You are on page 1of 4

ICTON 2010 Mo.C4.

Network Applications and Economic Considerations


for Fully Flexible Multi-Way ROADM/Optical
Cross-Connect Architectures
Richard M. Dorward, Fellow, IET
Ericsson Limited, United Kingdom
Tel: +44 1483 303 666, e-mail: richard.dorward@ericsson.com, RMDorward@theiet.org
ABSTRACT
This paper examines the flexibility and resilience requirements for multi-way ROADMs / optical cross-connects
in different network applications and the resulting economic implications for different architectures. It goes on to
review how optical (wavelength) switching can be cost effectively employed to complement electrical (OTN
and/or packet) switching in next generation network architectures, and the further requirements that this implies.
Keywords: WDM, ROADM, OXC, WSS, Optical bypass.

1. INTRODUCTION
In common with most hierarchical networks (for example, road networks with their arterial and feeder roads),
telecommunications networks have been traditionally built on high capacity point-to-point links (trunks) between
major centres, with “tree-and branch” networks to connect out from those major centres to individual termination
points.
Within any country, there will be some variations, dependent on the
local geography, but the predominant topology of the trunk networks is a
“nearest neighbour” partial mesh [1], for the same geographical and cost
reasons as in the road networks, so the degree of node meshing is relatively
limited and transit through centres is necessary for longer distance
connections. It also allows some degree of resilience by alternative routing
in the event of failure or congestion of a link. Within the
telecommunications network, the issue of alternative routing for link
protection also led to the creation of ring networks, particularly with the
advent of the SDH/SONET automatic protection schemes, as a more cost
effective route diversity topology to supplement and/ or replace the tree-
and-branch distribution networks in the larger, or metropolitan, distribution
environments.
The functions of flexible routing, protection and transit/bypass all
require switching and are a fundamental characteristic of the mesh and ring
architectures found in these networks. This is where ROADMs and optical
cross-connects, together with their lower capacity electrical switching Figure 1. Meshed core network [1]
counterparts (for example, SDH/SONET ADMs and DXCs) find their
applications.

2. FLEXIBILITY AND RESILIENCE PRINCIPLES


A typical SDH/SONET ADM and its extension to a DXC provide both an illustration of the network
requirements and a baseline against which their ROADM/OXC counterparts can be measured.
Both an ADM and a DXC are multi-port devices, in which the only real difference between line and tributary
ports is their capacity, so the ideal flexibility is any-to-any port connectivity, with the appropriate grooming
and/or consolidation between ports of different rates. This entails a full-connectivity, central switch, through
which all signals pass, and is the standard implementation today, although some of the early ADMs had more
limited flexibility, with only add/drop connectivity and no “loop-back” or “hair-pinning” between tributaries.
In terms of resilience, ideally, no failures should result in loss of network connectivity, other than
momentarily, as the result of protection switching and/or re-routing. In practice, this tends to be limited to
providing full protection against any single failure of a fibre/cable or in-service replaceable equipment unit.
However, it is usually complemented by layering and segmentation of the protection domains, which provides
some level of resilience to multiple failures and also facilitates fault location and planned maintenance activities.
The central switch is almost always automatically protected, generally on a 1+1 basis, independently of any
other form of protection. The line interfaces, for “East” and “West” in a simple 2-way ADM ring, and for each
direction in a multi-way DXC, are not generally protected locally, but are treated as forming part of the ring or
mesh network. This means that they are protected, together with the fibre/cable, by ring or mesh re-routing if
they, themselves, fail. However, they may also be protected on a 1+1 or 1:N point-to-point MSP basis for the

978-1-4244-779-/10/$26.00 ©2010 IEEE


ICTON 2010 Mo.C4.4

highest level of protection if multiple fibres/cables are available. On the client or tributary side, interface cards
may similarly have the option of being unprotected, MSP protected, or sometimes 1:1 or 1:N card protected.
Each of these protection domains is independent, so a failure in one does not prejudice operation of any other.

