WHITE PAPER

David W. Jenkins

Metro WDM Network Design & Evolution:

Positioning for the Transition to Optical Meshes

Principal Engineer,
Advanced Technologies,
Tellabs

Dale A. Scholtens
Senior Principal Engineer,
Network Evolution,
Tellabs

Executive Summary
A revolution in metro transport is underway, driven by revenue and
profit opportunities associated with triple-play services including voice,
video and data, and demands for increasingly flexible Ethernetbased bandwidth requirements from businesses. Many service
providers are considering Metro Wavelength Division Multiplexing
(WDM) to meet the anticipated demand. Reconfigurable Optical
Add/Drop Multiplexers (ROADMs) provide necessary flexibility for
the evolution of Metro WDM networks as demand for packet-based
services continues to rise. ROADMs are also one of the driving
factors in the cost of Metro WDM. Service providers are therefore
challenged in designing Metro WDM networks that are cost-efficient
at every stage of deployment, yet nimble in light of changing service
offerings and demand patterns.
An interconnected ring architecture is the most cost-efficient means
of deploying and evolving Metro WDM. Interconnected rings make
sense because emerging services are deep-sourced delivered
from relatively few hubbing points in a region and therefore
require more backhaul bandwidth than legacy networks. An
interconnected ring architecture exploits the cost advantage of
two-degree ROADM devices since it limits the need for more costly
higher degree ROADMs to the role of ring interconnection. Importantly,
though, it is wise to choose ROADMs with higher degrees of
connectivity than immediately needed at points of current and
probable interconnection, thereby leveraging the ROADMs flexibility
to future-proof the metro network against unpredictable levels of
demand growth.
It is more cost effective to engineer for dedicated protection of
individual lightpaths in the optical layer today than for shared mesh
protection. Dedicated path protection is preferred not only because
of the poor economics of shared path protection in low-degree
topological networks, but also because of architectural limitations
with todays ROADM technology.
Even with bandwidth demand potentially exploding, interconnected
ring architectures will suffice for several years. Migration from 10G
(Gbps) to 40G transport should be economically feasible in the next
couple of years, assuming downward cost trends on certain components
continue. Service providers are therefore well-advised to deploy
Metro WDM equipment that can support 40G transport, without
regeneration, across the same distances that 10G can be carried
today. By deploying 40G-capable equipment, service providers
position themselves for increased capacity on high-demand
lightpaths without forklift upgrades. Independent of 40G
deployment, this paper describes a non-intuitive technique of ring

division that essentially doubles capacity in a congested region of a metro

network. Having engineered extra ROADM connectivity at network
junctions to begin with, service providers can use the technique to
advance Metro WDM networks toward a mesh architecture. The
technique adds significant capacity in congested regions of a
network and requires small capital outlay for the benefit accrued.
By exercising the technique several times as different areas of a
network approach congestion, a service provider can gracefully
and cost-effectively migrate from rings toward a mesh.
Notwithstanding the transition to 40G, service providers should
therefore plan for todays Metro WDM deployments to evolve to
optical meshes. By seeding with ROADMs that support high-degree
connectivity as networks are initially built, the ring division technique
can be applied to transform interconnected rings to meshes as
demand warrants. Complemented by further advances in optical
component architecture, shared protection in Metro WDM networks
may eventually become cost-effective even achieving the
considerable savings in protection bandwidth observed in some
SONET/SDH meshes today.

Introduction
A revolution in metro transport networking is underway, driven by
revenue and profit opportunities from residential service bundles
that include Internet Protocol Television (IPTV), along with
increasing demand from businesses for IP- and Ethernet-based
services. Many service providers are investing in multiservice
access and transport networks to provide for these services,
deploying equipment that supports optical interfaces exclusively.
As SONET and SDH were standardized in the 1990s, concern arose
over the inefficiency of the dedicated protection schemes used in
SONET/SDH rings. Indeed, the driver for the Bidirectional Line Switched
Ring/Multiplex Section-Shared Protection Ring (BLSR/MS-SPRing)
was a desire to improve the protection efficiency of rings as compared
to the 1+1 dedicated protection defined by SONETs Unidirectional
Path Switched Ring (UPSR), which allocated 50% of available
bandwidth for protection. BLSR especially its transoceanic variant
improved upon UPSR by allocating protection bandwidth just
sufficient enough to accommodate traffic actually requiring
protection, and making use of protection bandwidth only during fault
conditions. In adopting this approach, the normally unused
protection bandwidth became available for a preemptible class of
traffic, improving the overall cost efficiency of transmission. Although
BLSRs preemptible traffic class was not widely embraced by service
providers for operational and marketing reasons, BLSR also

