Sie sind auf Seite 1von 70

Project P709

Planning of Full Optical Network


Deliverable 2
Basic factors influencing optical networks
Volume 5 of 5: Annex D Interaction between the optical and the
client layer

Suggested readers:
PNOs studying potential upgrade possibilities for their SDH networks
System engineers and network planners
Experts on standard bodies of ITU-T SG13 (Q 19), SG15 (Q 16, 17, 20) and
ETSI TM-1WG2/WG3
Researchers engaged in the field of optical transmission networks and
technologies

For full publication

May 1999
EURESCOM PARTICIPANTS in Project P709 are:

Finnet Group
Swisscom AG
Deutsche Telekom AG
France Tlcom
MATV Hungarian Telecommunications Company
TELECOM ITALIA S.p.a.
Portugal Telecom S.A.
Telefonica S.A.
Sonera Ltd.

This document contains material which is the copyright of certain EURESCOM


PARTICIPANTS, and may not be reproduced or copied without permission.
All PARTICIPANTS have agreed to full publication of this document.
The commercial use of any information contained in this document may require a
license from the proprietor of that information.
Neither the PARTICIPANTS nor EURESCOM warrant that the information contained
in the report is capable of use, or that use of the information is free from risk, and
accept no liability for loss or damage suffered by any person using this information.
This document has been approved by EURESCOM Board of Governors for
distribution to all EURESCOM Shareholders.

1999 EURESCOM Participants in Project P709


Deliverable 2 Volume 5: Annex D Interaction between optical and clinet layer

Preface
(Prepared by the EURESCOM Permanent Staff)
The advances in optical fibre transmission technology over the past years have kept
pace with the demand for increased bandwidth. In particular the introduction of the
WDM technology enables Telecom Operators to upgrade the capacity of their
networks by an order of magnitude. The evolution of photonics makes the
development of optical switching and routing structures in the core and metropolitan
part of the transport network possible.
As a consequence, the development of an optical network infrastructure will enable
the flexible, reliable and transparent provision of transport services for any type of
traditional and innovative services and applications. Taking into consideration the
current trends, the objective of network planning is to find the best possible balance
between network implementation cost, network flexibility, network availability and
survivability, subject to service requirements and topological constraints.
The aim of the P709 EURESCOM Project is to investigate a number of alternative
strategies for the planning of the optical transport network - with massive deployment
of WDM, OADM, and small size OXC- that will be used in a middle term future.
This is the second Deliverable (D2) of P709. D2 summarises the most important
factors that have to be taken into account when preparing the planning of optical
networks. Restoration and protection techniques implemented in optical networks are
assessed in terms of requirements, constraints on network planning and upgrading, as
well as their interaction with client layer functionalities. A study of resource allocation
and impact on network planning and upgrading is also presented.
We should remind the reader that the first P709 Deliverable (D1) provided an
overview over network architectures, which potentially may be used in the future and
D3 will give an analysis of the existing network planning methods, plus guidelines for
planning future optical networks.
The present Deliverable (D2) is a very useful study for Optical Network planners &
system engineers, and experts on Standard Bodies of ITU-T SG15 and ETSI TM1
(WG2 & WG3).

1999 EURESCOM Participants in Project P709 page i (v)


Volume 5: Annex D Interaction between optical and clinet layer Deliverable 2

page ii (v) 1999 EURESCOM Participants in Project P709


Deliverable 2 Volume 5: Annex D Interaction between optical and clinet layer

Table of Contents
Preface............................................................................................................................. i
Table of Contents ..........................................................................................................iii
Abbreviations ................................................................................................................. v
1 Introduction ............................................................................................................ 1
2 Overview of SDH architectures.............................................................................. 2
2.1 What is a network architecture? ................................................................... 2
2.2 SDH network layers ..................................................................................... 4
2.3 Classes of SDH network architectures ......................................................... 4
2.4 Trail protection............................................................................................. 5
2.4.1 Linear VC trail protection ............................................................... 5
2.4.2 Linear Multiplex Section protection................................................ 5
2.4.3 MS Dedicated Protection Rings ...................................................... 5
2.4.4 MS Shared Protection Rings ........................................................... 6
2.5 Subnetwork connection protection............................................................... 7
2.5.1 General characteristics of SNC protection ...................................... 7
2.5.2 SNC protection with inherent or non-intrusive monitoring ............ 8
3 Overview of optical architectures......................................................................... 10
3.1 What is an optical network architecture? ................................................... 10
3.2 Classes of architectures .............................................................................. 10
3.3 Capabilities and limitations of optical network architectures .................... 12
3.4 Architectures selected in P709 ................................................................... 12
3.5 Coloured Section Ring ............................................................................... 12
3.5.1 General .......................................................................................... 12
3.5.2 Description .................................................................................... 13
3.5.3 Protection ...................................................................................... 14
3.5.4 Functional model........................................................................... 15
3.5.5 Implementation.............................................................................. 15
3.5.6 Design rules................................................................................... 17
3.6 Optical Multiplex Section Shared Protection Ring .................................... 20
3.6.1 General .......................................................................................... 20
3.6.2 Description .................................................................................... 20
3.6.3 Functional model........................................................................... 21
3.6.4 Implementation.............................................................................. 22
3.6.5 Design rules................................................................................... 24
3.7 MWTN mesh.............................................................................................. 24
3.7.1 General .......................................................................................... 24
3.7.2 Description .................................................................................... 24
3.7.3 Functional model........................................................................... 25
3.7.4 Implementation.............................................................................. 26
3.7.5 Design rules................................................................................... 28
3.8 Conclusion ................................................................................................. 29
4 Dimensioning methods......................................................................................... 30
4.1 Dimensioning methods for multiple-ring OMS-SPRing or SDH
SPRing networks........................................................................................ 30
4.1.1 Types of suitable networks............................................................ 30
4.1.2 Routing problems in multiple-ring networks ................................ 31

1999 EURESCOM Participants in Project P709 page iii (v)


Volume 5: Annex D Interaction between optical and clinet layer Deliverable 2

4.1.3 The dimensioning method .............................................................32


4.1.4 Application to OMS-SPRing .........................................................34
4.2 Optical dimensioning methods ...................................................................35
4.2.1 Optical node structures in a meshed network ................................36
4.2.2 Dimensioning methods ..................................................................37
4.2.3 Application to a network example .................................................40
5 Optical/Electrical layers : management possibilities ............................................41
5.1 Management information processing .........................................................41
5.2 Possible interaction between SDH overhead and optical overhead............43
5.2.1 Constraints and limitations ............................................................43
5.2.2 Use of unallocated SDH overhead Bytes for optical
overhead.........................................................................................45
6 Dimensioning methods for optical/electrical layers..............................................49
6.1 The traffic grooming problem or how to minimise the number of
terminal equipment in the network.............................................................49
6.2 Dimensioning method based on simulated annealing.................................53
6.3 Application of the tool to network examples..............................................54
6.3.1 Results for 1+1 protection .............................................................54
6.3.2 Results for SPRing protection .......................................................57
6.3.3 Comparison between 1+1 and SPRing protection .........................58
7 Conclusion ............................................................................................................61
References ....................................................................................................................62

page iv (v) 1999 EURESCOM Participants in Project P709


Deliverable 2 Volume 5: Annex D Interaction between optical and clinet layer

Abbreviations
AIS Alarm Indication Signal
APS Automatic Protection Switching
DAt Dispersion Accommodation (in optical transmission section
layer)
DAc Dispersion Accommodation (in optical channel layer)
DCC Data Communication Channel
EMF various Equipment Management Functions
FFS For Further Study
LOS Loss Of Signal
MCF Message Communication Function
Mod Modulation/demodulation (O/E conversion)
OA&M Operation, Administration and Maintenance
OCH Optical CHannel section
OCH/Client_A Optical CHannel/Client Adaptation function
OCHOH Optical CHannel OverHead
OCH_T Optical CHannel Termination function
OM Optical Multiplexing
OMS Optical Multiplex Section
OMS/OCH_A Optical Multiplex Section/Optical CHannel Adaptation function
OMSOH Optical Multiplex Section OverHead
OMS_T Optical Multiplexer Section Termination function
opt. DCC Optical Data Communication Channel
OSC Optical Supervisory Channel
OTS Optical Transmission Section
OTSOH Optical Transmission Section OverHead
OTS/OMS_A Optical Transmission Section/Optical Multiplex Section
Adaptation function
OTS_T Optical Transmission Section Termination function
RDI Remote Defect Indication
SDH Synchronous Digital Hierarchy
TMN Telecommunication Management Network
WA Wavelength Assignment
WDM Wavelength Division Multiplex

1999 EURESCOM Participants in Project P709 page v (v)


Deliverable 2 Volume 5: Annex D Interaction between optical and clinet layer

1 Introduction
The scope of this Annex consists in proposing optical/electrical based architectures
and describing the management capabilities as well as dimensioning methods. Such
architectures correspond to the ones depicted in Deliverable 1.
The second chapter concerns an overview of SDH architectures. It describes different
SDH layers and their functionalities. Functional models are described according to
ITU-T recommendations namely G805 and G803. Different protections schemes are
given and compared.
An overview of optical architecture is presented in the third chapter. The concept of
optical architectures is given and classification methods as well. The optical network
layers are detailed and the architectures selected in Deliverable 1 of P709 are
developed. General information, functional models, proposed implementations as well
as design rules are given.
The dimensioning methods chapter provides a description of different architectures
mainly optical based ones as well as optical meshed networks. The dimensioning
problem is presented for each architecture under studies assumptions. Algorithms for
dimensioning MS-SPRing and OMS-SPRing based networks are presented and
theoretical references provided.
Concerning the meshed optical networks, the presentation focuses on node
dimensioning according to the assumptions described in Annexes A and B of this
Deliverable. Numerical results are also provided on the European network described
in Annex C of this Deliverable.
Thereafter, in addition to P615 project, a brief analysis of the management
possibilities of the optical layers in the conjunction with the client SDH electrical
layer is given. The informations management model used to control the optical layer
is described. Finally, the interactions between SDH layer capabilities and the optical
one are also described.
In the previous chapters, dimensioning optical networks is presented. As SDH is
client layer of the optical layer and according to the subject covered by this Annex,
such two layers should be dimensioned together. Dimensioning are different from the
ones used for the pure SDH networks or the pure optical networks. However, adapted
algorithms for dimensioning optical/electrical networks are presented. Finally
evaluated networks under given assumptions are presented.

1999 EURESCOM Participants in Project P709 page 1 (62)


Volume 5: Annex D Interaction between optical and clinet layer Deliverable 2

2 Overview of SDH architectures


The chapter starts by defining the concept of network architecture. Then, the layered
network structure of the SDH is presented and the classes of SDH network
architectures originating from the use of different protection techniques are
indentified. In the following two sections of the chapter, these architectures are looked
at in more detail.

2.1 What is a network architecture?


A network architecture is an abstract description of a networks functionalities and
configuration. Network functionalities that are usually used for distinguishing network
architectures are traffic routing and traffic protection functionalities. The configuration
of the network depends on the number of nodes and their interconnections.
An architecture potentially consists of a vertical and a horizontal description. The
vertical description uses the concept of network layers to identify and represent the
different network functionalities and their relationships. The horizontal description
uses the concept of network domains (subnetworks) to identify and represent
administrative/physical partitions within a network layer. The following figure
illustrates the concepts of network layers and domains.
Subnetworks Links

Specific path
layer network

Specific path
layer network

Transmission media
layer network

Access group A layer network

T1304500-95/d06

Layering view Partitioning view


(client/server association)

(a) Layering concept (b) Partitioning concept

Figure 1. Orthogonal views of layering and partitioning


It should be noted that each layer can have its independent topology. At the physical
level the topology coincides with the physical disposition of the cables implementing

page 2 (62) 1999 EURESCOM Participants in Project P709


Deliverable 2 Volume 5: Annex D Interaction between optical and clinet layer

the network. However, the topologies at other layers will be virtual (or logical)
topologies, defined by the interconnectivity between the network nodes.
Network architectures are formally represented by functional models. Functional
models follow the definitions and rules presented in the following ITU-T
Recommendations :
G.805 Generic Functional Architecture of Transport Networks (General
rules for all kinds of networks)
G.803 Architectures of Transport Networks based on the Synchronous
Digital Hierarchy (SDH) (Application of G.805 to SDH networks)
Draft G.otn Architecture of Optical Transport Networks (Application of
G.805 to optical networks)
In a functional model we can find the different network functionalities allocated to
different network layers. The layers contain transport processing functions, and
interact vertically according to a client/server relationship. Within each network layer
there may be different subnetworks interconnected by links in a horizontal
relationship. The following figure presents an illustrative functional model.
Trail
AP AP

Network connection Trail Client


Trail layer
termination termination
network
CP
SNC Link connection
TCP TCP

Client to
Client to server
server adaptation
adaptation
Trail
AP AP

Server
Trail Trail layer
termination termination network

SNC LC LC LC SNC

TCP CP CP CP CP TCP
T1304480-95/d04

Figure 2. Example of a functional model


The functional model of a network is especially important from a network
management point of view, since it is especially adapted to the TMN model. It is also
important to provide an abstract description of the network functionalities, which can
then be implemented in different ways using physical equipment.

