
up to 16 channels); a collector node will provide a single channel back
to the hub node. The company boasts that it will offer the fastest
per-channel optical protection switching—less than 1 ms.
When it comes to the SONET-vs.-DWDM argument, Chromatis
Networks, of Bethesda, Md., has taken the middle ground via its
Selective DWDM technology. The Chromatis product allows
operators to start out with a shared 1310 nm ring (short reach 1300
nm lasers used in SONET are cheaper than 1550 nm lasers), which
might be used for connecting DSL access muxes (DSLAMs) to the
CO. As the bandwidth needs of a particular DSLAM site increase,
the service provider could drop a new wavelength to that site and
perhaps someday upgrade each site to support a wavelength. A site
needing to access the shared 1310 nm ring would do a simple
pass-through of the 1550 nm ring.
With the Talon MX product Kestrel Solutions is expected to
announce in Q1 2000, the company has added yet another wrinkle in
the metro DWDM landscape with its frequency division multiplexing
(FDM) technology. The Kestrel product takes in a number of signals
(up to OC-48), converts them to electrical signals and then combines
them using FDM. It then does another optical conversion that, in
effect, allows multiple FDM channels to be carried on a single
wavelength, thus enabling a 10-Gbps DWDM channel to handle up to
64 OC-3 channels. Kestrel's approach allows any FDM channel to
be added or dropped at any node. The Kestrel solution supports the
full 3Rs (reshape, regenerate and retime)—something all DWDM
networks do every so often to clean up signals. Although the
3R-process is expensive in DWDM networks, Kestrel claims to have
found a way around the problem with a single conversion for all
channels, in contrast to DWDM, which requires a conversion for each
regenerated wavelength. “Traffic from each user device, for instance a
router, would occupy one FDM channel; a single input could also be
dispersed across multiple FDM channels,” said Dawn Hogh, Kestrel's
vice president of marketing.
It remains to be seen whether these new optical networking products
will be adopted as quickly in the metro/access network in the way that
DWDM technology has become popular in IXC networks. A huge
amount of capacity in long-haul networks is driving down the prices of
long-distance services. Meanwhile, the metro/access network remains
a bottleneck that probably cannot be addressed by a TDM
technology like SONET. This being the case, it's a good bet that
some service providers—sooner rather than later—will take the
plunge into relieving the bandwidth blues in this segment of the
network via a technology that has already proved itself in the network
core.
Sam Masud is senior technology editor at
Telecommunications®.
RSNo. 305 |