See tellabs.com for more information about Tellabs Solutions

2 METRO WDM NETWORK DESIGN & EVOLUTION:

POSITIONING FOR THE TRANSITION TO OPTICAL MESHES

formalized the notion of Nonpreemptible Unprotected Traffic (NUT)

bandwidth that does not require protection. In practice, NUT
enabled practical improvement in ring efficiency since bandwidth
that would have been reserved for protection became available to
carry yet more NUT traffic.
As SONET/SDH became widely deployed, rings were interconnected.
It was soon realized that if networks were constructed with greater
connectivity than afforded by interconnecting rings at one or two
points, relatively more protection bandwidth could be shared within a
network as a whole. Subsequently, there has been much research in
mesh network architectures and strategies for shared protection in
meshes. Indeed, some service providers made strategic decisions to
build meshes to capture the savings in protection bandwidth in the
last few years, recognizing that the cost of introducing the necessary
dynamic switching intelligence within network elements was more
than offset by protection bandwidth savings and simplified
provisioning operations systems.
In these same few years, Metro WDM systems have moved from
point-to-point applications for fiber relief to generalized application in
Metro WDM networks. Indeed, Metro WDM is critical to the business
cases of many service providers who recognize the potential explosion
in bandwidth demand over the next few years. Metro WDM has
become attractive due to increased levels of component integration,
exemplified by ROADMs that incorporate variable optical attenuation,
and electrical fabrics capable of grooming and switching both packet
and TDM flows at rates of hundreds of Gigabytes per second. The
Tellabs 7100 Optical Transport System (OTS), for example, can carry
44 10 Gbps wavelengths for nearly a thousand miles without 3R
regeneration, and is architected to support 40G transmission in the
future. By exploiting the systems sub-wavelength switching, service
providers can often justify deployment when as little as 30 or 40
Gbps of capacity is needed in an area, knowing that the systems
ROADM-based optical switches provide flexibility for ongoing
network expansion.
We discuss an approach to Metro WDM design that exploits the
ROADMs flexibility in order to minimize capital outlay for WDM
infrastructure over time. In addition, the availability of multidegree ROADMs resurrects questions around the efficacy of
optical shared mesh protection; we therefore discuss the feasibility
of various protection schemes in light of ROADM architecture and
overall network economics.

The ROADM Revolution

A two-degree ROADM subsystem is shown in Figure 1. The subsystem
comprises two ROADM devices connected back to back. It is
termed a two-degree subsystem by virtue of supporting two WDM line
interfaces (East and West). Just as with electrically-based add/drop
devices, a series of two-degree ROADMs can be joined to form
a ring.

While the architecture of different vendors devices varies, the

following connectivity is available:
n Adds

and drops are done in association with the WDM interface with
which a ROADM affiliates. For example, a transponder connected
to one of the add/drop port pairs of the West ROADM device in
Figure 1 can launch its signal toward the West interface, and can
terminate a signal from the West interface. It cannot, however,
launch or terminate in association with the East interface.
Adds

Adds

Selector/
Attenuator

Selector/
Attenuator

to/from

to/from

amplifier

amplifier

West

East

Splittler

Demux

Splittler

West to East to
East West

Demux

Express
Drops

Drops

Figure 1. Two-degree ROADM subsystem

n Pass-through

is effected via the splitter and selector/attenuator

functions. The optical power of each wavelength entering from
the West amplifier is divided by the West ROADM devices splitter.
Some power is diverted to the West drop port in case the signal
is being dropped, but the remaining power is forwarded to the
Selector/Attenuator in the East ROADM via the Express Interface
linking the West and East ROADM devices (red line in Figure 1).
The East ROADM devices selector/attenuator can be dynamically
configured to forward wavelengths received from the express
interface toward the East interface, or instead insert wavelengths
from its own add ports. The attenuation function assures that
all wavelengths are at appropriate power levels in relation to one
another before being amplified by the East amplifier.

n ROADM

devices that position a splitter function ahead of the

selector/attenuator function (which is the case shown in Figure 1)
also support drop-and-continue connections. A wavelength entering
from the West amplifier can be simultaneously dropped by the
West ROADM and forwarded through the East ROADM device to
the East interface. Drop-and-continue connectivity is costeffective for point-to-multipoint signals such as those associated
with broadcast and pay-per-view IPTV; it minimizes transport
bandwidth and reduces the amount of higher level switching
needed for these services.

The Cost of Connectivity

Many service providers commencing WDM rollout in a region begin
simply, with a single ring traversing the set of sites from which services
are initially to be provided. However, Metro WDM networks can
grow to connect cities and towns that may be hundreds of miles
apart. Therefore, capability to extend the initial ring in multiple
directions is needed. ROADMs provide this capability.