1999 EURESCOM Participants in Project P709 page 3 (62)


Volume 5: Annex D Interaction between optical and clinet layer Deliverable 2

2.2 SDH network layers


In order to understand SDH network architectures we must be acquainted with its
network layer model. This model is represented in the Figure 3, consisting of three
main layers : the circuit, path and transmission media layers. The layers can be divided
in sub-layers, as shown in the figure.
At the circuit layer we can have 2 Mb/s, 34 Mb/s or 140 Mb/s circuits, which are
mapped into Virtual Containers (VC) in the path layers (ex: 2 Mb/s into VC-12, 34
Mb/s into VC-3 and 140 Mb/s into VC-4). The path layer is split in two sub-layers,
since some VCs of lower capacity can be multiplexed in higher capacity VCs. This
results in lower- and higher-order paths.
The transmission media layer can be divided in a physical media layer, which depends
on the kind of transmission media used (optical fibre, coaxial cable, radio link) and a
section layer, associated to the netwok elements responsible for transmission: the
multiplexers and regenerators.
Circuit
Circuit Layer Networks Layer

Lower Order
VC-11 VC12 VC-2 VC-3 Path Layer

Path
Layer

VC-3 VC-4 Higher Order


Path Layer

Multiplex Section Layer Section


Layer

Regenerator Section Layer Transmission


Media
Layer
Physical Media Layer

Figure 3. Layered network model of the SDH


Each layer is supervised for its quality of service and must be managed in an
integrated way with the rest of the layers. As such, each layer associates overhead
information to the signals received from its client layer, which allows the
implementation of supervision and control functionalities.

2.3 Classes of SDH network architectures


SDH network architectures will be treated in this chapter according to the kind of
protection technique used. There are two classes of protection techniques:
SDH trail protection (at the section or path layer)

page 4 (62) 1999 EURESCOM Participants in Project P709


Deliverable 2 Volume 5: Annex D Interaction between optical and clinet layer

SDH subnetwork connection protection (at any path layer, on any physical
structure)
Trail protection is used to protect Multiplex Section Trails or VC Trails. It can be
applied to rings or point-to-point links. MS protection protects against failures in
specific Multiplex Sections, thus protecting all VCs simultaneously. VC trail
protection, on the other hand, is an end-to-end protection technique, and can be
applied to individually chosen VCs.
Subnetwork Connection Protection can be used to protect a portion of a path between
two Connection Points (CP) or between a CP and a Termination Connection Point
(TCP) or the full end-to-end path between two TCPs. It differs from VC trail
protection in the technique used to monitor the quality of the protected signal.
Contrary to trail protection techniques, SNC protection does not have inherent signal
quality monitoring capabilities, which results in the need to use specific monitoring
techniques. SNC protection schemes are classified according to the monitoring
technique adopted.
Protection techniques can also be bidirectional or unidirectional, as well as revertive
or non-revertive. Bidirectional protection switches both directions of traffic to
protection channels, even if only one traffic direction is affected by a failure.
Unidirectional protection switches only the affected traffic direction to protection
channels in case of a unidirectional failure. Revertive protection switches the traffic
back to the normal channels after correction of a failure. Non-revertive protection
keeps traffic running on protection channels after correction of a failure.
In the following sections we will analyse in more detail the different SDH protection
architectures.

2.4 Trail protection

2.4.1 Linear VC trail protection

Linear VC trail protection is a dedicated protection technique, which means that an


appropriate amount of capacity is available, over a separate physical route, for
protection of the normal traffic. This technique can be used on rings and meshes,
either on the LO or HO path layers. It is applied to VCs individually (not all VCs need
to be protected) and protects traffic on an end-to-end basis. Linear VC trail protection
can operate unidirectionally or bidirectionally. In the later case, extra traffic can be
carried on the protection path.

2.4.2 Linear Multiplex Section protection

Linear MS protection can be dedicated or shared. In the shared case, an MS trail with
protection capacity is available to protect one of a set of working MS trails. Linear MS
protection allows the simultaneous protection of all VCs in an MS trail in a point-to-
point link. This kind of protection technique can also be unidirectional or
bidirectional, allowing extra traffic to be carried over the protection trail in the
bidirectional case.

2.4.3 MS Dedicated Protection Rings

A basic MS Dedicated Protection Ring (MS-DPRings) consists of two fibres


supporting counter-rotating traffic. Only one fibre supports working traffic, the other

1999 EURESCOM Participants in Project P709 page 5 (62)


Volume 5: Annex D Interaction between optical and clinet layer Deliverable 2

being dedicated to protection. MS-DPRings require an Automatic Protection


Switching (APS) Protocol to operate. However, the detailed operation of these rings is
still under study by ITU-T.
In the event of a failure (span or node failure), the node immediately before the failure
re-routes the traffic from the working fibre to the protection fibre. The traffic then
circulates around the ring, in the opposite direction over the protection fibre, until
reaching the node immediately after the failure. This node routes the traffic back to the
working fibre. Thus, all traffic is restored except for the case of a node failure, where
traffic to/from the failed node is lost.

A B C A B C

F E D F E D

Figure 4. a) Schematic of a 6 node SDH ring network which employs two


unidirectional fibres. One of the fibres supports working traffic (grey) and the
other supports dedicated protection channels (white). Bi-directional traffic
(diverse routes) is shown between nodes A and D under normal operating
conditions; b) The same ring and traffic, but now with a cable cut type failure
between nodes B and C (one multiplex section). Traffic is ring-switched on to
(and off) the dedicated protection fibre by the nodes adjacent to the cut

2.4.4 MS Shared Protection Rings

MS Shared Protection Rings can have two or four fibres. The two-fibre version will be
described here. Each of the two fibres (one carrying clockwise and the other counter-
clockwise traffic) has a capacity which is divided equally between working channels
(time-slots) and protection channels (time-slots). The working channels on one fibre
are associated with the protection channels on the other fibre and vice versa. Only one
set of section overheads are carried on each fibre. In a two fibre scheme, where the
ring is rated at STM-16, each span can carry eight AU-4s of priority (protected) traffic
on the working channels. Under normal conditions the protection channels can be used
to carry additional low-priority traffic (e.g. PSTN) which is dumped when protection
capacity is required. Traffic will normally be routed on the most direct route to its
destination node, although sometimes the longer routing will be preferred due to
capacity limitations and in order to balance the traffic load on the ring. Connections
are made through SDH ADMs at the path layer level according to the destination
node.
In the event of a span failure, priority traffic is switched from the working channels of
one fibre to the protection channels of the other fibre. This operation, conducted at the
node immediately preceding the broken span, causes the traffic to change direction
and go the long way around the ring until it reaches the node on the other side of the
broken span. The traffic is then switched back on to the working channels of the
original fibre where it continues to its destination node. Bidirectional switching and an

page 6 (62) 1999 EURESCOM Participants in Project P709


Deliverable 2 Volume 5: Annex D Interaction between optical and clinet layer

APS protocol are essential for the proper operation of the protection scheme in an MS-
SPRing.
Figure 5a depicts a 6 node ring under normal conditions with bi-directional traffic
between nodes A and D while Figure 5b shows the same traffic under span failure
conditions.
Note that multiplex section shared protection can also be used to protect against
intermediate node failures. In the case of multiple cable cuts, the ring splits into two or
more segments, and service can be maintained between any two nodes providing they
are in the same segment.

A B C A B C

F E D F E D

Figure 5. a) Schematic of a 6 node SDH ring network which employs two


unidirectional fibres. The capacity of each fibre is divided between working
(grey) and protection (white) channels. Bi-directional traffic is shown between
nodes A and D under normal operating conditions; b) The same ring and traffic,
but now with a cable cut type failure between nodes B and C (one multiplex
section). Traffic is ring-switched on to (and off) the protection channels at the
nodes adjacent to the cut

2.5 Subnetwork connection protection

2.5.1 General characteristics of SNC protection

Subnetwork Connection Protection is a linear protection technique that can be used to


protect a portion of a path between two Connection Points (CP), or between a CP and
a Termination Connection Point (TCP,) or the full end-to-end path between two TCPs.
Not all VCs in a MS have to be protected by this technique, as it can be applied to
individual VCs. SNC is typically used to protect segments of a VC trail that traverses
several administrative domains, as for example when two nodes in belonging to
different network operators are linked by a VC trail. The following figure illustrates
this case:

1999 EURESCOM Participants in Project P709 page 7 (62)


Volume 5: Annex D Interaction between optical and clinet layer Deliverable 2

Administrative borders
between operators

VC trail

SNC Protection SNC Protection SNC Protection

Figure 6. Use of SNC protection to protect a VC trail on a segment by segment


basis
SNC protection can be applied to rings, meshes or mixed topologies. Both uniform or
diverse routing of traffic is allowed. Only unidirectional protection switching is
allowed so far, since bidirectional switching would require a dedicated overhead,
which is not available, for supporting an APS protocol. SNC protection also supports
reversive and non-reversive modes of operation.

2.5.2 SNC protection with inherent or non-intrusive monitoring

Contrary to trail protection techniques, SNC protection does not have inherent signal
quality monitoring capabilities, which results in the need to use specific monitoring
techniques. SNC protection schemes are classified according to the monitoring
technique adopted. Currently, there are two kinds of monitoring techniques for SNC
protection described in [1]: inherent monitoring (SNC/I) and non-intrusive monitoring
(SNC/N)
In SNC/I the protection switching criteria are derived from the server layer
information. For example, a Server Signal Fail (SSF) indication can be used by the
SNC protection system to initiate a protection switching action. Thus, the failure
detection is performed by the server layer and the protection switching is performed
by the client layer. This kind of protection protects traffic against failures in the server
layer. If SNC/I is used to protect a VC-4, the server layer is the MS layer. If the path
to protect is a VC-12, the server layer is a HOP (e.g.: VC-4).
SNC/N implements a trail termination in the path layer itself. This trail termination is
only able to read the path overhead to monitor the path status, not being able to alter
in any way the bytes in this overhead. Thus the name of the technique. SNC/N is able
to protect the path against failures in the server layer, and failures and degradations in
the client layer. The failure detection and the protection switching are performed bu
the client layer.
The comparison of the different SDH protection architectures allows the following
conclusions to be drawn:
For uniform or adjacent node traffic paterns the MS-SPRing provides more
capacity than MS-DPRings or path protected rings.
For hubbed traffic paterns the most appropriate architectures are the MS-DPRing
and the path protection.

page 8 (62) 1999 EURESCOM Participants in Project P709


Deliverable 2 Volume 5: Annex D Interaction between optical and clinet layer

Path protection can be applied to all transmission rate (STM-n, with n=1,4,16)
and can be used with all topologies where two physically independent trails are
available.
The APS protocol of MS-SPRings is optimised for AU-4 operation (in Europe).
Therefore, add & drop should preferentially be performed at the VC-4 level.
MS-SPRings are only advantageous, when compared to MS-DPRings, when
working at the STM-16 level, because at the STM-4 level the number of available
AU-4s is not enough to provide real benefits.
Two-fibre MS-SPRings cannot operate at the STM-1 level.
After this overview of SDH architectures, the second section will present an overview
of optical architectures.

1999 EURESCOM Participants in Project P709 page 9 (62)


Volume 5: Annex D Interaction between optical and clinet layer Deliverable 2

3 Overview of optical architectures


The chapter starts by defining the concept of optical architecture. Then, a
classification method for architectures, based on the allocation of the main network
functionalities to network layers, is proposed. In the following sections of the chapter,
the optical architectures selected by P709, as well as OAM aspects of the optical
layers are looked at in more detail.

