|
Enterprises have an insatiable appetite for bandwidth
within buildings, between sites over the wide area,
and out into the Internet. Fiber optic systems are the
preferred mode of transmission because of their
inherent capacity, low loss, reach, immunity to
electromagnetic and radio frequency interference (a
major cause of errors in twisted pair systems) and
undetected tapping, and their lower latency (photons
move faster than electrons). Fiber optic transmission
systems are used extensively within enterprises in
many forms and for many applications, including
in-building Ethernet risers and campus extenders,
high-capacity interfaces between and to storage and
application servers, metropolitan aggregation of all
inter-site traffic, and of course multi-Mbps access to
the public network and the Internet. Understanding the
characteristics of these systems, particularly as IP,
Ethernet, and optics converge, is a key challenge for
IT managers.
FIBER DIET
There are two kinds of fiber optic cables: multimode
fiber (used predominantly within buildings) and single
mode fiber (used in the metro and wide area).
Multimode fiber has a relatively thick core of glass
with a diameter of 50 to 162.5 microns allowing the
light beam to bounce around a fair amount. Its
attraction is that it is easier and cheaper to couple
the light source to the fiber because of the larger
core diameter. Single mode fiber has an 8 to 10 micron
core that more tightly guides the light, with
resulting lower loss of energy. In both cases, this
glass core cylinder is encased in a cladding glass
tube, commonly resulting in a strand of 125 micron
diameter. The refractive design of the core and the
cladding glass is designed to keep the light in.
Bending radius specifications have to be followed if
breakage is to be avoided.
Fiber optic cables have multiple wavelength
windows; these wavelengths exhibit substantially lower
attenuation than other wavelengths. Multimode
transmission often takes place using the 850 nm
window, while today's single mode systems use 1310 and
1550 nm. A convention has emerged to describe optical
systems in terms of these wavelengths (measured in
nanometers), rather than frequency, which is the speed
of light divided by the wavelength. Visible light has
a wavelength of from 400 to 700 nm, while the above
wavelengths fall into the near-infrared band.
Within some buildings, structured copper and fiber
cabling may have been installed when the building was
originally built, though typically constrained to
risers (vs. every desktop). In others, fiber cables
may have to be pulled, which may or may not be an
issue depending on the age and function of the
building. Outside, fiber optic cables are generally
buried, with the cost of trenching, conduit
installation, and fiber pulling ranging from $20 to
$150 per meter (trenching costs may represent 90
percent of the installation cost). While these costs
may seem high, once installed, optical cables have
economic lives measured in decades. In either case,
since the cost of the actual cable is measured in
cents per meter, multi-strand cables are used with
anywhere from a handful to many hundreds of strands
per cable, not all of which would be lit. A market has
emerged for unlit strands being swapped between
service providers to improve market coverage. The
costs of buried fiber has led to alternatives such as
using utility poles, and to explorations of innovative
approaches such as shallow "trenches" just under the
road surface and using sewer systems.
SEEING THE LIGHT
Lighting up the optical pipes is done using light
sources such as light-emitting diodes and lasers, the
former for shorter distances. Separate fibers are
normally used for each direction, though there are low
capacity systems that are bi-directional. These
optical sources run on a single wavelength of light
and are pulsed at up to 10 Gbps. They can be
integrated into a vast array of enterprise networking
devices, including Ethernet switches, routing switches
and routers, ATM switches, storage and application
servers, video codecs, and SONET muxes. SONET muxing
was designed to combine circuits operating at a broad
range of speeds onto a single wavelength operating at
bit rates of up to OC192 or 10 Gbps. Enterprises have
used SONET ring technologies for a number of years as
an efficient way of connecting a large number of
metropolitan sites.
Another form of multiplexing is called Dense Wave
Division Multiplexing (DWDM). This form of
multiplexing logically creates multiple parallel lanes
on the optical highway. In carrier backbone
applications, DWDM allows fiber capacities of 1.6 Tbps
by multiplexing 160 wavelengths, each carrying up to
10 Gbps of SONET, IP, and ATM. This has dramatically
decreased the cost per fiber bit-mile in these core
network applications. SONET and DWDM systems can be
standalone or combined in various ways to define a
range of communications solutions. For example, in
metropolitan applications, payload independent DWDM
systems have been developed which simultaneously
support traditional SONET, IP, and ATM, and also
various forms of storage networking and mainframe
channel extension traffic. A large number of
enterprises deploy these systems because of their cost
effectiveness and because they will accept any optical
payload, thus allowing the enterprise IT organization
to transparently handle increases in Ethernet speed
(e.g., going from 100 to 1,000 Mbps) as well as the
evolution of storage networking and mainframe channel
extension protocols.
GOING THE DISTANCE
The extension in reach of optical systems has been
equally dramatic with state-of-the-art capabilities
allowing light transmission of 2,500 miles without
regenerators, using instead amplifiers that boost the
signal every 50 to 80 miles!