See tellabs.com for more information about Tellabs Solutions

3 METRO WDM NETWORK DESIGN & EVOLUTION:

POSITIONING FOR THE TRANSITION TO OPTICAL MESHES

ROADMs supporting up to eight-degree connectivity are available

today, and ROADMs supporting ten degrees are on the horizon.
These devices enable rings to be interconnected, and of course the
ability to construct WDM meshes. Higher-degree ROADMs also give
service providers the freedom to rearrange network topology over
time as different services are rolled out, and to react swiftly to
changes in service demand patterns.
However, higher-degree devices are limited by internal losses
attributable to the splitter and selector functions, as well as the
increased complexity associated with selector/attenuator function.
This is evident even with a four-degree ROADM subsystem, illustrated
in Figure 2. Because of the number of express interfaces increases
geometrically with the degree of connectivity, ROADM cost also
increases with subsystem degree.

North
Selector/
Attenuator

East

Splittler

Demux

West
South

Why does this matter? The cost of a WDM network is dictated by

its major media plane components. Broadly, these are:
spans (amplifiers): WDM is attractive because its
amplification costs are lower than the corresponding costs of
stacked SONET/SDH multiplexers to carry equivalent bandwidth.
However, the decision to deploy a WDM link must be taken
carefully because the stepwise cost is still significant. Service
providers should therefore leverage existing capacity in their WDM
networks as effectively as possible before deciding to add links.

Figure 2. Four-degree ROADM

n WDM

switching (ROADMs): As suggested above, the flexibility

afforded by ROADMs comes at a price. To control network costs,
then, service providers must strike a balance. Higher-degree
connectivity is clearly justified at major hubs and interconnection
points, and low degree connectivity is generally all that is
necessary in outlying areas. But in a metro network there is often
a foggy area for which the appropriate degree of connectivity is
debatable in light of forecasting uncertainty involving service
mixes, service take-up rates and the like.

E
A
D

n Optical

adaptation (WDM transponder lasers): Transponder costs

closely follow the addition of traffic to the network, and are a
lesser consideration since their costs more directly align with
service orders. Still, design and evolutionary strategies that
minimize the number of transponder lasers for example, by
reducing need for regeneration tend to minimize the cost of a
network overall.

F
B
C

n Optical

Tellabs has been engaged many times to plan Metro WDM network
rollouts that are capital-efficient, achieving the balance that service
providers seek. We consistently observe that the lowest cost approach
to Metro WDM is to construct rings traversing many service hubs
and Central Offices (COs) in a region, with local distribution from
these rings as necessary. This is perhaps counter-intuitive given the
savings attributed to SONET/SDH meshes over the last several years.
However, our modeling shows that higher-degree ROADMs allow for
better utilization of WDM links in architectures of interconnected
rings (network degree closer to 2.0 than, say, 3.0) as opposed to
mesh designs (say, degree approaching 3.0 or higher) by deferring
the need for additional amplifiers and even entire WDM systems
in time.

2-degree ROADM
X-degree ROADM

Figure 3. Using ROADMs to defer WDM links

This approach can be appreciated by comparing the costs of two

independent WDM rings that traverse a pair of locations in common,
versus the same pair of rings constructed using a common WDM
link between those locations, as illustrated in Figure 3. In this case,
the cost of the three-degree ROADMs trades favorably against the
cost of a pair of two-degree ROADMs and an additional link between
sites A and B. First, the three-degree ROADMs eliminate cost
associated with interconnecting optical demands between the rings
within those sites. Second, and more importantly, the higher-degree
ROADMs defer the expenditure for the pair of amplifiers on the
second WDM link between sites A & B until the capacity of the

See tellabs.com for more information about Tellabs Solutions

4 METRO WDM NETWORK DESIGN & EVOLUTION:

POSITIONING FOR THE TRANSITION TO OPTICAL MESHES

shared link is exhausted. Given typical mixes of protected and

unprotected demands, this does not occur until both rings are at
least half full.
In modeling large networks, we have seen that mesh architectures
exhibit the same deficiency as independently engineered WDM
rings. When lightly loaded, meshes require relatively more links to
service a given demand set than interconnected rings, and WDM
links have associated costs. Therefore, the time-weighted cost of a
mesh especially at early stages of Metro WDM network build-out
when few demands need to be accommodated is higher than the
cost of a network of rings interconnected by multi-degree ROADMs.
The disadvantage of the mesh is not overcome until links become
heavily used, which occurs only over time. The situation can be
appreciated by considering a group of, say, 8 nodes requiring protected
services. When the number of lightpaths needed is few, the
minimum cost to link the nodes is a single ring comprising 8 links.
Any other structure will require more links, hence greater capital
outlay early on.
Clearly, a ROADM-based deployment strategy involving interconnected
rings makes sense, but exactly how much connectivity should a service
provider allow for over time? For example, Figure 3 depicts threedegree ROADMs. But four-degree ROADMs make more sense if it is
anticipated that the link between sites A and B will in fact eventually
block, since the second WDM link can be added between sites A
and B without impacting existing traffic.