3.1 What is an optical network architecture?


An optical network architecture is an abstract description of a network where some or
all functionalities are allocated to the optical layers of the layered model. An optical
architecture inserts three new network layers between the regenerator section layer of
the SDH layered model and the physical media layer (optical fibre): the Optical
Channel (OC) Layer, the Optical Multiplex Section (OMS) Layer and the Optical
Transmission Section (OTS) Layer. Presently, it is expected that an optical network
architecture supports several types of client technologies in the upper layers. The
resulting layered model would look as in the following picture:

IP
ATM
SDH PDH
WDM
Figure 7. Interrelation of client layers and the WDM layer

3.2 Classes of architectures


Many different functionalities can possibly be considered as criteria for classifying
architectures. Here, only the traffic routing and traffic protection functionalities of the
network will be taken into consideration.
Traffic routing in a network is directly related to the networks physical and/or logical
topologies. This functionality is carried-out by Cross-Connects (XCs) or Add-Drop
Multiplexers (ADMs), and can be performed electrically or optically. More
specifically, routing can be done in the following network layers :
SDH Low Order VC or SDH High Order VC layers
Optical Channel layer using optical switch or optical cross-connect
Advanced networks supporting high volumes of traffic must ensure that network
outages, with the associated traffic and revenue losses, are minimised. To achieve this,
restoration and/or automatic protection techniques are implemented in the network.
Restoration consists of re-routing traffic from damaged links via any available
capacity in a different route to the same destination. It is usually performed manually
or with the intervention of a centralised system. Protection is built into the equipment,
relies on pre-assigned capacity in diversely routed links and is performed

page 10 (62) 1999 EURESCOM Participants in Project P709


Deliverable 2 Volume 5: Annex D Interaction between optical and clinet layer

automatically by the equipment whenever a network damage occurs. Modern


networks have already implemented electrical protection techniques. With the
development of advanced optical technology, network protection can also be
implemented optically. As such, protection can be achieved in the following network
layers :
SDH Low Order VC or SDH High Order VC (path protection)
SDH Multiplex Section layer (MSP: Multiplex Section Protection)
Optical Channel layer using optical switch or optical cross-connect
Optical Multiplex Section layer using optical switch
Considering that the above mentioned network functionalities can be implemented in
the electrical or optical layers of the architecture, independently of the network
topology, we have four classes of network architectures:
1. Electrical Routing/Electrical Protection Switching
2. Optical Routing/Electrical Protection Switching
3. Electrical Routing/Optical Protection Switching
4. Optical Routing/Optical Protection Switching
According to the work of previous EURESCOM Projects P413 and P615, these four
classes of architectures represent a logical evolutionary path from todays technology
(Class 1, SDH-based) to future all-optical technology (Class 4), with intermediate
steps in Classes 2 and 3. This evolution is illustrated graphically in the following
figure.
2. Electrical Protection
and
SDH protection Optical Routing
available (e.g.: Coloured Section Rings)

4. Optical Protection
1. Electrical Protection and
and Optical Routing
Electrical Routing (e.g.: all-optical network)
(e.g.: present-day SDH systems)
3. Optical Protection
and
SDH protection Electrical Routing
not available (e.g.: optically protected ring)

Present time

Figure 8. Time scale for the introduction of optical functionalities


It should be noted that in general, in an optically routed network, routing will be partly
done in the electrical layers, by the SDH equipment (finer granularity), and partly in
the optical layers, by fixed or re-arrangeable wavelength routing (coarser granularity).
Therefore, classifying an architecture as optically routed puts the emphasis on the
additional degree of freedom allowed by routing traffic in the optical domain. On the
other hand, a basic difference between optical and electrical routing should be pointed
out: electrical routing is implemented logically (using routing tables in the SDH
equipment), while optical routing is really a physical operation.

1999 EURESCOM Participants in Project P709 page 11 (62)


Volume 5: Annex D Interaction between optical and clinet layer Deliverable 2

3.3 Capabilities and limitations of optical network architectures


The following conclusions were reached by EURESCOM Project P615 based on a
functional comparison of optical, SDH and mixed architectures.
Optical routing allows an additional degree of freedom (flexibility) for routing
traffic in a network.
The routing connectivity in an optical network is more limited than in SDH when
wavelength conversion is excluded. Fortunately, this does not decrease the
capacity efficiency in WDM ring networks.
When an optical network carries SDH traffic, then optical channel (OCH)
protection is functionally equivalent to existing SDH protection. Therefore, there
is little interest for OCH protection right now.
Optical multiplex section (OMS) protection, having simultaneous protection of
all wavelengths, can replace the SDH protection systems of all the channels,
providing potentially a huge reduction of equipment and cost reduction in
exchange for the reduced level of protection.

3.4 Architectures selected in P709


Project P709 in its Deliverable 1 selected a number of optical network architectures
for further study in the other Deliverables. The selected architectures were the
following :
Coloured Section Ring (CS-Ring)
Optical Multiplex Section Shared Protection Ring (OMS-SPRing)
MWTN Optical Cross Connect-based Mesh (MWTN Mesh)
In the following sections, this chapter will present detailed information on these three
architectures. The information is taken from Annex A of EURESCOM Project P615s
Deliverable 1. For each architecture the information presented includes a general
description, a functional model, proposed implementations and design rules.

3.5 Coloured Section Ring

3.5.1 General

fibres: 2
physical topology: ring (max. number of nodes limited by number of
wavelengths and by SDH protection mechanism)
logical topology: mesh
routing: SDH path layer followed by optical channle layer
protection: Linear MS (1+1 MSP assumed)
span failure: yes
multiple span failure: only if both spans are in the same SDH MS
intermediate SDH node failure: no if node is used for o/e/o redirection

page 12 (62) 1999 EURESCOM Participants in Project P709


Deliverable 2 Volume 5: Annex D Interaction between optical and clinet layer

intermediate opt. node failure: yes (looks like span failure to SDH nodes)

3.5.2 Description

The Coloured Section Ring, invented and patented by France Telecom, employs
wavelength division multiplexing to improve the capacity of standard SDH two-fibre
rings. This architecture is an example of optical routing and electrical protection.
The standard SDH node (an SDH ADM) is supplemented by a two-channel OADM or
two single-channel OADMs as shown in Figure 9. The OADMs do not possess any
cross-connect functionality. Each node on the ring (see Figure 10a) is directly
associated with two other nodes using two unique wavelengths. Any other nodes on
the ring can be reached by going through either of these two directly linked nodes.
They in turn will be linked to other nodes, etc. Note that Figure 10a) depicts
wavelengths linking adjacent nodes, however, in order to maximise the capacity of the
ring, it is preferable to have direct links on the busiest routes (e.g. say A to D, and A to
F), and indirect on the quieter routes (i.e. traffic from A to B, C and E would route via
D or F). Traffic undergoes o/e/o wavelength conversion at the directly-linked nodes
and at any other intermediate nodes preceding the destination node.

east SDH ADM west

OADM 1 OADM 2

Figure 9. Schematic of add & drop node used in the coloured section ring.
Working traffic is shown by the thick lines and protection traffic is shown by the
thin lines, however, their capacities are identical. Here, the OADM functionality
has been broken into two separate single channel OADMs, but could also be
fulfilled by a 2-channel OADM. The OADMs by themselves have no cross-
connect functionality, but through signals are switched to different wavelengths
by the SDH ADM
The size of the ring in terms of number of nodes is directly limited by the number of
wavelengths available (one wavelength per node), therefore a sixteen wavelength
system can support 16 (or fewer) nodes. An additional restriction may occur
depending upon the particular SDH protection mechanism employed.
With this basic scheme, the maximum capacity that can be allocated between any two
nodes on the ring is limited to two optical channels (two wavelengths), say 2 x 2.5
Gbit/s for an STM-16 system. Of course, this extreme situation would prohibit this
particular pair of nodes communicating with any other nodes on the ring, and in
general we would limit the capacity between any two nodes to a level much less than

1999 EURESCOM Participants in Project P709 page 13 (62)


Volume 5: Annex D Interaction between optical and clinet layer Deliverable 2

two channels. However, if additional capacity is needed at one site, there is no reason
why we cannot site several optical nodes at the same geographic site.
Traffic entering the ring at an SDH ADM is connected through a particular interface
(east or west in a simple two-port ADM) according to the destination address. At this
point, the signal is allocated to a wavelength which ensures it is only connected to one
of the other SDH ADMs on the ring. If this node is not the final destination, traffic is
connected through to the other optical interface of the SDH ADM and is reallocated to
another wavelength. This process repeats until a connection is made between the
originating and destination nodes. Note that although optical routing is used to bypass
many of the en-route SDH ADMs, SDH path layer routing is used at the other
intermediate nodes where opical/eletrical/optical wavelength conversion occurs.

3.5.3 Protection

The coloured section ring uses linear MSP to protect traffic. Dedicated protection
(1+1) with single-ended switching has been assumed for the effects of this discussion.
In the event of a span failure, traffic is protected by duplicate traffic which is sent the
long way around the ring on the same wavelength (the protection route), see Figure
10b. Therefore the SDH ADMs require duplicate optical interfaces for both directions
(east & west) and switches on the receive side to allow connection either to the
working or protection routes.
In general the scheme cannot protect against multiple span failures, but where one
wavelength is used to cross several spans and the second span failure occurs on the
same multiplex section as the first span failure, then traffic will remain protected. The
coloured section ring can also be affected by intermediate node failures, e.g. nodes
where wavelength conversion occurs using the SDH ADM. These failures cannot be
protected against by the multiplex section protection. At other intermediate nodes,
traffic would not be affected by failure of the SDH ADM, but would be affected by
complete OADM failure. However, this latter type of failure may be protected by the
multiplex section protection scheme.

SDH ADM

OADM

A B C A B C

F E D F E D

Figure 10. a) Coloured Section Ring under normal operating conditions.


Protection traffic is not shown, b) Coloured Section Ring showing protection
mechanism. Protection traffic for the broken span is shown by thin green line

page 14 (62) 1999 EURESCOM Participants in Project P709


Deliverable 2 Volume 5: Annex D Interaction between optical and clinet layer

3.5.4 Functional model

The basic functional architecture of the CS ring node is illustrated in Figure 11. The
architecture of the 1+1 MSP is based on information provided in ITU-T Standards
G.783 and G.841. The MSPC ellipse represents the switch required to receive traffic,
and also the permanent connections to send duplicate traffic for protection purposes.

HP

MSPC

MS

RS

OC

OMS

WEST OTS EAST

Figure 11. Functional architecture of a coloured section ring node showing the
duplicate protection circuit (shaded), and the MSP switch on the receive side
The MSPC switch (receive side) is activated by a Signal Degrade or Signal Failure
alarm which is normally detected in the trail termination function in the Multiplex
Section layer (lower termination function in the MS layer, below the MSPC). An APS
protocol is not used unless dual ended switching is required (i.e. 1:1 protection instead
of 1+1, see G.841).

3.5.5 Implementation

The Coloured Section Ring associates optical routing on a 2-fibre ring with linear
SDH multiplex section protection. It is an example of an architecture with optical
routing and electrical protection.
This architecture is based on an SDH ring where ADMs are optically interconnected.
The basic principle is to associate a wavelength with each SDH Multiplex Section
(between two ADMs regardless of their physical location in the ring), using Multiplex
Section Protection. Protection is done in the SDH layers using linear MS protection.
Working and Protection Multiplex Sections are diversely routed on the ring utilising
the same wavelength on different fibres (Figure 12).
The benefits of CS-Ring architecture rely on an increased transmission capacity by
implementing WDM and on a new level of flexibility to define a node logical order to
reduce the transit traffic through the nodes.

1999 EURESCOM Participants in Project P709 page 15 (62)


Volume 5: Annex D Interaction between optical and clinet layer Deliverable 2

Working Section 1

Node 2
Node 1
Protection Section 2
Section 1

Section 4

Node 4 Node 3
Section 3
OADM

Figure 12. Scheme of a four node Coloured Section Ring


Wavelength routing is used in fixed filtering configuration and wavelength reuse is
adopted in order to reduce the number of implemented wavelengths. The functionality
of the OADM is described in Figure 13. In a CS-Ring, two wavelengths are assigned
in each node. The most efficient and cost effective technology for a two wavelength
add-drop device in the nm range channel spacing is the multilayer technology.

Drop n* i Add

N* i
N* i
Figure 13. Functionality of an Optical Add Drop Multiplexer with spectral reuse
Table I shows the insertion loss at the different ports of an OADM with two drop or
add wavelengths in case of fixed filtering and spectral reuse.
Number of add or drop wavelengths input - output input - drop port ; add port - output
2 wavelengths 2 dB 1 : 1.5 dB ; 2 dB
2 : 2 dB ; 1.5 dB

Table I - Insertion loss of a two wavelength multilayer OADM without


connectors
To implement a CS-Ring, an optical layer is needed for wavelength routing. On a
standard ADM basis, two extra line boards fitted with MS dedicated wavelength are
implemented for protection in each ADM. The standard boards are replaced by
dedicated wavelength short distance line boards for working state. Figure 14 shows a
detailed implementation on the optical layer in node 2 assuming splices between the
filters. Optical connectors are used at each pigtail of the drop or add port and the
input/output fibre section. The insertion loss for a wavelength going through the

page 16 (62) 1999 EURESCOM Participants in Project P709


Deliverable 2 Volume 5: Annex D Interaction between optical and clinet layer

optical layer of a node is 2 dB and the insertion loss of working signal drop/add port is
1.5 dB (2 dB for protection signals).

OADM ; InsertionLoss : 2 dB Optical Connector


Multilayer Filter
Figure 14. Detailed implementation of filters at node 2 of a CS-RING
In terms of ring fibre span, the size of the CS-Ring architecture is limited by the
protection link power budget. To overcome the standard 26 or 28 dB power budget,
optical amplification can be used according to the needs. In order to amplify only the
protection links, the implementation of optical amplification in the middle of the node
optical layer is suggested according to Figure 15. Note that the position of the filters
used for protection signals is different from the previous case. For architecture
comparisons, we assume that the optical layer has no chromatic, no polarisation effect
and optical crosstalk gives no penalty on transmission.

