|
Archive
Flanagan Consulting
Network Analysts and Consultants
"We Have
the Experience"
-------------------------------------------------------------------------
ViewsLetter on Provisioning
Sept. 2002
#5
Lead Analysis
"LAMBDA" ON DEMAND IN OPTICAL CORE
MAY LEAD AUTOMATION OF CARRIER NETS
We framed the first statement of the provisioning problem, in VLoP#1,
partly in terms of the reluctance of carrier operations people to give
up
manual control of connection routing. So far, manual provisioning
has
been possible, despite causing significant delays in turning up new
services.
The huge complexity in setting up "clear wavelength" paths in all-optical
core networks makes manual operations less likely to succeed in the
future. Lambdas (from the Greek letter used by scientists and engineers
for a specific wavelength) aren't as simple to set up as T-1s or SONET
channels. Those digital transmission methods enjoy an inherent
regeneration of the signal in every node. Hardware vendors also
have
worked out ways to ignore many of the underlying problems and impairments
by automatically adjusting gain of amplifiers, sensitivity of receivers,
and shapes of pulses.
As long as light pulses stay optical, they are analog and accumulate noise
and distortion with distance. Optical amplification increases noise
as
well as the distorted signal. At some point, transmission equipment
must
convert signals to the electrical domain to regenerate them and remove
noise.
Remember when telephone networks relied on analog transmission?
Remember
loss plans? Inserted loss and gain? Translations between frequency
division multiplexed (FDM) channels? Every link and physical port
needed
"alignment" when installed--and periodic "tweaking" to maintain
performance.
In many ways, optical wavelengths behave like analog circuits:
-- A loss plan accounts for the drop in signal strength at each switch
point, filter, and wave division multiplexer (WDM).
-- Characteristics of individual fibers influence the needed optical power
to launch signals into a fiber, the sensitivity of the receiver,
and how
much compensation is needed for the various types of signal dispersion.
-- Changes in wavelength may be needed to find a free path across
successive WDM links--when first provisioned or, even more likely, when
rerouting to recover from an outage.
-- Long fibers need a specific amount of amplification to restore signal
levels.
All of these factors change with time, specifically with the temperature,
stress, and age of the fiber. Yes, you can provision and manage
such an
optical network with notes in a 3-ring binder and a command line interface
(CLI) for remote control. But your customers better be patient when
they
want service.
In the optical core we may have what Buckminster Fuller called "the thin
edge of the trial balloon." The need for configuration tools to
set up
lambdas may be the opening through which automation enters carrier ops.
Two Who Do
This issue of VLoP looks at two companies founded on the assumption that
(some day) clear-channel optical connections will be set up automatically.
Innovance Networks (www.innovance.com) and Meriton Networks
(www.meriton.com) designed products that respond to signalling
(instructions from another machine). While neither _requires_
operator
intervention, both _allow_ it because that's what carriers
still demand.
Both families of products automate what was formerly manual setup:
-- tests for loss across a fiber;
-- power level adjustment of laser sources;
-- selection of a specific wavelength on each WDM link.
Innovance Networks
Aimed at the long-haul market, Innovance's Transparent
Photonic Switch
(TPX 1000) combines WDM, pure photonic or all-optical switching (used
where possible), and optical-electrical-optical (OEO) switching (applied
where necessary for grooming a channel or regenerating a signal).
Sounds
like the old rule for LANs: bridge where possible, route where necessary.
There are some similarities.
If an optical signal arrives at a node where it doesn't terminate, and
the
signal doesn't need regeneration, the TPX can pass it on without
converting it to electrical form. This "express connection" may
amplify a
signal before inserting it in the next fiber link, but doesn't terminate
the optical path. As a result, it doesn't need optical muxes,
electronics, detectors, or lasers, a design holds down hardware cost.
Innovance (www.innovance.com) developed its own switch element in
partnership with an optical component vendor. Details are not yet
announced, but it is said to deal with individual wavelengths with a
"broadcast and select" approach that inherently supports multicast.