Cost Factor

ROADM vendors are of course focused on reducing costs of their

devices. However, relatively more attention is being given to
two-degree devices than devices of higher degree. We foresee that
ROADM subsystems will track along a knee-shaped curve suggested
by Figure 4. This cost trend reinforces the case for interconnected
rings, and in fact argues that service providers should opt for
high-degree connectivity at ring interconnect points (and other
probable points of network expansion) as a hedge against uncertain
demand forecasts.

Such a strategy nicely balances ongoing capital outlay with long-term

flexibility. Though it requires some number of higher-degree ROADMs
during build-out, it defers most expenses for WDM links until they are
needed and enables network expansion without service interruption.

Implications of Deep-sourced Services

Metro WDM bandwidth demand is primarily driven by demand for
new services, for example, as IPTV rollout takes place throughout
a region, or as businesses increasingly demand Ethernet and IPbased services. Service providers intent on shuttering COs may also
strategically migrate portions of existing transport networks onto their
WDM infrastructure to minimize operational expense. As the metro
network expands, many service delivery points become increasingly
deep-sourced, and there is corresponding need to aggregate traffic
toward relatively few service delivery points. The effect of deep
sourcing is apparent, for example, with the transition from TDM
voice to VoIP: The number of switching points in a region becomes
far fewer, but the amount of traffic aggregated toward each point
increases.
But voice traffic pales in comparison with the bandwidth needed
for other services. On the consumer front, as IPTV transitions from
a broadcast to a unicast (video on demand) model the bandwidth
associated with the service may grow by two orders of magnitude,
from a few Gbps per Video Serving Office (VSO) to well over a hundred
Gbps per 10,000 subscriber VSO.1 Even basic Internet service is on
a speed ramp driven by competitive forces and visually rich, Web-based
applications: Internet edge routers handled traffic predominantly
sourced from 56 kbps access circuits just a few years ago, but 3-6
Mbps access is now common, and 30-100 Mbps service is appearing
as fiber is deployed deeper and deeper in access networks.
On the commercial front, businesses increasingly seek higherbandwidth WAN connectivity, not only to facilitate day-to-day
communications between sites but also, for example, to maintain
mission-critical storage. In the United States, law and policy
concerning record retention and disaster recovery such as
Sarbanes-Oxley, HIPAA, and SEC 17a-4 are furthering the
demand. Enterprise Strategy Group estimates that digital archive
capacity will grow 90% year-on-year worldwide through 2010 to total
28,000 Petabytes.2 This rate of accumulation is stimulating demand for
managed storage and archival solutions, creating significant
opportunities for Storage Service Providers (SSPs). SSPs tend to
deploy just a few server sites in metropolitan regions, in turn creating
a need for aggregation and backhaul to those sites.

10

Degree of Connectivity

Figure 4. ROADM cost profile

1 Assume each subscriber has one High Definition TV (HDTV) and two Standard Definition TVs (SDTV). Assuming advanced coding, HDTV requires about 9Mbps and SDTV requires 3Mbps. In a pure
broadcast model for which 150 SDTV and 40 HDTV channels are delivered, (150 x 3) + (40 x 9) = 810Mbps is delivered independent of the number of subscribers. In a unicast/VoD model, we have
(10,000 x (9Mbps + (2 x 3Mbps)) = 150 Gbps.
2 See: Web summary of report entitled Digital Archiving: End-User Survey and Market Forecast 2006-2010, Enterprise Strategy Group, January 2006.

See tellabs.com for more information about Tellabs Solutions

5 METRO WDM NETWORK DESIGN & EVOLUTION:

POSITIONING FOR THE TRANSITION TO OPTICAL MESHES

Even traditional private line services are arguably deep-sourced (or,

at least deep-connected) as are the SONET/SDH network-builder
services used internally to service provider networks. Emerging Ethernet
private line and protocol-independent wavelength services are also
deep-sourced. In practice, the routing of such services tends to be
through preferred hubbing points as a matter of policy for example,
through sites with test heads or higher order multiplexing facilities.

The implication of deep-sourcing is straightforward: Once an architecture

is decided upon, it is unlikely that the degree of connectivity in a
given area of the metro network needs to change significantly over
time since new services are likely to be sourced through the same
hubs as existing services. Even if a new hub is added, it is likely that
the hub will be positioned along an existing ring. Thus, a Metro WDM
network initially laid out as interconnected rings is to likely propagate
along the rings since this requires only the addition of lasers as
bandwidth demands increase. Still, higher-degree ROADMs offer
service providers the flexibility to migrate toward meshes if required,
and so are a cost-effective hedge for network evolution.