3. OPTICAL DOMAIN SWITCHING TECHNOLOGY EVOLUTION


While DWDM with optical amplification is accepted as the most economic high capacity transport technology,
the immaturity of optical switching technology and the resulting costs, optical losses and limited optical
functionality have driven the need for much simpler switching architectures, with a consequent reduction in the
levels of flexibility and resilience provided by the transparent optical domain.
The nearest all-optical equivalent of a basic electrical ADM/DXC, for example, uses a protected central
optical switch matrix, which was the target architecture for a number of manufacturers, and sophisticated micro-
electro-mechanical systems (MEMS) technology and products were developed at a relatively early stage [2].
However, even without provision for line or tributary protection, this approach has so far failed to reach the
required size/capacity and cost points required for widespread economic deployment, and optical switching has
tended to evolve more from simpler 2-way (R)OADM architectures, with the more limited connectivity of the
early electrical ADMs. The integration of the wavelength multiplexing/demultiplexing functions in the
Wavelength Blocker (WB) components provided a crucial implementation advantage, and the predominant
optical switching deployed today uses WB and, more recently, its extension to Wavelength Selective Switch
(WSS) switching technology, and ROADM-based architectures.

4. RING-BASED NETWORKS
4.1 Two-way ROADM architectures
The classic 2-way ROADM architecture is based on operation
in a ring network, with the ring providing the alternative path WSS unit

x x x x x x
for network protection, exactly as it does in an SDH/SONET
ADM ring. However, whereas in the SDH/SONET ring the line Tx/Rx Tx/Rx
cards are the only part of the ADM that are protected by the WDM WDM
x x x x x x

East
ring, in the ROADM case, there is no separation of the West
protection domains between the line interfaces, the mux/demux
and switching functions, and the transponder client interfacing WSS unit
functions. So, the entire network element becomes effectively Mux /
Demux
just two separable East- and West-facing halves, protected by
the ring. Normally the East- and West-facing transponders just
provide a pair of client interfaces for external protection, or they
Transponders Transponders
can be combined with passive couplers/splitters on their client
sides and switching between the transponders to provide Figure 2. Two-way WSS ROADM
a single protected client interface.
Separation of the transponder protection from the line side/ring protection would be possible by having a pair
of transponders (or a single unprotected transponder) with the capability to switch to either line direction on their
line interface sides. However, that generally reduces the capability to detect failures elsewhere in the network,
on which transponder protection switching is based, and introduces additional loss at a more critical point in the
network. The problem is that the switching function can no longer be based on all the digital parameters of the
two incoming signals, as detected at the electrical level in the transponders, but only on simply detectable optical
parameters, such as loss of light. So, separation of transponder protection, particularly in its unprotected single
transponder configuration, is rarely used in DWDM networks, where the capacities involved normally demand
the highest possible level of monitoring and physical transponder protection.
4.2 Electrical layer inter-working
Apart from the simple case where transponders provide the final, high capacity client interfaces, the DWDM
layer interfaces directly into the electrical layer on the client side of the transponders. Interfacing pairs of
transponder client interfaces into the electrical layer merely demands the equivalent pairs of interfaces on the
line side of the electrical switch, or merging these with the transponder functions
In general, and particularly for next generation flexible networks, it is assumed that network operation should
require the absolute minimum of manual reconfiguration. This entails the pre-provisioning of fully tuneable
transponders or line interfaces on the electrical switch and the use of the “colourless” ROADM architecture,
where WSS-based demultiplexing provides flexible allocation of wavelengths, or “colours” to them [3].
In terms of resilience, the East/West 2-way ROADM architecture and separability matches the way the
electrical layer is protected and place no additional demands on the optical layer architecture.