OA

Figure 15. Detailed implementation of filters with optical amplification


As the optical layer is not reconfigurable, supervisory wavelength is not needed as far
as there are no in line amplifiers. An embedded channel in the SDH overhead can be
used to transmit supervisory informations to the manager.

3.5.6 Design rules

General contents of the design rules


In general the design rules should contain information of the physical limitations of
the architectures to be studied. The physical limitations (architecture by architecture)
can be described with the following specifications :
maximum number of nodes that can be included into the given architecture

1999 EURESCOM Participants in Project P709 page 17 (62)


Volume 5: Annex D Interaction between optical and clinet layer Deliverable 2

maximum length of a span between two neighbouring nodes without and with
optical amplification (OA) implemented in the nodes.
maximum length of the protection path
On the other hand there are not only physical but different kinds of capacity
limitations that can be identified in an architecture. These capacity limitations can
depend on the applied technology or on the architecture itself, thus it should be
specified architecture by architecture.
The capacity limitations can be described with the following specifications:
Maximum number of available wavelengths in the architecture (if it is limited
only by the current state of the technology and not the given implementation of
the architecture, it should be noticed so)
Maximum number of the available wavelengths in one node (if it is limited only
by the current state of the technology and not the given implementation of the
architecture, it should be noticed so)
Maximum capacity of an optical channel (i.e. maximum capacity that can be
carried by one wavelength)
The physical limitations and the capacity limitations are studied in this chapter. The
maximum number of available wavelengths in each architecture and per node is
specified.
The general assumptions adopted to calculate the ring maximum length are :
fibre attenuation : 0.28 dB/km (according to G.692)
an optical connector at each end of a fibre section : 0.3 dB mean loss and 0.1 dB
standard deviation. As many connectors are used in a ring, a statistical basis is
adopted (ETSI recommendation M 1009) assuming 3 times the standard deviation
to calculate the related insertion losses. In the Tables, all results are rounded at
the upper integer number of dB.
Using OA at each node, two extra connectors are considered to enable the
implementation of the amplifier board.
A 28 dB optical budget is considered enabling transmission rate up to STM-16
with joint engineering. With OA in each node, the protection span can be
considered as a multi-wavelength link with in-line amplifiers for the related
wavelengths. To evaluate the span length, we considered an attenuation range
between the OAs from 22 dB to 33 dB. According to commercial products, a 33
dB attenuation range is feasible with only one in-line amplifier and 22 dB
attenuation range is feasible with typically 6 in-line OAs and 8 wavelengths.
The span lengths with OA reported in Table II and Table III are rather pessimistic
because they are based on G 692 recommendation and commercial system capabilities
using +13 dBm or +15 dBm output level Optical Amplifiers.
Physical limitations
In CS-Rings, the primary parameter is the number of nodes that is equal to the number
of wavelengths. This number is limited either by the maximum number of nodes in an
SDH ring (16) or the maximum number of wavelengths in a wavelength multiplexer
allowed by the current state-of-the-art technology; we assume a node implementation
according to Figure 16.

page 18 (62) 1999 EURESCOM Participants in Project P709


Deliverable 2 Volume 5: Annex D Interaction between optical and clinet layer

OADM Fibre 1 OADM


OADM Fibre 2 OADM
OADM ; InsertionLoss : 2 dB Optical Connector Optical Amplifier
Figure 16. Node configuration of CS-Ring without and with Optical amplification
The working path is always shorter than the protection path, therefore the latter should
be considered in calculation of the maximum span lengths.
This is shown in the Table II; the result is rounded.
Number of nodes 3 4 5 6 7 8 9 10 to 16
Max. span length without OA. 37 21 14 9 5.9 3.8 2.2 no
longest protection path without OA 75 64 55 46 35 26 17 no
Max. span length with OA at each node 107 92 89 82 77 76 75 for further
study

Table II - Physical limitations of a CS-Ring (length in km)


Table II is filled up to 9 nodes which is the maximum number of nodes achievable
without OA. with the defined parameters and the physical implementation depicted in
Figure 15. With OA in a 9 node CS-Ring, 8 OA are implemented in a protection path.
Besides undersea transmission, no commercial equipments are available for a link
longer than around 650 km including more than 6 OA. The length limitation is due to
the maximum level of chromatic dispersion (12800 ps) compatible with STM-16
transmission.
Capacity limitations
The number of wavelengths in a CS-Ring is identical to the number of nodes. For the
time being, the size of a CS-Ring with OA including more than 9 nodes requires
further studies because this implementation implies more than 9 wavelengths which is
commercially available but we need also links with more than 8 in-line amplifiers.
This implies an accumulation of spontaneous emission and a degradation of SNR
which must be calculated.
Regarding the flexibility of wavelength routing to define a logical order of the nodes,
we must take into account the physical implementation to know how many nodes are
really crossed by the protection path. Using this flexibility, the protection path length
may be reduced in some cases and thus the span length.
According to Table II results, the CS-Ring full span with OA is in the range 320 km (3
nodes) to 675 km (9 nodes) which is the maximum length of a STM-16 path on G 652
fibre due to chromatic dispersion. Larger size CS-Ring will imply the use of a reduced
chromatic dispersion fibre or chromatic compensation.
For ring comparisons in Deliverable No. 2 of EURESCOM P615, the span length of
each span is 10 km. According to the results in Table II, no optical amplification is
needed in a 5 node CS-Ring and 2 nodes must be fitted with OA in a 8 node CS-Ring.

1999 EURESCOM Participants in Project P709 page 19 (62)


Volume 5: Annex D Interaction between optical and clinet layer Deliverable 2

3.6 Optical Multiplex Section Shared Protection Ring

3.6.1 General

fibres: 2
physical topology: ring
logical topology: mesh
routing: optical and SDH path layer
protection: optical multiplex section shared protection
(dual-ended switching)
span failure: yes
multiple span failure: yes
node failure: yes

3.6.2 Description

Protection at the optical multiplex level could be beneficial. It would enable protection
of all optical channels simultaneously, with a low number of optical switches, giving it
a cost advantage over electrical protection. Of course, such an advanced functionality
requires a proper management system. Although OMS-protection does not protect
against defects in the network layers, like multiplexing / demultiplexing, it protects
against cable-cuts, which is an increasingly important issue for public network
operators.

a) b)

A B C A B C

E D E D

1, 3, 5, ...
2, 4, 6, ... Fibre failure

Figure 17. Two-fibre OMS-shared protection ring architecture. a) working state,


b) protection state
This section covers one specific example, namely, the Two-Fibre Optical Multiplex
Section - Shared Protection Ring (OMS-SPRing). In this architecture, the capacity on
each of the two fibres is shared1 by working and protection traffic. For example, a
sixteen-wavelength WDM system would have eight channels allocated for working
traffic and eight for protection traffic on each of the two fibres. In the event of a
failure, working traffic on one fibre would be switched over to the protection capacity

1 An alternative to shared protection is dedicated protection when a dedicated fibre is used for
protection traffic only.

page 20 (62) 1999 EURESCOM Participants in Project P709


Deliverable 2 Volume 5: Annex D Interaction between optical and clinet layer

on the other fibre. Therefore the eight working channels on one fibre have the same
wavelengths as the eight protection channels on the other fibre and vice versa. The
operation of this protection mechanism is illustrated in Figure 17.
Additionally, the OMS-SPRing provides optical traffic routing based on wavelength
selection by the nodes.

3.6.3 Functional model

When optical multiplex section protection (i.e. between OADMs) is considered in a


ring architecture, then, realistically, dual-ended switching must be used. Single-ended
switching would result in duplicate traffic generated for each OMS which would
interfere with itself on the long protection routes unless wavelength conversion was
employed at each node. Dual-ended switching should also enable the implementation
of wavelength reuse.
The functional model for this 2-fibre, OMS-SPRing is shown in Figure 18. It should
be noted that this diagram has been simplified between the HP and OC layers, since it
shows only one or two channels being added and dropped, whereas one channel would
be used for each of the other nodes on the ring in order to fully exploit optical routing.

HPC
HP

MS

RS

Shared
OMSP
OC 2 fibres

OMSPC

OMS
WEST EAST
OTS

Figure 18. Two-fibre OMS-SPRing node with cable cut on east side
The functional model shown here is based on the SDH MS-SPRing functional model
depicted in G.841. The standard G.681 implies that the OMS and OTS functions
should handle the same number of wavelengths. This model appears to violate this
restriction in G.681, but it is believed that it circumvents G.681 by using functions
having an identical channel capacity in the OMS and OTS layers, but that the OMS
Adaptation function is only part-populated, e.g. it would be an OMSA_16 function in

1999 EURESCOM Participants in Project P709 page 21 (62)


Volume 5: Annex D Interaction between optical and clinet layer Deliverable 2

the case of a 16 wavelength system, but it would only accommodate working channels
or protection channels, not all 16.
The challenge in EURESCOM Project P615 was to study the functional model of the
node shown in Figure 18, and convert it to a physical picture of a ring network. It
should be noted that switching occurs on both sides of a cable cut (dual-ended
switching).

3.6.4 Implementation

This part deals with the implementation of a two-fibre OMS-SPRing with a meshed
logical node connectivity: each node has a link with all the other nodes on a dedicated
wavelength. The transmission capacity on each of the two fibres is shared between
working and protection traffic. Therefore, in a 7 node ring, 6 wavelengths are needed
per fibre. OADMs are implemented in each node for wavelength routing and dual
ended optical switching enables protection in case of cable cut. Management of the
optical layer is not considered in this chapter. Figure 19 presents a typical
implementation of the optical layer in a node. Wavelength allocation is studied in
Deliverable No. 2 of EURESCOM P615. Opto-mechanical 22 swiches are suited for
protection as far as insertion losses are low (1 dB typical) and switching speed is in the
20 ms range.
The proposed node configuration includes one OADM per fibre and a 22 cross-bar
optical switch implemented at each side of the node. Optical amplifiers can be used.
The best place for the amplifiers seems to be between the OADM and optical switch
in order to use both optical amplifiers in protection state to cope with the extra losses.

OS1
EDFA OS2
OADM

OADM

Figure 19. Configuration of a node with a cross-bar optical switch


An alternative configuration for the node is depicted in Figure 20. It includes one
OADM per fibre and two 21 optical switches.

OS1 EDFA OS2


OADM

OADM
OS4 OS3

Figure 20. Configuration of a node with 21 optical switch

page 22 (62) 1999 EURESCOM Participants in Project P709


Deliverable 2 Volume 5: Annex D Interaction between optical and clinet layer

In both nodes, only half of the fibre capacity is used to transport working traffic, the
other half being reserved for protection.
Regarding the optical amplification, the best location seems to be before the optical
switches, but it should be noted that insertion losses are increased in the protected
state.
In a 5 node architecture, 3 wavelengths are needed in each node for a full logical
mesh. Insertion losses are 2 dB for a wavelength through an OADM. For 5 node
architecture comparisons, we assume 2 dB insertion losses between the add and drop
ports.
Figure 21 and Figure 22 depict the node configuration in working and protection
switched states in order to compare them on the insertion loss basis.

2dB 4dB 3 dB
1dB 1dB
OADM
2dB 2dB

3 dB

2dB 2dB
OADM
2dB
Figure 21. Cross-bar node configuration in a 5 node OMS SP ring (Working state)

2dB
1dB 1dB
OADM
2dB 2dB

Cable Break
2dB 2dB
OADM
6dB 2dB

Figure 22. Cross-bar node configuration in a 5 node OMS SP ring (Protection state)
Multilayer filters are adopted for architecture comparison with 1 dB loss per dropped
channel because they are available and they exhibit the lower insertion losses and
costs than grating devices.

1999 EURESCOM Participants in Project P709 page 23 (62)


Volume 5: Annex D Interaction between optical and clinet layer Deliverable 2

3.6.5 Design rules

The configuration shown in Figure 23 was used to obtain the information presented in
Table III on physical limitations of OMS-SPRings.

OS1 OS2

Figure 23. Schematic implementation of a node in an OMS-SPRing

Number of nodes 4 5 6 7 8 9 to 16
Max span length without OA. 4 2 - - -
longest protection path without OA 16 11 - - - -
Max. span length with OA 92 78 59 51 36 for further study

Table III - Physical limitations of OMS-SPRing (in km)


According to Table III, the ring full span is in the range of 368 km (4 nodes) to 288
km (8 nodes). The reduction in size as the number of nodes is increasing is due to an
increasing number of wavelengths used in the ring. This limitation could be overcome
by the implementation of a more powerful OA used only for protection links. The
OMS-SPRing with 9 to 16 nodes needs further studies by means of software tools to
simulate an optical amplified WDM network.