3
3

Degree N ROADMs
2-degree ROADMs
1-degree Spurs

Figure 5. Metro WDM network of interconnected rings

ROADM Support for Optical Layer Protection

Figure 5 depicts an Metro WDM network designed for deep-sourced
services, comprised of interconnecting rings (degree approximately
2.4 ).3 While the 1+1 path protection schemes of SONET/SDH
immediately come to mind, it is useful to evaluate whether shared
optical protection can be more effective than dedicated schemes,
along with the feasibility of implementing such protection within
ROADMs. To assess, consider the following:
protection in individual rings. Shared protection is effected
by either turning traffic around at a point of failure (like BLSR/
MS-SPRing), or redirecting it from its source upon failure
(transoceanic BLSR). As previously described in association with
Figure 1, a wavelength arriving on a given WDM interface can either
be dropped or expressed along to another WDM interface, but it
cannot be returned to the interface on which it arrived. Therefore,
BLSR-like optical protection is infeasible with todays ROADM
architectures. However, even if ROADMs could cost-effectively be
re-architected to support turnaround, the network-level benefit
would be minimal: The efficacy of turnaround in low-degree
networks is poor, nominally around 5-6% savings in distancebandwidth product.4

4
A

3
3

n Shared

n Shared

protection network-wide. Referring again to Figure 1 and

its associated text, wavelengths added/dropped by a given ROADM
device affiliate with that devices WDM interface. In the general case
of an N-degree ROADM subsystem, a given laser/photodiode can
launch/terminate a signal on only one of the subsystems N WDM
interfaces. Consequently, the ability to provide shared protection is
compromised. It can be supported along the interior of a lightpath
where light both enters and leaves a node on WDM links, and

4
C

Working Path
Protection Path

Figure 6. SNCP-protected lightpath

therefore traverses the express interfaces. But shared protection is

not possible on the first and last links of a lightpath since the
added/dropped wavelength does not traverse an express interface.
To fully support shared protection requires that added/dropped
wavelengths be de-coupled from individual ROADM devices
which requires an additional optical matrix between
transponders and the ROADM device array, or its functional
equivalent. Thus, even though shared protection has been shown
viable in networks of suitable degree, the cost and complexity of
further ROADM architectural changes must be traded against the
worth of a fully functional shared protection architecture.

3 The degree of a network is the average of the degree of the networks nodes.
4 See: Wayne D. Grover and John Doucette, Design of a Meta-Mesh of Chain Subnetworks: Enhancing the Attractiveness of Mesh-Restorable WDM Networking on Low Connectivity Graphs,
IEEE Journal on Selected Areas in Communications, Vol. 20, No. 1, January 2002, pp.47-61.

See tellabs.com for more information about Tellabs Solutions

6 METRO WDM NETWORK DESIGN & EVOLUTION:

POSITIONING FOR THE TRANSITION TO OPTICAL MESHES

n Perhaps

more significantly, past analyses ascribing value to

shared mesh protection in SONET/SDH networks presume that
each node is capable of both space and time switching. In the
WDM realm, ROADMs provide space switching alone. The second
degree of freedom in WDM is frequency switching changing the
color of light along its path. Today, there is no way to do frequency
switching cost effectively other than with back-to-back
transponders. However, their connectivity is limited to designated
pairs of WDM interfaces because of the ROADM device limitations
described previously. Significant advances in optical technology
are needed to provide the needed two-dimensional flexibility
arguably a blending of semiconductor laser and ROADM
technologies. Again, such changes will come only with additional
component complexity (hence, cost) and must be weighed against
gains attributable to shared protection.

n Dedicated

path protection. UPSR and SDHs Sub-Network Connection

(SNC) protection are reference techniques for path protection.
UPSR and SNC both require 1+1 signal bi-cast at the transmitting
edges of the protection domain, along with corresponding selector
functions at the receiving edges. UPSR protects a path within
individual rings, while SNC protects the subnetwork as a whole.
Techniques analogous to UPSR and SNC can be used for
individual wavelengths in the WDM domain.5

n ROADM

devices inherently support the bi-cast functionality

necessary for UPSR-style optical channel protection. However,
experience with dual-homed SONET rings has revealed the
complexity of selection across such interconnects, and the
difficulty of managing bandwidth along the common link between
sites (for example, the link between sites A and B in the lower half
of Figure 3) is well-known.

n In

WDM networks, SNC is preferred. A demand protected by SNC

is illustrated in Figure 6. SNC effectively creates a two node ring
for each protected demand. In contrast to UPSR, SNCs
protection structure is independent of the underlying WDM link
topology, meaning that the underlying WDM infrastructure can
evolve independent of the provisioning of its optical clients. SNC
requires only that diverse routing be possible through the network,
and that a single bi-cast/select function be provided at each
endpoint (at A and B in Figure 6). There is no difficulty crossing
ring interconnects, as is the case with UPSR.

n SNC

is also amenable to automation using standard routing

protocols such as OSPF. Routing protocol-based techniques can
handle not only demands protected within the WDM network, as
illustrated, but also client-protected demands (for example, at the
IP layer), thereby affording protection for lasers at either end of a
subnetwork connection without increasing the complexity of
provisioning operations.