2
ICTON 2010 Mo.C4.4

5. MESH-BASED NETWORKS
5.1 Multi-way ROADM architectures
Extension of the ROADM architecture to a multi-way network node should meet, at least, the same levels of
flexibility and resilience against single faults as in the 2-way case.
Simple extension of the basic architecture consists, in either the “coloured” or “colourless” architectures,
merely of replicating the line ports and associated transponders, and using some of the “mux/demux” ports on
each line port 1×N WSS unit to provide a full mesh between them [3]. This provides full wavelength routing
flexibility between the line ports and an exactly equivalent level of network resilience and protection for transit
traffic, but not so readily for the transponders and the add/drop and terminating traffic.
For a fixed traffic matrix or one in which slow/manual reconfiguration is acceptable, it is easy enough to
identify and install the relevant pairs of transponders for protection, as required, even if they are spread between
various different line ports, but this becomes unworkable for any dynamic network and traffic matrix. In this
case, the add/drop ports and transponders need to be flexibly pre-provisioned, which entails provisioning for the
maximum number of channels to be accessed from each line direction, so a “colourless” architecture is much
more efficient: furthermore, flexible/dynamic pairing for protection requires another level of switching to be
introduced. This switching layer could either be a protected optical matrix, which would be of a more viable size
in this case as it only has to deal with the add/drop and terminating traffic, or it could be the electrical switching
layer itself, although the inefficient distribution of add/drop ports increases the size of either switch significantly,
and partially cancels out the benefits of optical bypass in the electrical case.
5.2 Electrical layer inter-working in a mesh network
Much more efficient provisioning for the total number of add/drop ports and transponders in a multi-way node
can be achieved with a central bank of those functions, rather than having to handle the maximum number on
each line port. Similarly, protection switching and inter-working with the electrical layer can be simplified by
having just two groups of transponders or electrical switch ports, in a 1+1 East/West protection pair, as in the
2-way case. As the basic architecture associates the add/drop direction and transponders with each line port, an
architecture providing a flexible add/drop direction (and centralised transponders) is known as “directionless”.
One simple way of implementing such an architecture is to use the directional flexibility between the line
ports and provide an additional pair of them for the add/drop and termination function [4]. Since the “local”
add/drop port is no longer needed, the fan-in/out of the WSS’s is only increased by one, and that is for the
second (protection) port. However, this use of “internal” line ports introduces the issue of wavelength
contention. While it is natural that external line ports should only support a single instance of each wavelength,
this is not acceptable in a central add/drop or termination function, which must be able to handle as many
instances of each wavelength as there are external line ports, to provide “contentionless” transponder access.
The simple architecture can be extended to avoid contention by matching the number of internal to external
line ports [5], but this again complicates the provision of 1+1 protection, or requires the electrical switch to be an
essential part of it, unless such an architecture is further extended to provide two separate protection groups.
Then, for “contentionless” transponder access, each group needs to provide half the matching number of internal
line ports, rounded up to an integer, all of which further increase the fan-in/out requirements and losses of the
WSS’s and couplers involved. This increased flexibility/resilience also comes at a significant cost as the
increasing numbers and functionality of the WSS’s become increasingly expensive (although still not as
expensive as a duplicated central matrix switch architecture!).
What is really required in this area is “contentionless” M×N WSS technology, but this is only a concept so
far. Many other architectures have been proposed to meet the full range of requirements, see e.g. [6] [7], but
most involve relatively large optical matrix switches, which continue to be expensive niche products, or large
arrays of smaller optical switches. So, research is still ongoing for the optimum flexible and resilient,
“colourless”, “directionless” and “contentionless” architecture, and the components to support it economically.

6. OPTICAL BYPASS
For both SDH/SONET-based TDM traffic and IP/Ethernet-based packet traffic, the existing electrical layer is
focussed on service delivery and switching for traffic aggregation and grooming. In general, it has to deal with
very fine granularities and huge numbers of services, which require large amounts of complex digital processing
at the bits and bytes level, and it can do this readily as a function of the sophistication and low cost of the silicon
CMOS integration that it employs. However, as the bit rate increases, the power consumption increases and
more complex and expensive SiGe and other technologies are needed to operate at the speeds required for the
latest generation of transport rates.
The sheer capacity and signal transparency provided by the all-optical domain make it an essential complement
to the existing electrical aggregation and switching network, providing the transport and optical bypass required
when the capacities and bit rates involved make purely electrical technologies less viable, from both a cost and