3.7 MWTN mesh

3.7.1 General

fibres: 2 per link


physical topology: mesh
logical topology: mesh
routing: optical + SDH
protection: OC or SDH path
MS failure: yes
multiple MS failure: yes
node failure: yes

3.7.2 Description

The Multi-Wavelength Transport Network (MWTN) Project, part of the RACE


program, set out to develop ideas for a future broad-band flexible transport network

page 24 (62) 1999 EURESCOM Participants in Project P709


Deliverable 2 Volume 5: Annex D Interaction between optical and clinet layer

employing the optical network layers. MWTN was seen from the outset as a broad-
band network designed to overlay existing networks. The fundamental building block
of the MWTN architecture is the Optical Cross-Connect (OXC) node (Figure 24).
Optical cross-connects are particularly useful for meshed network architectures, where
nodes have to route traffic from three or more directions.
The MWTN OXC node has the ability to redirect (route) traffic according to
wavelength. Inbound traffic is selected first by its incoming route, and second by
wavelength. Optical cross-connects are then used to redirect this traffic onto outgoing
routes. Therefore each optical channel can be selected and redirected as required. In
addition, optical channels may be added or dropped using a conventional digital cross-
connect (DXC).
Traffic is transferred between the OXCs and the DXC via tuneable receivers and
transmitters. The DXC may also be used for fine granularity processing (down to the
VC-12 level), such as grooming, AU routing, and monitoring. The DXC can also be
used to perform wavelength conversion, since all-optical wavelength conversion is not
included in the OXC functionality. Both the OXCs and the DXC can be used to restore
service in the event of a network failure.
TUNEABLE FILTERS OSSs

DXC
TUNEABLE Tx Rx

DROP ADD

Figure 24. Reconfigurable OXC proposed by the MWTN consortium. The node
shown handles traffic from three directions. Both pre- and post-amplification
(triangles) are used to boost signals. Optical channels are selected by tuneable
filters and power equalised (not shown) before cross-connection. Traffic may be
added / dropped or groomed using the DXC in conjunction with the tuneable
transmitters (Tx) and receivers (Rx)
For a WDM system employing n wavelengths, the node requires n OXCs to have full
cross-connect functionality. Each OXC must have an input (and corresponding output)
port for each fibre accessing the node with additional ports for each DXC.

3.7.3 Functional model

The functional model of the MWTN mesh comprises routing functionalities in the
electrical and optical layers. The DXC switch is located above the SDH Multiplex
Section layer but below the Higher-Order Path functions. The OXC function is located
between the OMS and OC functions because the optical channels can be selected
using tuneable filters.

1999 EURESCOM Participants in Project P709 page 25 (62)


Volume 5: Annex D Interaction between optical and clinet layer Deliverable 2

HP

MS

RS

OC

OMS

OTS

Figure 25. Functional model for the MWTN OXC illustrated in Figure 24. OC
protection has not been shown in this diagram, but could be represented by an
expanded OC layer with additional termination and adaptation functions, along
with an OCPC matrix

3.7.4 Implementation

Optical nodes are studied in European Projects as MWTN, OPEN, METON,


PHOTON or MEPHISTO to develop ideas for a future broad band flexible transport
network designed to overlay the existing networks. Optical Cross Connects (OCC) are
particularly useful for meshed network architectures where nodes have to route traffic
from different directions. Taking into account analysis made within EURESCOM
P615 and an overview of the other Projects dealing with OCC, P615s choice was
guided by these functionalities :
a non-blocking reconfigurable node
a reliable configuration (solution including several medium size switch matrices
is preferred)
switching speed
optical properties (insertion loss, crosstalk etc.)
commercial availability
According to these guidelines, P615 proposed two OCC configurations. The first one,
depicted in Figure 26, is reconfigurable, has low losses with no regeneration and no
wavelength conversion. It is based on the first stage for demultiplexing input channels,
the middle stage for switching and the final stage for multiplexing. Optical amplifier is
implemented at the output of the OCC.

page 26 (62) 1999 EURESCOM Participants in Project P709


Deliverable 2 Volume 5: Annex D Interaction between optical and clinet layer

demultiplexer space switch multiplexer O.A.

input fibre 1

input fibre m

from local node Supervisory and management to local node


Figure 26. MWTN node configuration for meshed architecture comparison
If n wavelengths are used per fibre and m fibres are connected at the input of the OCC,
n switches are needed ; each matrix has m input and output ports related to the number
of fibres. An extra matrix is used for dropping or adding channels for the local node.
Assuming three input and output fibres and 8 wavelengths per fibre, the available
optical components are grating type devices with 3 dB typical insertion loss in micro-
optic technology plus 1 dB for addressing wavelength supervisory. An opto-
mechanical matrix has 2 dB insertion loss. For an input - output channel, a typical
insertion loss is 11 dB including 1 dB for optical connectors. Narrowing effect of
cascaded nodes is out of the scope of this chapter. With 16 wavelength per fibre,
typical insertion loss is 12 dB. The OCC depicted in Figure 26 is suited for a small
size meshed architecture with an end to end restoration policy because there is no
wavelength conversion.
One wavelength is dedicated for supervisory in order to get fault, reconfiguration and
protection/restoration management.
In case of a larger size network with subnetwork protection/restoration, the OCC node
configuration depicted in Figure 27 is proposed. This configuration can be taken as a
further step after Figure 26 configuration, taking into account MWTN and OPEN
Projects, but P615 suggested the implementation of the wavelength translation with
electrical conversion because not all optical converters were mature.
In the first part of the OCC, the wavelength is chosen with a tuneable filter then routed
through space switches. Implementation of wavelength converters enables each signal
to reach an output whatever is the input wavelength. Assuming m input - output fibres
and n wavelength in each fibre, the first stage of the OCC is m optical couplers (1n)
followed by a switching stage including n space switches (mm). Wavelength is
selected by tuneable filters (micro-optic Fabry-Perot filters : 2 dB typical insertion loss
including wavelength locking). The final stage is a wavelength converter and an
optical multiplexer. Assuming 4 fibres with 8 wavelengths per fibre, total insertion
loss before detection is around 16 dB. Insertion loss of the final stage is 4 dB before
optical amplification. Assuming 16 wavelengths per fibre, insertion losses are
respectively 19 dB and 4.5 dB.

1999 EURESCOM Participants in Project P709 page 27 (62)


Volume 5: Annex D Interaction between optical and clinet layer Deliverable 2

Space Tunable AO
Switch filtre Wl converter
1 1
m
input fibre 1
n

input fibre m

to local node
from local node
Supervisory and management

Figure 27. Optical Cross Connect with wavelength conversion


Optical Cross Connects are proposed in 2 configurations. The first one (Figure 26)
includes only passive devices with 11/12 dB insertion losses and is suited for small
size meshed networks. The other one (Figure 27) includes wavelength converters with
electrical conversion and the insertion losses are 16 dB before detection and 4 dB after
wavelength conversion.

3.7.5 Design rules

The configuration of the node, adopted for architecture comparison, is defined in


Figure 28. According to the traffic demand considered in Deliverable No. 2 of
EURESCOM P615, the size of the network is defined on a number of fibre basis and
number of wavelength per fibre.

demultiplexer space switch multiplexer O.A.


Supervisory and management

input fibre 1

input fibre 3

8(1/2 sdh LTE)


n*STM-1
Drop Add

Figure 28. MWTN node configuration for meshed architecture comparison

page 28 (62) 1999 EURESCOM Participants in Project P709


Deliverable 2 Volume 5: Annex D Interaction between optical and clinet layer

The maximum fibre span is 100 km. As only 8 wavelengths are used per fibre, P615
assumed the implementation, in each node, of only one Optical Amplifier in each fibre
link. No in-line amplifiers were needed.
The implementation described in Figure 28 exhibits 3 fibres with 4 wavelengths per
fibre There are also add/drop possibilities for local node and the total number of
optical channels crossing the optical cross-connect is 16. The related cost is 16*(the
average cost of one optical channel in an OCC) and we must add the cost of the line
terminal equipments and of the electrical DXC4/4. For a bi-directional link, this cost
must be multiplied by 2. So, for the Optical Cross-Connect, an average cost is defined
as a number of optical channels crossing the OCC.

3.8 Conclusion
This chapter defined what is meant by optical network architecture: a network
architecture that implements some or all network functionalities in the optical layers of
the layered model. Choosing the traffic routing and protection functionalities as
classification criteria, the chapter then proposed a classification method for
architectures, taking into account whether the chosen network functionalities were
allocated to the electronic or optical layers of the functional model.
Then, the network architectures selected by EURESCOM Project P709 were identified
and, for each one, a detailed description was presented. The description included an
overview of the architectures functionalities, a functional model, a discussion of
implementation options and basic design rules.
The optical and electrical architectures are now well known. The next section will
study the problem of dimensioning and some methods to solve it will be proposed.

1999 EURESCOM Participants in Project P709 page 29 (62)


Volume 5: Annex D Interaction between optical and clinet layer Deliverable 2

4 Dimensioning methods

4.1 Dimensioning methods for multiple-ring OMS-SPRing or SDH


SPRing networks
This section considers a way of dimensioning simple networks, which are made-up of
a collection of rings. The rings can be OMS-SPRing optical rings, or SDH MS-
SPRings.
The suggested dimensioning method is based on planar graph theory. It is
unfortunately only applicable to networks where all the traffic generating nodes can be
drawn so that they appear on the edge of the network, however this includes many
simple networks which comprise a few rings. It also gives the minimum possible
number of rings required to carry the traffic, and thus gives an optimal solution.

4.1.1 Types of suitable networks

The used of shared protection rings in telecommunication networks is currently


increasing due to some attractive characteristics of SDH MS-SPRing rings:
The network is built up of small scaleable elements (ADMs) which can be
cheaper than SDH digital cross connects needed for mesh networks.
Shared protection rings are usually more efficient in terms of fibre usage than
1+1 protection in rings.
Although restoration in mesh networks can be more bandwidth-efficient, it is
slower with a state-of-the-art restoration time of one minute. It generally provides
a lower quality of service than MS-SPRing protection where cable breaks cause
less than 50ms of downtime.
Many telecommunication operators are therefore choosing to deploy backbones based
on interconnected MS-SPRings. Routing in MS-SPRing based networks is at the same
time more complex than routing across a single 1+1 protected ring, but is also more
amenable to analytical techniques, since routing a demand involves only choosing the
working path only (the protection path being implicit). With 1+1 protected mesh
networks the protection as well as the working path must be chosen, resulting in
greater complexity. In effect, using MS-SPRing based networks bring us back to the
problem finding sets of unprotected paths across a network, a problem which has
been covered extensively in the literature, with regards not only to
telecommunications.
The principles of the MS-SPRing are also applicable to WDM rings such as the OMS-
SPRing, and we therefore consider in this contribution ways to dimension both optical
and SDH shared protection rings.
An example of a network composed of multiple rings, which is suitable for
dimensioning using the proposed method, is shown in Figure 29. The key
requirements of such a network are that:
The network is planar, i.e. can be drawn on a plane such that no links cross one
another.
All the nodes where traffic originates or terminates are on the exterior
boundary of the network.

page 30 (62) 1999 EURESCOM Participants in Project P709


Deliverable 2 Volume 5: Annex D Interaction between optical and clinet layer

Any nodes which are inside the network boundary must have no traffic
terminating, and must also be even (i.e. the sum of link capacity adjoining that
node must be even). In a network made of rings, this is always the case.
Unfortunately this can rule out many large networks, but if the ring design is
performed carefully, many smaller networks (less than ~20 nodes) meet these
requirements. As shown in Figure 29, rings can be chosen to intersect one another.
Once the geographical placement of rings have been determined, and the ring
capacities (e.g. STM16 MS-SPRing, STM64 MS-SPRing, 16 or 40 OMS-SPRing),
then the number of stacked rings required to carry the given traffic demands must be
calculated. The next section outlines the possible problems in such an approach.

1 3

4
5

Figure 29. Example of a multiple ring network suitable for dimensioning using
the proposed method

4.1.2 Routing problems in multiple-ring networks

Figure 30 shows a demand between two nodes named Start and Finish. A number
of possible routes can be chosen. For example, the Start node sits on both rings 4
and 5, so either ring could carry the first leg of the traffic. Three possible paths are
shown. One uses rings 1,2 and 4, the other two both use rings 1 and 5, but travelling
on opposite sides of the rings. If the demand is larger than 1 VC4 (or wavelength) it
may also be split and several paths be used at the same time.

1999 EURESCOM Participants in Project P709 page 31 (62)


Volume 5: Annex D Interaction between optical and clinet layer Deliverable 2

Finish

1 3

4
5

Start
Figure 30. Possible routes for a given demand
A certain ring usage choice will impact on the in-ring routing, and consequently on the
ring loading and free capacity. Similarly the choice of the gateway node between rings
can impact on the total numbers of rings needed to route all the traffic within the
network.
Faced with all these possible routings and ring choices, it can be difficult to determine
the minimum number of rings required to transport a given set of demands.
The dimensioning method proposed here consists of:
Estimating the number of stacked rings required for each geographical ring
placement
The model then calculates whether this number of rings can carry all the traffic
demands
The number of rings needed can then be increased or reduced until the smallest
possible number required has been found
The proposed method can also be extended to find the routings, but this is not
explained in this contribution.