Before leaving the topic of protection, it is useful to consider the strategy

for interconnecting WDM rings. Importantly, dual interconnection
comes at no cost. In Figure 6, bandwidth demand to the core ring
through point C cannot be protected on a node-diverse basis. For
the topology illustrated, the particular ring involved could have been
arranged to connect the core ring at both nodes C and D, not just
C alone. The costs are the same; six ROADM devices and associated
amplifiers are used whether the rings are singly or doubly connected
(i.e., 4+2 = 3+3).6

Leveraging Interconnected Rings

The Stealthy Path to Mesh
Metro bandwidth demand will inevitably increase. Because services
are deep-sourced, it is likely that congestion will first occur in the
core of the network perhaps on a ring connecting major cities and
towns in a region. Congestion manifests as wavelength blocking
where wavelength assignments for existing lightpaths are such that
a new traffic demand cannot be carried through the network on light
of a single color.
Service providers have several choices when encountering congestion,
with relief available at both lightpath level and WDM link level as
indicated in Table 1. In choosing amongst relief strategies, service
providers desire to minimize capital outlay; at the same time they
want to be confident that expenditures are appropriate in context of
longer term network evolution.

Technique

Network CapEx Efficiency

Lightpath Based

Reroute or change color

of existing lightpath

Ideal: Zero cost (but seldom

possible)

Insert regenerator

Poor: One demand served at cost

of two transponders

Upgrade speed

Better: 4x capacity for 2.5x cost

(in near future)

WDM Link Based

Bypass regional ring

Poor: Viable only if large number of

demands arise between two edges

Bisect regional ring

(add one WDM shortcut)

Poor: Provides for small to moderate

number of additional demands

Divide regional ring into two

interconnected rings (add
two WDM links)

Excellent: Near doubling of core

network capacity

Table 1. Efficiency of various congestion relief approaches

5 Refer to ITU G.872 for a formal definition of SNC. For purposes of this discussion, we restrict ourselves to 1+1 SNC schemes.
6 To be fair, it may be impractical to provide dual interconnect everywhere for reasons of geography or lack of sufficient demands requiring true node-diverse routing. In such cases, there is a small
compromise in overall service availability to traverse a single interconnect node.

See tellabs.com for more information about Tellabs Solutions

7 METRO WDM NETWORK DESIGN & EVOLUTION:

POSITIONING FOR THE TRANSITION TO OPTICAL MESHES

3
5
3

3
3

3
Figure 7. Ring bypass and bisection

At the lightpath level, costs of congestion relief can be fairly well

synchronized with increases in demand. Options at this level are:
n Rerouting

and/or changing color. In a few cases, relief can be

obtained at no cost by changing the routing or color of an existing
lightpath in order to free resources for a new demand. Importantly,
the ROADMs flexibility coupled with control-plane based
approaches to lightpath provisioning7 makes reroute and color
change more feasible than has historically been the case. Using
the control plane, such changes can be performed quickly, and
without repeated interaction with OSSs since only the final result
needs to be communicated northbound. Still, this technique is
very situational since an appropriate pattern of demands must
already exist in the network. Service Level Agreements (SLAs)
must also allow momentary interruption of service if unprotected
lightpaths are manipulated.

n Inserting

regenerators. Regeneration comes at a cost that is small

compared to WDM link addition; it also allows the routing of lightpaths
in different areas of the network to be decoupled. If the incidence
of blocking is low, regeneration makes sense. However, overall
cost-effectiveness of the network is compromised if very much
regeneration is needed, since equipment slots are tied up for
regeneration instead of servicing additional demand.

n Introducing

I
L

higher bandwidth lightpaths. Moving from 10G to 40G

transmission provides the most leverage, and service providers will
surely exploit this approach as it becomes economically feasible.
Electronically controlled dispersion compensators that allow 40G
lightpaths to travel as far as 10G lightpaths are appearing today.
Though not cost-effective yet, 40G devices appear to be on path
for mass introduction. Coupled with cost reductions in 40G tunable
lasers, we expect the historical 2.5x cost/bandwidth ratio for 40G
transponders to be achieved within the next couple of years (that
is, 40G price ~= 2.5 x 10G price). Pragmatic service providers will
therefore avoid wavelength blocking by upgrading congested areas
of their networks to 40G, freeing 10G lightpaths in the process. If