3
ICTON 2010 Mo.C4.4

power consumption perspective. Recognising that all high capacity core links use DWDM transport, the
addition of optical wavelength switching, in its simplest case, allows the transit traffic at a node to bypass the
electrical switches, thereby reducing the number of transponders and the size, cost and power consumption of the
electrical switch needed at that node. This can be simply evaluated in the case of traffic demands at the same
granularity as the transport wavelengths, such as 10Gb Ethernet over 10G wavelengths, and analysis of real
customer traffic demands at a typical 4-way node in the German core network showed transponder savings of
nearly 70% (and resulted in that customer moving from a point-to-point DWDM network to one using WSS-
based optical switching).
However, for mixed traffic, with different traffic demand and transport SDH/SONET
rates, this does not necessarily scale well: while simple wavelength
bypass is good for fully-filled wavelengths, it is likely to be less economic
than electrical switching for light and diverse traffic demands. The
economics depend on network topology and the nature of the traffic flows
within it, as well as the relative economics of the electrical and optical
technologies themselves, and studies have shown that a hybrid TP
MXP
optical/electrical switch offers the best overall, scalable solution [8].
In such a hybrid switch architecture, the level of traffic grooming that
can be done economically in the electrical domain is key to the overall
cost-effectiveness. One solution gaining increasing international favour
with both customers and equipment vendors is the use of an intermediate
sub-wavelength OTN [9] switching layer to provide a coarser granularity
(and lower cost, size and power consumption) than that of the packet- and
SDH/SONET-based service switches. Its low switching latency is also a
significant advantage to many customers. There will still be cases where
direct transponder/muxponder access is the most efficient, and cases
where the routers can fully fill wavelengths, so the final result is a multi- Figure 3. Core node functionality
layer hybrid core node architecture of the sort shown in Figure 3.
7. CONCLUSIONS
Optical transport network topologies demand both 2- and multi-way optical switching node architectures, but
naturally multi-way technologies, such as large optical matrix switches, have never reached the size/capacity and
costs required for widespread economic deployment, and optical switching technology has evolved more from
simpler 2-way (R)OADM architectures.
The predominant optical switching deployed today, using WSS ROADM technology, is optimal for 2-way
and relatively static multi-way nodes, but its extension to the flexibility required in the next generation of fully
dynamic multi-way mesh networks comes at a significant cost as the increasing numbers and functionality of the
WSS’s required become increasingly expensive. So, research is still ongoing for the optimum transparent optical
switching node architecture and the components to support it most economically.
Nevertheless, the ability of optical bypass to reduce electrical switch sizes means that high capacity WDM
transmission and fully transparent optical switching remain highly cost-effective technologies to complement
electrical (OTN and/or packet) switching in dynamic next generation transport networks.
ACKNOWLEDGEMENTS
The author would like to acknowledge the inputs from his colleagues in reviewing and refining the concepts in
this paper and, particularly, in providing the results of the network example quoted.
REFERENCES
[1] C.V. Saradhi et al.: Control Plane Issues in Cross-Layer Optimized Dynamic Optical Networks, in Proc. ICTON 2009,
Ponta Delgada, Portugal, Jun. 2009, paper Mo.B3.2.
[2] Calient Networks: Calient Networks Exhibits All-Photonic Switch Architecture at SUPERCOMM 2000, Calient
Networks Press Release, Sunnyvale, CA, USA, May 2000
[3] R.M. Dorward: Comparing the different technology approaches to ROADM’s and understanding the implications on
cost and performance, IIR Conference on WDM & Next Generation and Optical Networking, Cannes, France, Jun. 2007.
[4] P. Roorda, B. Collings: Evolution to colorless and directionless ROADM architectures, in Proc. OFC/NFOEC 2008, San
Diego, CA, USA, Feb. 2008, paper NWE2.
[5] Ericsson AB: Switch Node, International Patent Application No. PCT/EP2008/060893, Aug. 2008.
[6] S. Thiagarajan et al.: Direction-independent add/drop access for multi-degree ROADMs, in Proc. OFC/NFOEC 2008,
San Diego, CA, USA, Feb.2008, paper OThA7.
[7] J.J. Chen et al.: En Route to grouping-constraint free, colorless directionless ROADMs, in Proc. OFC/NFOEC 2010,
San Diego, CA, USA, Mar. 2010, paper NThF1.
[8] L Blair, S. Thiagarajan: Impact of moving to 100Gbps on the cost per bit of a USA national network, in Proc.
OFC/NFOEC 2010, San Diego, CA, USA, Mar. 2010, paper NME4.
[9] ITU-T: Interfaces for the Optical Transport network (OTN), Recommendation G.709, Dec. 2009.

You might also like