4.1.3 The dimensioning method

The dimensioning method is based upon the Okamura-Seymour theorem [2], and its
extension due to A. Frank [3].
The Okamura-Seymour theorem states that if G is a planar graph with a set of nodes V
and edges E, and H is a set of k demands {si,ti}, i=1,, k, then if:
All si and ti are on the exterior boundary of the graph

page 32 (62) 1999 EURESCOM Participants in Project P709


Deliverable 2 Volume 5: Annex D Interaction between optical and clinet layer

The graph is even (the sum of the edge capacity adjoining a given node, plus the
sum of the demands originating or terminating at the same node, must be even for
all nodes in V)
Then provided that for any cut which separates the network into 2 connected
subnetworks, the sum of the edge capacity removed to perform the cut must be
superior of equal to the sum of the demands spanning the cut (see Figure 31 for an
example).
To be more precise, let X be a subset of V. Let D(X) be the set of all edges with one
end in X and the other in V-X, and let (X) be the set of all demands with one terminal
in X and one in V-X. If |D(X)| represents the sum of the capacity of all edges in D(X),
and |d(X)| represents the sum of size of all demands in (X), then there is enough
capacity in the network to carry all the demands if and only if:
For all X V, |D(X)| - |(X)| is even and non-negative (1)
It is obvious that this condition is required, but in Reference [2] it is proved that it is
sufficient.
Set X
16 16
(a) (b) 7

16
16 6
16 3 9 4

16
16 16

16 16 3

16 6
Fibre infrastructure with capacity in VC4s Demands in VC4s
(c) (d)
7

16

A cut of the network with capacity Demands across the cut with
D(X) of 32 VC4s (X) bandwidth 32 VC4s

Figure 31. Example of making a cut of a network to check if sufficient


transmission capacity exists. In this case the cut condition is just satisfied, the cut
is tight
Two important conditions limit the applicability of the theorem in telecommunication
networks:
The parity or evenness condition
The requirement that all nodes be on the edge of the network.
The necessity of the parity condition is demonstrated in Figure 32a. It can be seen that
all nodes are on the exterior boundary, the cut criterion is satisfied (there is enough
capacity to transport 2 VC4s both demands across any cut of the network), but the
parity condition is not satisfied therefore the demands cannot be routed.
The necessity of all demands originating or terminating on the outside boundary is
shown Figure 32b: all nodes are even in this case, and the cut criterion is satisfied, but
still the demands cannot be routed.

1999 EURESCOM Participants in Project P709 page 33 (62)


Volume 5: Annex D Interaction between optical and clinet layer Deliverable 2

2 demands
of 1 VC4 each

(a)
(b)

Capacity = 1 VC4
on all edges

Figure 32. Examples demonstrating the necessity of the conditions


The restriction due to the parity criterion can be alleviated in two ways:
By making the network even. In a network consisting of multiplex rings, this can
be done by increasing some demands by up to 1 VC4 (or wavelength) to make all
nodes even. The condition (1) can then be applied to check whether ring capacity
meets demand.
By applying a more complex criterion described in [3]. This extra parity criterion
is based on increasing demands to make the network even, but in an optimal way.
The condition which states that the terminals for the demands lie in the exterior
boundary of the network cannot be removed so easily: routing and dimensioning for
such general networks has been shown to be a NP-complete problem. It is not known
either by how much the required capacity is underestimated by simply satisfying the
cut condition for general network topologies, and this is any interesting study to
perform.
To apply this theorem to our multiple ring network dimensioning method, we first
choose ring placements Ri , i=1,,k. Each Ri is a subset of the edges E used by ring i.
Once ring placements have been chosen, the number of physical stacked rings Ni at
each ring placement i are estimated, as well as the working capacity Ci of the ring (i.e.
8 VC4s for a 2-fibre STM16 MS-SPRing, 16 wavelengths for a 2-fibre 32 OMS-
SPRing). With this data, we calculate the capacity CE of every edge of the network:
For all rings i, and all edges E, CE=CE+Ni Ci if E Ri
With these edge capacities, the condition (1) is verified for all possible cuts of the
network. If the condition holds, then the network capacity provided by the sets of rings
Ri is sufficient to satisfy all the demands. The numbers Ri can then be reduced, until
the minimum number of rings with sufficient capacity has been found.
Note that the algorithm described above does not provide routings for all demands. It
only indicates whether the ring capacity is sufficient to carry all demands, given an
optimised routing. The algorithms leading to the proofs in [2] and [3] can in fact be
extended to provide routings for all demands, see [3] for details.

4.1.4 Application to OMS-SPRing

The method described above is sufficient for dimensioning a network using SDH MS-
SPRing rings, provided time-slot interchange is allowable. Optical rings such as the
OMS-SPRing do not allow wavelength conversion, and in this case the extra
constraints imposed by wavelength continuity have to be satisfied.

page 34 (62) 1999 EURESCOM Participants in Project P709


Deliverable 2 Volume 5: Annex D Interaction between optical and clinet layer

To apply the dimensioning method to OMS-SPRing, the following approach is


suggested :
Firstly, dimension the network (i.e. find the number of rings required) by using
the method described above.
Find the routings of all the demands using the algorithms described in reference
[3].
Once all routings have been determined, break up each route into sub-routes
when they traverse multiple rings, so that each sub-route now only travels across
one ring. This is carried out to reflect the fact that wavelengths will generally be
regenerated or demultiplexed between rings, to groom traffic or combat signal-to-
noise degradation. Wavelength conversion between rings is therefore allowed in
the majority of cases.
Use a wavelength allocation algorithm such as that described in [4] to allocate a
wavelength to each sub-route. We have seen that the wavelength continuity
constraint rarely results in a much greater capacity requirement [5].
A dimensioning method for networks composed of multiple rings of the MS-SPRing
or OMS-SPRing type has been proposed. Although it is restricted to a certain class of
networks, it may nevertheless find wide application for dimensioning networks
composed of a small number of rings, and having up to around 20 node.
The advantage of the proposed method is that it dimensions the network assuming an
optimal routing is used. It therefore gives an optimal answer, where the minimum
number of rings are used.
The method is not applicable to networks using rings with 1+1 protection (e.g.
DPRings), as the routing methods are completely different. Dimensioning such
networks is not a trivial task, (e.g. different routes using different sets of rings may be
chosen for some demands). Algorithms for dimensioning networks consisting of
multiple rings would therefore be of great interest.

4.2 Optical dimensioning methods


This chapter is related to the dimensioning methods in the optical layer. Some
resource allocation algorithms have already been described together with the
description of the restoration and protection techniques studied in Annexes A and B.
Therefore we will focus on the design and dimensioning of optical nodes in meshed
networks, which is one of the major element to be considered in the interaction
between the optical layer and the client layers.
Figure 33 depicts the typical configuration of an optical cross-connect in the optical
layer. As it is described, the local ports of the OXC may be connected to various client
layers (e.g. ATM, SDH STM-n layers). In this chapter, we analyse the possible
structures for the OXC in a multi-client applications context. Several possible
structures are evaluated on the EON network case, showing a compromise in the
optical layer, between the size of optical elements and the characteristics of optical
transmission.

1999 EURESCOM Participants in Project P709 page 35 (62)


Volume 5: Annex D Interaction between optical and clinet layer Deliverable 2

Optical switch matrix

D M
M U
U X
X
Optical
fibres
D M
M U
U X
X

Local
ports
Optical layer

SDH (STM-16) Layer

ATM layer

Figure 33. Optical node with interconnection to several client layers


The possible interdependence of client layers adds complexity to the dimensioning
problem. In Deliverable 1, several architectures have been selected, depending on the
interdependence of the layers (for instance IP/ATM/SDH/WDM compared to
IP/SDH/WDM). In this part of the study, we consider the dimensioning of the optical
layer and we assume other client layers have already been dimensioned. This means
the demands in optical channels have been fixed for each client layer.
In the next paragraph, two structures will be proposed and dimensioning methods for
each structure will be described. An application to example network is given in the
last paragraph.

4.2.1 Optical node structures in a meshed network

Two options have been investigated, on the basis of the possible restoration schemes
mentioned in Annex A, to fulfil network requirements for survivability. The design of
more detailed structures in terms of optical components, as well as the full description
of their technological options, is not addressed here. In particular, the need for optical
amplification in the node is not considered here, since it does not impact on the node
functionality. However, this should be a crucial parameter for techno-economical
studies.

page 36 (62) 1999 EURESCOM Participants in Project P709


Deliverable 2 Volume 5: Annex D Interaction between optical and clinet layer

Switch Switches

fibre

DMUX MUX

Figure 34. OXC structures proposed for path restoration scheme


Structure 1 (on the left) consists of a single switch matrix, connecting all optical
channels passing through the node or dropped to the local ports.
It could be upgraded by a wavelength conversion stage added after the NxN switch.
Structure 2 ( on the right) consists of a layered structure, each single matrix
connecting all same wavelengths passing through the node or connected to local ports.
This structure offers a smaller switch size than the previous structure, thus it may be
more feasible. However the upgrade with wavelength converters is no more allowed.

4.2.2 Dimensioning methods

Considering the possible schemes for resource allocation selected in Annex A, the
following table describes the cases to be used for optical layer.
Node structure 1 Node structure 2
Wavelength conversion (VWP) X
No wavelength conversion (WP) X X
Tunable source X X
Table IV - Possible schemes for resource allocation in optical layer
In order to provide a comparison between the two node structures in terms of
dimensioning, we will apply dimensioning methods with the WP case for resource
allocation.

1999 EURESCOM Participants in Project P709 page 37 (62)


Volume 5: Annex D Interaction between optical and clinet layer Deliverable 2

Node dimensioning can be done as follows:


STRUCTURE 1
We first allocate resource in the network (wavelength channels, fibres) for each case
of failure. Then for each node, all possible cases of (single) failure in the network are
simulated and the size of the node is updated (maximum between current size,
initialised to 0 and the sum of input channels and local channels).
Dimensioning methods is summarised in the following diagram.
Initialise all matrix
sizes to 0

Simulate single
failure (link or
node)

Resource allocation
(wavelength, fiber)

For all nodes

Update matrix size

Figure 35. Dimensioning method for structure 1

page 38 (62) 1999 EURESCOM Participants in Project P709


Deliverable 2 Volume 5: Annex D Interaction between optical and clinet layer

STRUCTURE 2
Each node is described by a set of N matrices where N is the maximum number of
wavelengths to be used in the network. We first allocate resource in the network
(wavelength channels, fibres) for each case of failure. Then for each node, all possible
cases of (single) failure in the network are simulated and the size of each sub-matrix is
updated (maximum between current size, initialised to 0 and the sum of input channels
and local channels at the corresponding wavelength).
Initialise all matrix
sizes to 0

Simulate single
failure (link or
node)

Resource allocation
(wavelength, fiber)

For all nodes

For all sub matrix

Update matrix size

Figure 36. Dimensioning method for structure 2

1999 EURESCOM Participants in Project P709 page 39 (62)


Volume 5: Annex D Interaction between optical and clinet layer Deliverable 2

4.2.3 Application to a network example

The two previous methods have been applied to the European network example
described in Annex C.
Assumptions were as follows :
No wavelength conversion case
End-to-end restoration : two end-to-end paths were assigned for each demand on
the basis of the shortest-path algorithm
Wavelength and fibre allocation was achieved through Graph Colouring methods
used in Annex C.
Spare resource was allocated by simulating each possible single failure (node or
link) and re-calculating resource allocation.
The sizes of switches were rounded up to the closest 2n integer.
As far as structure 1 is concerned, 2 nodes in the network would require 64x64
switches, 8 nodes require 128x128 switches and one node requires a 256x256 switch.
Compared to the state of the art of optical switching, it shows that structure 1 would
hardly be feasible under the traffic and network assumptions taken here.

100
90
80
70
60 N=4
50 N=8
40 N=16
30
20
10
0
2x2 4x4 8x8 16x16 32x32 64x64

Figure 37. Total number of switches in the network for various capacity per fibre (N)
Figure 37 reports the total number of optical switches of given sizes in the network for
structure 2, for various capacity per fibre. The influence of the number of wavelengths
per fibre on the switch size points out the trade-off between the transmission costs and
complexity (N) and the switching cost (related to the size of the switches).

page 40 (62) 1999 EURESCOM Participants in Project P709


Deliverable 2 Volume 5: Annex D Interaction between optical and clinet layer

5 Optical/Electrical layers : management possibilities


The OAM (Operation, Administration and Maintenance) requirements to the optical
networks have been already defined in P615 project [6]. The present chapter is a short
analyse of the possibilities of management of the optical layers, taking part of an SDH
electrical client layer.
First, the processing of the management information required to control the optical
layers, is described. Then, the question is what kind(s) of signal could be defined to
support the optical overhead data (optical OAM parameters and TMN data) ? Is it
possible to use unallocated bytes of the SDH client overhead to handle the OAM
optical parameters (or only some of them) ?
Is it possible also to use the SDH layer to transport the TMN data exchanged between
the management system and the optical nodes ?