But first some physics. Reach is dictated by three
main physical characteristics of optical systems:
attenuation, dispersion, and non-linearities of the
fiber (at various wavelengths). Optical systems are
all-digital, sending digital pulses representing ones
and zeros at an extremely high rate. Ideally, the
pulses that are transmitted are received as sent at
the other end of the optical pipe. Attenuation
decreases the strength of the signal, ultimately
making it hard for the receiver to find the signal in
the noise. Dispersion spreads the nice clean pulses,
ultimately corrupting adjacent pulses. Non-linearities
particularly come to play in DWDM systems, since these
support multiple wavelengths, each of which is
impacted differently by the transmission media.
As a result, it's quite common to transmit signals
over single-mode fiber for distances up to 25 miles at
1,310 nm, and 50 miles at 1,550 nm, typical
metropolitan distances. These distances are achieved
by using the highest power sources that can be sent
without interfering with other signals. One way of
going further is to insert regenerators in the path,
but this typically goes beyond what enterprises are
willing to do themselves. In carrier networks, these
regenerators can be standalone, as you might have in
an undersea cable, or they can be integrated into
add-drop muxes, which add new traffic to the fiber or
extract it from the cable to deliver to a user,
somewhat analogous to a bus stop (as opposed to
terminal muxes, such as on the enterprise premises,
which might be analogous to bus terminals where all
traffic is terminated).
However, regenerators are expensive and difficult
to manage, while add-drop muxes may not be
functionally required. There are a number of
sophisticated techniques that have been developed to
extend the reach of optical systems. Examples include:
- Special, more expensive non-dispersion and/or
low attenuation fiber.
- Externally modulated transmitters using a very
high-speed physical shutter in front of a
continuously running laser, rather than turning
the laser on and off quickly.
- Inserting a length of fiber, called a Dispersion
Compensation Module, which distorts the pulse in
the opposite way of the distortion the pulse
suffers on the fiber.
- Programmable control of the shape of the pulse
(particularly its width) to enhance reach
performance.
- Optical amplifiers that magnify the signal; two
competing/complementary technologies are called
Erbium Doped Fiber Amplifies (EDFAs) and Raman
optical amplifiers.
- A high-tech technique called Solitons, which
generates specially shaped pulses which interact
with each other as they go down the fiber to
compensate for dispersion in the fiber and other
non-linearities. This is a critical technique to
achieve ultra-long transmissions.
These are definitely interesting, but beyond the
scope of most enterprises, who prefer to turn to
experts to meet their long haul needs.
BEYOND FIVE NINES RELIABILITY
According to Meta Group, the average loss of revenues
due to a one-hour outage is $1,000,000, though this
varies by company size and industry. That's the
measurable loss. Outages often mean loss of customer
contact or diminished ability to serve the customer
well, potentially resulting in loss of a profitable
customer to the competition. Understanding how optical
networks have been designed to even exceed the
reliability of the phone network can be instructive,
particularly if it means bringing this level of
reliability to IP networks.
Fiber optic networks can be configured as mesh,
ring, and star networks, the latter two in fact being
cases of partial meshes. When running Ethernet over
dark fiber, reliability is achieved either by running
protocols such as Multi-Link Trunking (whereby
multiple point-point fibers are treated as a trunk
group) or by higher-level mechanisms (e.g., dynamic IP
routing). This is very much business as usual in that
the optical links are treated as any other physical
media. The development of SONET has established a new
level of reliability through Layer 1 mechanisms,
including end-to-end path protection and node-by-node
link protection with very fast recovery (i.e., 50 msec).
Because of their inherent simplicity, SONET rings
are used in one of two primary modes:
- Dedicated path protection, whereby data to be
protected is sent simultaneously on both sides of
the ring and the receiver chooses the best signal
(termed Uni-directional Path Switched Ring).
- Protection is shared on all fibers on the ring
by reserving half the line capacity between
adjacent nodes for protection (termed
Bi-directional Line Switched Ring).
That being said, what the enterprise sees is
ultra-high level of service. There is another
implication. Lessons learned from designing highly
resilient SONET rings are being applied to the design
of metro DWDM solutions and towards the
standardization of Resilient Packet Rings for IP and
Ethernet traffic. These latter two areas are of direct
interest to enterprises, since many are deploying
private solutions based on these solutions. The next
level of opportunity is to evolve IP routing systems
to fully leverage these optical developments.
THE LAST PHOTONIC WORD
The improvements in optical systems have been
dramatic, doubling every nine months in price
performance. This contrasts with Moore's Law, which
states that processing improvements are doubling every
18 months. We expect this to continue with more
bandwidth per wavelength, more wavelengths per fiber,
integration of optics with IP and Ethernet, and
photonic switching using sophisticated technologies
such as tiny mirror-based Micro ElectroMechanical
Systems (MEMS), and bubble inkjet/waveguide
technologies.
These developments in optical technologies
challenge the assumptions that have lead to the
widespread distribution of server, storage, and
routing intelligence across the network. Rethinking
enterprise networking in light of optical technologies
can lead to dramatic simplification, lower total cost
of ownership across IT, and freeing up of resources
for strategic investments.
Tony Rybczynski is director of strategic marketing and technologies
for Nortel Networks' Enterprise Solutions unit. For more information,
visit the company's Web site at www.nortelnetworks.com.
E-mail questions or comments to tonyryb@nortelnetworks.com.
[ Return
To The July 2001 Table Of Contents ]
|