Figure 8. Regional ring division

traffic is growing at 40% per year, for example, a complete 10G to

40G upgrade defers additional WDM links for nearly four years.8
At the WDM link level, costs of creating capacity are higher, and thus
greater benefit to the network as a whole should be expected when
links are added. It is unwise to add links that cannot be wellutilized very soon, or do not create significant additional lightpath
capacity in the network as a whole. At the WDM link level, we
consider the following options:
n Bypassing

existing rings. Refer to Figure 7, which increases the

degree of nodes A and B on the edges of the network previously
illustrated. By increasing both of these nodes to degree three, a
WDM link (red dashed line) can be added between them. In
context of the network as a whole, bypass links such as this have
little topological value. They do not prove cost-effective unless
there is significant unprotected traffic between the sites they are
connecting usually many wavelengths. Otherwise, protected
traffic between A and B must use the existing network anyway.9
Also, demands originating from nodes A and B must detour into
the distribution rings in order to make use of the link.

n Bisecting

a ring. The intuitive approach to blocking in the core of

the network is to increase the degree of connectivity between a
pair of sites to relieve blockage. The link between nodes C and D
(blue dashed line) in Figure 7 illustrates this case. The technique
appears attractive from the standpoint of cost since only a single
link is being added. And it seems that by adding such links over
time, the network can evolve toward a mesh. However, the
networks degree of connectivity and therefore the flexibility to
route arbitrary demands is not significantly influenced by
adding just a few such links. In a network providing deepsourced services, the practical effect of such additions is to
provide about half of the anticipated bandwidth. Once again, the
fallacy is apparent when considering the proportion of traffic
requiring protection, since protection traffic must traverse the
existing network.

7 A companion paper covers ASON/GMPLS control-plane based operation of WDM and sub-wavelength networks.
8 Whether using 10G or 40G transmission, aggregating and switching traffic to fill lightpaths completely at network junction points is critical to metro WDM efficiency. Equipment such as the Tellabs 7100 OTS
supports integrated packet and TDM switching and aggregation, typically offering 30% savings at ring interconnects compared to use of external equipment for the purpose.
9 It makes no difference whether the WDM network (a) provides protection and diverse routing on behalf of simplex client demands, or (b) carries pairs of client-protected (or load-shared) demands on diverse
paths. Either way, pairs of information flows traverse the network diversely between pairs of nodes. Hence, freedom to create diversely routed pairs of lightpaths is a measure of a topologys effectiveness.

See tellabs.com for more information about Tellabs Solutions

8 METRO WDM NETWORK DESIGN & EVOLUTION:

POSITIONING FOR THE TRANSITION TO OPTICAL MESHES

n Dividing

a ring into two rings. Counter-intuitively, adding pairs of

links between sites on a ring is actually much more cost-effective
than bisection. This is shown in Figure 8, where nodes A and H
have been increased by two degrees each. In Figure 8, the
addition of the two links divides the core ring into overlapping
upper and lower rings. The upper ring comprises nodes A-B-C-DE-F-G-H, while the lower comprises A-H-I-J-K-L.

n By

dividing a ring approaching congestion into a pair of rings,

capacity becomes available on each of the newly created rings. In
the example shown, the second link addition is well worth its cost,
since the bandwidth of the entire core of the network is essentially
doubled: If the ring being divided has the capacity of n wavelengths,
the addition of the two links shown provides for n wavelengths
in each of the two rings created. This doubling of bandwidth also
reduces the likelihood of wavelength blocking as demands enter
the core of the network; thus the cost of the second link is further
offset through the reduced likelihood of regeneration for purposes
of changing colors.

To summarize, lightpath-level relief via rearrangement or regeneration

is appropriate when little additional demand is anticipated for some
time. Where demand is increasing in a pattern consistent with existing
lightpath routes, 10G to 40G upgrade should be considered to forestall
relatively greater expenditure on WDM links. Service providers are
therefore wise to choose WDM equipment that will support 40G
upgrade without service interruption, and transport 40G lightpaths at
the same distances as 10G lightpaths without having to upgrade
existing ROADMs or amplifiers.
When links must be added, service providers should opt for ring
division adding pairs of links over ring bypass and ring bisection.
Apart from doubling capacity in the area of the network where the
division occurs, ring division leads to a more fully connected
network over time. Using ring division, service providers can be
assured of providing bandwidth in areas of the network where it
is most needed. Ring division also allows a graceful migration to
meshed architectures. For instance, five division operations in the
network of Figure 8 will change its degree to about 2.8, a level at
which shared mesh protection begins to show some benefits.