5.1 Management information processing


The management information required to manage an optical network are :
TMN data for requests and notifications exchanged between the management
system and the optical nodes.
OAM parameters for signalling, monitoring and supervision information
exchanged between optical adjacent nodes.
The combination (TMN data + OAM parameters) defines the optical overhead data.
The Technical Annex B of P615 project [6] define the OAM parameters of equipment
boxes for each optical layer (OCH, OMS and OTS). It defines also how optical
overhead data is processed by the Termination functions at each optical layer, and how
TMN data is processed by the Message Communication Function at OCH layer and
OMS layer :
B Ac
DAc
OCH_T O C H /client_A
O CHO H T M od
C arried out b y
term ination
function:
S
overhead
g en eration an d EM F E quipm ent specific m onitoring, e.g.:
term ination laser p ower & bias
D eal w ith alarm s ,
e.g.:
-O C H (sup ) loss of
su pervis ory
-O C H (sup ) loss
of tim ing /fram e
-etc...

M CF
Q

TM N

Figure 38. Optical channel layer elements

1999 EURESCOM Participants in Project P709 page 41 (62)


Volume 5: Annex D Interaction between optical and clinet layer Deliverable 2

Ac Am
OM S_T OM S/OCH_A
O MSOH WA OM
Carried out by
termination
function:
S
Equipment specific monitoring, e.g.:
overhead
monitoring:
EM F Signal level of added & dropped wavelengths opt. DCC
-signal level Frequency of demux channels
-error checking
Demux power supply voltage
Alarms:
-OM S (sup) loss of Equipment specific configuration:
supervisory Selection of added/dropped channels
-OM S (sup) loss Adjustement of demux frequency
of timing/frame
-etc...

M CF
Q

TMN

Figure 39. Optical multiplex section elements

Am Am
DAt
Carried out by OTS_T OTS/OMS_A
termination OTSOH OSPt FFS
function:
overhead
monitoring:
-signal level
S
-error checking
Alarms:
EMF Equipment specific monitoring, e.g.:
total power input/ouput
-OTS (sup) loss of OA pump laser power/current temperature
supervisory
power supply voltage
-OTS (sup) loss
of timing/frame amount of dispersion
-OTS (sup) AIS Equipment specific configuration:
-OTS (sup) - Error signal level setting
-OTS (sup) FERF Laser bias current/amplifier gain/temp. control opt. DCC

opt. DCC
MCF

Figure 40. Optical Transmission section layer elements

page 42 (62) 1999 EURESCOM Participants in Project P709


Deliverable 2 Volume 5: Annex D Interaction between optical and clinet layer

The following diagram identifies the different possibilities of Management Signal


implementation to support the optical Overhead:

OCH OCH OMS, OTS

Management Client signals Surmodulation of Separate optical External


Signal support (1..n) client signals supervisory communication
(1..n) signal (osc) channel

Management STM-16 STM-4 STM-1 Specific STM1 IP Specific TCP/IP,


bit bit X25,
Signal type
protocol protocol ISDN, ...

Electrical SDH ? Optical OAM


parameters TMN data
overhead
?

Optical overhead

OAM functions Configuration Alarm Performance


Management Management Monitoring

Figure 41. Overview of management signal implementation possibilities


The two transitions symbolised by two dashed arrows on the previous diagram
between Optical overhead and Electrical SDH overhead are studied in the next
section.

5.2 Possible interaction between SDH overhead and optical


overhead
The general idea of this section is to look at the possibilities and the constraints of an
implementation of the optical overhead taking part of the SDH overhead, i.e. some
unallocated bytes of the SDH overhead. (We suppose that the electrical clients are
exclusively SDH signals. Of course, in case of different clients signals transmitted
over different wavelengths on the same fiber (PDH, SONET, ATM,...), this
implementation of overhead is no more applicable).

5.2.1 Constraints and limitations

An Electrical processing is needed at each optical channel level to process the


optical overhead. This constraint implies the use of OTUs (Optical Translating
Unit) in OXC, OADM, or Optical Line Terminal nodes, in order to modify the
SDH overhead unallocated bytes with the received optical OAM parameters
related to the OTS, OMS and OCH. A possible implementation is shown
hereafter.

1999 EURESCOM Participants in Project P709 page 43 (62)


Volume 5: Annex D Interaction between optical and clinet layer Deliverable 2

Management
OperationSystem

TMN interface

Client network

SDH
SDH SDH
SDH
Rx
Rx Rx
Tx

Optical Transport network OTU OTU


OTU
OTU receiver
receiver
receiver transmitter

OTOH OTOH
OMOH OCOH OMOH
extraction insertion
CTIO
1 1
OTOH1 OTOH2

2 2 Optical
OLA1 OLA2 Optical Optical
DMUX MUX
osc Switch osc
3 Matrix 3

4 4

Figure 42. Implementation of OXC-node with processing of optical overhead


transferred into unallocated bytes of the SDH signals SOH
OTS and OMS overhead information should be carried separately over a
dedicated wavelength (osc), and can be added in each SDH client overhead,
with the OCOH by the OTUs. With this solution, the data frame supported by
the optical supervisory channel (osc) should be large enough to encapsulate
OTS overhead of a limited number of adjacent Optical Amplifiers (for example
20 in Figure 43):
20 20 20

3 OTOH3
2 OTOH2 2 OTOH2
1 OTOH1 1 OTOH1 1 OTOH1
osc
OXC OA1 OA2 OA3 OXC

Figure 43. Example of dedicated wavelength to carry OTS and OMS overhead
information
Different STM-n signals may be transported on the same optical network, via
different transponders, as illustrated below in Figure 44:

page 44 (62) 1999 EURESCOM Participants in Project P709


Deliverable 2 Volume 5: Annex D Interaction between optical and clinet layer

Transponder A 1
STM1
2 Optical
Transponder B
STM4
MUX
3
STM16 Transponder C

Transponder C 4
STM16

Figure 44. Example of a 4 WDM system, transporting different STM-n signals


on 4 optical channels
In such scenario, we can only use unallocated bytes of STM1 overhead (see next
paragraph).
In case of failure of the SDH signal, the management of the optical layers is no
more operational.
The use of allocated bytes of SDH overhead for optical management purpose
should be forbidden, to avoid non operating effects of SDH layer (due to
overhead modifications) and mismanagement of the SDH Management System.

5.2.2 Use of unallocated SDH overhead Bytes for optical overhead

5.2.2.1 Unallocated bytes of SDH overhead

RS overhead MS overhead Bytes for national use

Media dependent bytes

1 49 97 14
4
A1 A A A1 A A2 A A A A2 A J0 Z0 Z0 Z0 *
1 1 1 2 2 2 2
B1 E1 F1
D1 D2 D3
H1 H H H1 H H2 H H H H2 H H3 H H H3 H3
1 1 1 2 2 2 2 3 3
B2 B2 B2 B2 B2 K1 K2
D4 D5 D6
D7 D8 D8
D1 D1 D1
0 1 2
S1 M E2
1
STM-16 overhead

==>856 unallocated bytes

1999 EURESCOM Participants in Project P709 page 45 (62)


Volume 5: Annex D Interaction between optical and clinet layer Deliverable 2

1 13 25 36
A1 A1 A1 A1 A1 A1 A2 A2 A2 A2 A2 A2 A2 J0 Z0 Z0 Z0 * * * * * *
B1 E1 F1
D1 D2 D3
H1 H1 H1 H1 H1 H1 H2 H2 H2 H2 H2 H2 H2 H3 H3 H3 H3 H3 H3 H3 H3 H3 H3 H3 H3
B2 B2 B2 B2 B2 B2 K1 K2
D4 D5 D6
D7 D8 D8
D10 D11 D12
S1 M1 E2

STM-4 overhead

==>198 unallocated bytes

1 4 7 9
A1 A A A2 A A J0 * *
1 1 2 2
B1 E1 F1
D1 D2 D3
H1 H H H2 H H H3 H H3
1 1 2 2 3
B2 B2 B2 K1 K2
D4 D5 D6
D7 D8 D8
D1 D1 D1
0 1 2
S1 M E2
1
STM-1 overhead

==>30 unallocated bytes

5.2.2.2 Definition of a byte structure for optical Overhead

optical tracing bytes


OCH trace (J2): it provides an OCH identifier, to do end-to-end monitoring of optical
route.

page 46 (62) 1999 EURESCOM Participants in Project P709


Deliverable 2 Volume 5: Annex D Interaction between optical and clinet layer

optical monitoring bytes


The purpose is to detect Degradation of Signal, at input of each OA (Degradation of
OTS), at input of Multiplexer (Degradation of OMS) and also for each optical channel
(Degradation of OCH). The method of measurement and calculation of the Quality of
Signal should be studied. The following bytes could be defined to code the quality of
signal :
B4 for the signal quality of OTS layer
B5 for the signal quality of OMS layer
B6 for the signal quality of OCH layer
optical APS bytes
At least, two bytes (K4,K5) should be necessary to perform Automatic Protection
Switching of OCH layer :
K4 : to code the Protection Switching message of OCH
K5 : to code OCH AIS and OCH RDI alarms detection
optical DCC bytes
Optical DCC bytes should transport optical TMN data between adjacent nodes. The
number of new DCC bytes should be fixed according to the data rate that is necessary
to transfer the messages that are listed in the following table :
TMN message data optical layer
configuration of the OXC Switch matrix OCH
configuration of the optical Multiplexer OMS
control of the OCH connectivity, with OCH identifiers OCH
restoration of OCH trails OCH
alarms
Loss Of Signal (LOS) OTS, OMS, OCH
Loss Of Power (LOP) OTS, OMS, OCH
Power Degradation OTS, OMS, OCH
Trail Identifier Mismatch OMS, OCH
Alarm Indication Signal (AIS) OTS, OMS, OCH
Remote Defect Indication (RDI) OTS, OMS, OCH
performance parameters
Total input power OTS, OMS
Total output power OTS, OMS
Input power per channel OCH
Output power per channel OCH
Input OSNR OCH
Ouput OSNR OCH
OA pump current OTS
OA pump temperature OTS
Multiplexer temperature OMS
Table V - Link between TMN message data and opical layer

1999 EURESCOM Participants in Project P709 page 47 (62)


Volume 5: Annex D Interaction between optical and clinet layer Deliverable 2

As the data messages are specific to a layer, the optical DCC bytes can be structured
in OTS DCC, OMS DCC and OCH DCC. The following rates are given as a proposal
of use of unallocated STM1 overhead bytes :
4 bytes (D13-D16) for OTS DCC --> 256 kbit/s rate
2 bytes (D17-D18) for OMS DCC --> 128 kbit/s rate
8 bytes (D19-D26) for OCH DCC --> 512kbit/s rate
Tracing bytes J2 for OCH trace identifier
Monitoring bytes B4 for OTS Signal quality
B5 for OMS Signal quality
B6 for OCH Signal quality
APS bytes K4 for OCH APS command
K5 for OCH AIS and RDI alarm detection
DCC bytes D13-D16 for OTS DCC
D17-D18 for OMS DCC
D19-D26 for OCH DCC
total 20 bytes
Table VI - Summary of optical overhead bytes
The total is less than the number of unallocated STM1 overhead bytes (30)
All of these bytes are written into electrical overhead of each STM client signal by the
OTU. Of course, the bytes related to OMS and OTS sections are repeated in each SDH
client overhead, but this can be easily handled and masked by the management
system.

page 48 (62) 1999 EURESCOM Participants in Project P709


Deliverable 2 Volume 5: Annex D Interaction between optical and clinet layer

6 Dimensioning methods for optical/electrical layers


This chapter considers the task of dimensioning and designing a WDM network when
the underlying client is SDH and the traffic demands are given in VC4s. To simplify
our task, we only consider here the case where the WDM network is :
A WDM ring
1+1 protection (SNC-P) is used at the VC4 level, or
MS-SPRing protection is used in SDH rings running over the WDM
Our task is to minimise the cost of the WDM ring, whilst satisfying the traffic
demands.
First, the problem considered is explained in more detail. Secondly, a dimensioning
method based on the concept of simulated annealing is suggested. Thirdly, the
suggested method is applied to simple test scenarios and evaluated

6.1 The traffic grooming problem or how to minimise the number


of terminal equipment in the network
In Annex C, dimensioning methods for various types of WDM networks were
discussed, but in all cases the input to the WDM network dimensioning process was
the number of wavelength paths required between pairs of points. The network
dimensioning process was effectively by layer.
The process of optimising each layer independently from the others (usually starting at
the top and moving successively closer to the physical layer) may not always achieve
the overall best design, but it simplifies the network design task considerably. Also, in
practice, clients of WDM networks can be of various types (e.g. SDH, PDH, ATM, IP,
Gigabit Ethernet), and a completely general method for joint optimisation of the
optical and client layers is probably impractical.
Nonetheless, such a two-layer optimisation is here attempted. To simplify the task, a
joint process of dimensioning the optical layer (a WDM ring which provides fibre and
repeater savings) and the SDH layer (which provides protection either using the MS-
SPRing protocol in ADMs as shown in Figure 45, or using SNC-P protection at the
VC4 level in a 4/4 DXC as shown in Figure 46) is considered.
In both cases, the WDM ring uses optical add-drop multiplexers (OADMs) to access
wavelengths at the nodes where they must be dropped. These OADMs may be
frozen, as in the first prototypes of this technology proposed by system vendors, or
else they may be reconfigurable. Typically, an OADM will provide access to any
chosen wavelength in both directions (East and West). The assumption that the WDM
ring is already installed, with an OADM at every node, is made.
In the case where SNC-P protection in a 4/4 DXC is used, the dimensioning process
aims to determine and minimise the number of SDH terminal multiplexers, and
OADM wavelength block filters required. For assumption, the SDH terminal
multiplexers will be installed in pairs so as to access a wavelength in both directions at
a node. Since the traffic is 1+1 protected, this constraint does not seem to result in a
higher network cost.