They
say the switch can control power level at the same time.
Their Dynamic Line System (DLS) is a subset of the TPX, an optical
add/drop mux (OADM). It contains the variable wavelength multiplexing
elements, tunable lasers, and photonic switching components, but not as
much OEO as the TPX.
Both products have building block components that include Erbium Doped
Fiber Amplifiers (EDFAs), Raman amplifiers, dispersion compensation, and
continuous testing capabilities. The nodes adjust for changes in
the link
due to ageing, repairs, and non-linear effects in the glass. All
good
stuff, but it does mostly what it's told.
The major advance is in the control plane. Here Innovance created
a
topology database, a routing engine that engineers paths for each lambda,
and a "pricing engine" that calculates the cost of each engineered path.
That cost may be based on dollars spent, but can also reflect other
influences. For example, the last wavelength on a link can be priced
higher than when many are available.
A network of TPX and DLS nodes, with the routing engines, will find the
four "best" paths across the network to meet a request for a wavelength
connection. Requests may come from an operator at the control console
who
indicates the desired end points. Or a device (router, switch) can
signal
for the connection using the standard for the optical user-network
interface (O-UNI) drafted by the Optical Internetworking Forum (OIF).
The single best path could be turned up automatically, but carriers still
demand hands-on control. The operator then picks one, which is
provisioned without further manual control. Sixty percent of their
R&D
goes into software.
Meriton Networks
The Metro Area Networks (MAN) core is Meriton's target,
specifically for
wavelength services. While their 7200 Optical Add/Drop Switch (OADX)
is
OEO, the electrical components are "transparent," which means they will
follow an input optical signal at any rate from 100 Mbit/s to 2.7 Gbit/s.
The receivers and lasers don't care what the bit rate is, either, so the
carrier can offer wavelengths without caring how the subscribers uses
them--but carriers probably won't, at least for now.
Carriers want to charge more for faster transmission speed (pricing to
value). They have asked for control of the top bit rate (an equivalent
to
that small capacitor across the twisted pair of a d.c. alarm circuit that
prevents it from being used for T-1). Increased administrative and
billing costs in the face of competition may make transparent service
just
as profitable--and more appealing to potential customers.
Meriton (www.meriton.com) made different trade-offs in designing for the
MAN. Shorter distances require less expensive lasers and detectors,
making it practical to provide an OEO transceiver for every wavelength
in
every node. Always having an electrical stage for every signal means:
-- it's easy to move between wavelengths by switching an electrical bit
stream;
-- any wavelength is just as good as any other on every fiber link,
avoiding stranded assets;
-- multicast is possible by electrical bridging;
-- when appropriate, signals may be processed (groomed, routed,
compressed, etc.);
-- monitoring and management are easier at the electrical level, keeping
hardware costs down.
With the electrical signal always available, a normal router can get any
information it needs quite easily. Routers in each node run Link
Management Protocol (LMP), part of Generalized Multiprotocol Label
Switching (GMPLS), to discover network topology and all wavelengths on
each link. GMPLS can also find a path when requested, but carriers
are
reluctant (still) to let the network do this job on its own.
But a MAN
of OADXs can set up a connection based on signaling (GMPLS) with no
operator intervention. Note that the connections is a transparent
wavelength (unless choked) and has nothing to do with the way MPLS defines
packet forwarding--the OADX "forwards" light pulses.
both Innovance and Meriton separate the control plane from the data plane.
These two companies differ in many ways, and address different market
segments, but they share the vision. Play on.
LINKS
-- Visit our web site for information about our consultant and analyst
services, speaking engagements, and updates on other activities.
-- Special thanks to www.webtorials.com for hosting our web site and
ViewsLetter.
-- For sponsorship information, email us at info@flanagan-consulting.com
or phone 703.855.0191.
-- Links for this issue:
www.meriton.com
www.innovance.com
"Flanagan Consulting" and "ViewsLetter" are
Service Marks of W. A. Flanagan, Inc.
Updated: 11 June 2003
|