Mesh and Eventually Shared Mesh Protection

We have shown that building a Metro WDM mesh outright is an
inefficient use of capital. Service providers will therefore prefer to
engineer interconnected WDM rings for some time. However, neither
comfort with ring architectures grounded in SONET/SDH experience
nor the transition to 40G should dissuade service providers from
architecting ring-based networks for migration to meshes. In
deploying Metro WDM, service providers should think about the
evolved topology of the Metro WDM network, which will likely be a
mesh comprising some number of high-degree nodes intercon
nected by chains of two-degree nodes. The key to migrating is to
install higher-degree ROADMs than are immediately needed at
probable points of ring interconnection in essence, scattering
space switching around at important junctions in the network,
independent of whether the switching capability will immediately be
used. By strategically deploying higher-degree ROADMs today and
using the ring division technique we describe, service providers can
gracefully evolve their Metro WDM networks toward meshes in
concert with whatever demands arise, when they arise, and with
capital outlay aligned with those demands.
As a matter of policy, meshes may come sooner than suggested by
bandwidth demand alone. Survivability trumps capital constraints that
ordinarily guide network architecture and deployment; the recent
experience with Hurricane Katrina is a reminder. Service providers
may therefore choose to engineer extra WDM links early on to
provide a slightly higher degree of connectivity than available with
interconnected rings. This increases the likelihood that paths will be
available for critical services, even in the face of multiple failures.
The Automatic Switched Optical Network/Generalized MultiProtocol
Switching (ASON/GMPLS) control plane is an enabler in this
strategy. For example, a critical SNC-protected service can be
provisioned by ordinary means. But in case both working and
protection paths fail, the control plane can be used to quickly
recalculate a temporary working route using available capacity,
even preempting NUT traffic on the extra links if necessary.
Whether driven by demand or concerns over survivability, we can
expect that Metro WDM networks initially engineered as rings will
evolve to meshes. As connectivity increases, service providers will
surely turn their attention to shared mesh protection. The attention is
justified in the face of decreasing revenue-per-bit from packet-based
services. SNC requires 100% additional bandwidth on raw basis,
and sometimes over 200% in terms of distance-bandwidth product
(when the comparative lengths of working and protection paths are
considered). In networks with sufficient connectivity, shared
protection can save 10-30% of raw protection bandwidth, and often
bring distance-bandwidth product for protection down to the 60%
range, or even lower.10

10 See John Doucette, and Wayne Grover, Comparison of Mesh Protection and Restoration Schemes and the Dependency on Graph Connectivity, 3rd International Workshop on the Design
of Reliable Communication Networks (DRCN 2001), Budapest, Hungary, October 2001.

See tellabs.com for more information about Tellabs Solutions

9 METRO WDM NETWORK DESIGN & EVOLUTION:

POSITIONING FOR THE TRANSITION TO OPTICAL MESHES

To achieve such savings in distance-bandwidth product requires

components that are not yet available components that offer the
freedom to change color (and ideally, perform full regeneration) as
needed along shared paths. Architecturally, this functionality must
be closely coupled with higher-degree ROADM devices even
directly integrated. Today, color changers are being researched but
none is near commercialization, much less integration with a
ROADM. Therefore, it will be some time before such components
are available.
Fortunately, service providers have time and current technology on
their side. ROADMs provide the flexibility needed for present Metro
WDM buildouts. Ring division using high-degree ROADMs will provide
capacity for years to come, independent of the timing of the
migration to 40G. As a leading vendor of optical transport gear,
Tellabs will continue to influence development of WDM components
ROADMs, amplifiers and lasers alike. Service providers can
therefore be assured that their Metro WDM networks will remain
capital efficient as WDM technology continues to advance.

North America

Asia Pacific

Europe, Middle East & Africa

Latin America & Caribbean

Tellabs
One Tellabs Center
1415 West Diehl Road
Naperville, IL 60563
U.S.A.
+1 630 798 8800
Fax: +1 630 798 2000

Tellabs
3 Anson Road
#1401 Springleaf Tower
Singapore 079909
Republic of Singapore
+65 6215 6411
Fax: +65 6215 6422

Tellabs
Abbey Place
2428 Easton Street
High Wycombe, Bucks
HP11 1NT
United Kingdom
+44 871 574 7000
Fax: +44 871 574 7151

Tellabs
Rua James Joule No. 92
EDIFCIO PLAZA I
So Paulo SP
04576-080
Brasil
+55 11 3572 6200
Fax: +55 11 3572 6225

The following trademarks and service marks are owned by Tellabs Operations, Inc., or its affiliates in the United States and/or in other countries: TELLABS, TELLABS and T symbol, and T symbol. Statements herein may contain projections or other
forward-looking statements regarding future events, products, features, technology and resulting commercial or technological benefits and advantages. These statements are for discussion purposes only, are subject to change and are not to be construed
as instructions, product specifications, guarantees or warranties. Actual results may differ materially. The information contained herein is not a commitment, promise or legal obligation to deliver any material, code, feature or functionality. It is intended to
outline Tellabs general product direction. The development, release and timing of any material, code, feature or functionality described herein remains at Tellabs sole discretion.
2011 Tellabs. All rights reserved. 74.1717E Rev. B 2/11