1999 EURESCOM Participants in Project P709 page 49 (62)


Volume 5: Annex D Interaction between optical and clinet layer Deliverable 2

OADM
1 2

West East

Coloured
aggregate
STM16
Terminal STM1
multiplexer
4/4 SDXC SNC-P for
Transit VC4 terminating VC4
STM1 access
Figure 45. Architecture for WDM ring with SNC-P protection in 4/4 DXC

OADM
1 2

West East

MS-SPRing
ADM with
coloured
aggregates

Transit VC4 Terminating VC4


Figure 46. Architecture for WDM ring with MS-SPRing client providing
protection
In the case where MS-SPRing protection is used, the dimensioning process aims to
determine locations and numbers of SDH ADMs and OADM wavelength block
filters. the constraint that every SDH ADM interfaces at the same wavelength on both
the east and west sides is imposed. Since every SDH ring must be closed around the
ring, this constraint cannot result in higher network costs. The problem is therefore
equivalent to minimising the number of SDH ADMs.
Whether SNC-P protection or MS-SPRing protection is used, therefore, the
dimensioning process outlined in Figure 47 attempts to minimise the number of

page 50 (62) 1999 EURESCOM Participants in Project P709


Deliverable 2 Volume 5: Annex D Interaction between optical and clinet layer

OADM wavelength blocks and SDH ADM or terminal multiplexers (Terminal


equipment). In contrast to previous dimensioning approaches where the number of
wavelengths were minimised, in this case the reduction of the number of wavelengths
used is no longer attempted, with respect of the constraint of capacity of the WDM
ring. This is because the incremental cost of lighting up one wavelength is purely the
cost of the ADMs/Terminal multiplexers/transponders which are needed.
Note that this network dimensioning task is not necessarily specific to WDM
networks, and the same arguments could be used for purely SDH networks (for
example, the ring stacking problem when SDH rings run directly over fibre).
However, there can be a greater incentive to minimise fibre usage when fibres are
scarce, than when using a WDM system with many wavelengths.

INPUTS OUTPUTS

Optimise transponder and


WDM ring with
SDH LTE/ADM
flexible OADMs
locations
VC4 traffic matrix
wavelength
4/4 DXC client
assignments
layer
VC4 routings

Figure 47. Two-layer dimensioning process


In order to minimise terminal equipment, traffic travelling to different destinations
may be combined to make better use of wavelengths. Figure 48 shows an example
where demands from Node 4 to Node 3 and Node 4 to Node 2 are combined between
Nodes 4 and 2 and carried over the same wavelength.
One STM-16: containing OADM 1
8 STM-1 from node 4 to node 2 and
8 STM-1 from node 4 to node 3
SDH 4/4 cross-connect

INPUTS

OUTPUTS
OADM 4 OADM 2

8 STM-1 from node 4 to node 2


dropped using 4/4 DXC

OADM 3
The remaining 8 STM-1s
to node 3

Transponder and
SDH LTE MUX

Figure 48. Different traffic sources and destinations in WDM ring

1999 EURESCOM Participants in Project P709 page 51 (62)


Volume 5: Annex D Interaction between optical and clinet layer Deliverable 2

In Figure 49, it is shown that minimising wavelengths or terminal equipments are


sometimes mutually exclusive options. In both cases, demands of 8 STM1s, between
nodes 1&3, and 2&4, are transported. The first option uses the least number of
terminal equipment, and 2 wavelengths, whereas the second option uses only one
wavelength, but 50% more terminal equipment.
Another example of the optimisation procedure is shown in
Figure 50. In this scenario, a fully-meshed traffic demand of 1 STM-1 between each
pair of nodes is required. Allocating one wavelength for each demand would be
wasteful: 20 terminal multiplexers, and 3 wavelengths would be required, and each
STM16 wavelength would be used to 1/16th of its capacity.

OADM1 OADM

OADM 4 O A DM 2 OA D M OADM

OADM3 O AD M

Dem ands from 4 to 2 ( eight STM -1 )and Dem ands from 4 to 2 ( eight STM -1 )and
from 1 to 3 ( eight STM -1 ) from 1 to 3 ( eight STM -1 )

2 w avelengths and 4 transponders 1 wavelengths and 6 transponders

Figure 49. Number of wavelength versus number of transponders


The solution given by the optimising tool is shown in
Figure 50. One wavelength is used, most of the way around the ring in a horseshoe
configuration, and 8 terminal multiplexers are required. The loading of the
wavelengths along the used spans are 4, 6, 6, and 4 STM1s.
Demands from node 1
Capacity limit = 16 STM-1
Capacity limit = 16 STM-1 Used capacity = 6 STM-1
Used capacity = 6 STM-1
Demands from node 2

1 Demands from node 3


Demands from node 4

5 2

Capacity limit = 16 STM-1


Capacity limit = 16 STM-1 Used capacity = 4 STM-1
Used capacity = 4 STM-1

4 3 Wavelength
number one

Fully meshed demands case

1 wavelength and 8 transponders

Figure 50. Traffic grooming in a five-node WDM ring

page 52 (62) 1999 EURESCOM Participants in Project P709


Deliverable 2 Volume 5: Annex D Interaction between optical and clinet layer

If the traffic in this scenario is increased from 1 to 3 STM-1s between each pair of
nodes, then using the same routing, a capacity of 18 STM-1s would be required on
spans 1-5 and 1-2, leading to a solution using 2 wavelengths, and 12 terminal
multiplexers. The optimisation tool, in contrast, prefers to complete the ring by
lighting up wavelength 1 on span 3-4, and finds a 1-wavelength solution requiring 10
terminal multiplexers.

6.2 Dimensioning method based on simulated annealing


A dimensioning method based on simulated annealing has been developed to groom
and route VC4s in order to minimise terminal equipment costs over a WDM ring. This
method can be applied for a network where the WDM ring provides fibre relief and
node-bypass for through traffic, and the traffic protection is carried out in the SDH
layer (either using SNC-P or MS-SPRing). The procedure is very similar to the
simulated annealing tools developed purely for wavelength routing in Annex C, and is
depicted in Figure 51. The differences are:
Each wavelength now has a capacity of 16 STM1s (when SNC-P is used) or 8 STM-1s
(when MS-SPRing is used), instead of a capacity of 1 demand previously.
The cost of using available wavelengths around the ring are now very low, and are
only included to prevent clearly nonsensical routings, and force the usage of the lower
numbered wavelengths rather than choosing them at random.
Most of the network cost is now associated with links internal to the nodes, which join
the traffic sources and sinks to the wavelength planes. These links represent terminal
equipment (transponders, pairs of terminal multiplexers, SDH ADMs) as well as the
necessary filters required to add-drop a wavelength at a node.
This model ensures that if a single VC4 needs to be accessed in one wavelength at one
node, then an internal link (hence terminal equipment, transponders etc) is required.
However, if no VC4s need to be added or dropped at a node, the wavelength can pass
through transparently without incurring any additional cost.
Bundling wavelengths
in same fibre (for diversity)
Very low cost associated
with wavelength
number Wavelength planes

Links representing
terminal equipment
& OADM
(b) wavelength block

SDH grooming incurs cost Traffic sources and


sinks
Network topology
(a)

Figure 51. Schematic of simulated annealing model for 2-layer optimisation

1999 EURESCOM Participants in Project P709 page 53 (62)


Volume 5: Annex D Interaction between optical and clinet layer Deliverable 2

Diverse routing is possible if 1+1 (SNC-P) protection is used. This is done by adding a
large penalty cost when both working and protection paths of a VC4 are found to use
the same cable.
When MS-SPRing is assumed in the client layer, protection is implicit, and VC4s are
only routed one way around the ring assuming that each wavelength can only carry 8
working STM1s.

6.3 Application of the tool to network examples


The combined SDH/WDM dimensioning tool was applied to some simple scenarios
using 1+1 and SPRing protection. A 5 node WDM ring was considered. In all cases, a
uniform fully-meshed traffic pattern was assumed, but the size of the demand between
each pair of nodes was varied between 1 and 16 VC4s.
Given this full mesh traffic demand, 3 different dimensioning approaches where
considered:
The first approach used the dimensioning tool to minimise the number of Terminal
multiplexer pairs (1+1 case) or ADMs (SPRing case), by clever grouping and routing
of the VC4s.
The second approach simply stacks SDH rings on top of one another, and WDM is
only used between nodes. At every node, every wavelength is demultiplexed by a
terminal multiplexer of ADM.
The third approach assigns a wavelength to every demand, no matter how small. For
this case, 1+1 protection and SPRing protection are equivalent, since when SPRing
ADMs are used, each SDH ring has only two nodes.

6.3.1 Results for 1+1 protection

For the case of 1+1 (SNC-P) protection, the numbers of terminal multiplexers are
plotted in Figure 52 for the three approaches. Demultiplexing every wavelength at
every node is only economical for very low traffic. When the traffic exceeds 8 VC4s
between each node, it is interesting to note that assigning a wavelength to every
demand requires the same number of terminal equipment as the more optimised
version, which uses less wavelengths when the traffic does not exceed 10 VC4s
between each nodes. The wavelength usage for the 3 methods is plotted in Figure 53.

page 54 (62) 1999 EURESCOM Participants in Project P709


Deliverable 2 Volume 5: Annex D Interaction between optical and clinet layer

60

Optimised
50
Demux at every node
Number of TM pairs required Wavelength for every demand
40

30

20

10

0
0 2 4 6 8 10 12 14 16 18
traffic between every node pair (VC4s)

Figure 52. Numbers of terminal multiplexer pairs required for full mesh connectivity in
a 5-node ring network, versus the traffic volume, assuming 1+1 protection
It is interesting to note that in many cases, the network dimensioned to minimise
terminal equipment uses more wavelengths than when each wavelength is processed at
every node.

12

10
Number of wavelengths required

4 Optimised

Demux at every node


2
Wavelength for every demand

0
0 2 4 6 8 10 12 14 16 18
Traffic between every node pair (VC4s)

Figure 53. Numbers of wavelengths required for full mesh connectivity in a 5-


node ring network, versus the traffic volume, assuming 1+1 protection

1999 EURESCOM Participants in Project P709 page 55 (62)


Volume 5: Annex D Interaction between optical and clinet layer Deliverable 2

The optimised configuration for a fully meshed demand of 2 VC4s per node requires
14 terminal multiplexers and two wavelengths, and is shown in Figure 54.

5 2

4 3

Figure 54. Terminal multiplexer placements and wavelength usage determined for
a full mesh demand of 2VC4s per node pair, for 1+1 protection at the VC4 level
In the case when 8 VC4s are required between every pair of nodes, the optimised
solution uses only 15 terminal multiplexer pairs and only 5 wavelengths, compared to
the case where each demand is given a wavelength (which requires 20 TM pairs and
10 wavelengths). The optimised configuration, as produced by the simulated annealing
tool, is shown in Figure 55. Each wavelength supports 2 demands of 8 VC4s each,
which completely fills the wavelength. Each wavelength is only accessed at 3 nodes
on the ring.

1 2 3

8 VC4s 4 5 LTE
accessed pair

Figure 55. Placement of terminal multiplexers for 1+1 protection of a demand of


8 VC4s between each node pair. 15 TM pairs and 5 wavelengths are required.
The black lines represent a ring capacity of 8 VC4s

page 56 (62) 1999 EURESCOM Participants in Project P709


Deliverable 2 Volume 5: Annex D Interaction between optical and clinet layer

6.3.2 Results for SPRing protection

For SPRing protection, the optimisation tool was used to minimise the number of
ADMs required as a function of traffic level. For comparison, the case where ADMs
are placed at every node (stacked rings) is also considered, as well as the case where
only 2-node SPRings (equivalent to 1+1 protection) where used. The number of
ADMs required for the different cases are shown in Figure 56.

35

30
Number of ADMs required

25

20

15

10 Optimised

Demux at every node


5
Wavelength for every demand
0
0 2 4 6 8 10 12 14 16 18
Traffic between every node pair (VC4s)

Figure 56. Numbers of ADMs required for full mesh connectivity in a 5-node ring
network, versus the traffic volume, assuming SPRing protection
It can be seen that by optimising traffic routings, fewer ADMs are required than by
deploying 2-node rings, for traffic of up to 12 VC4s between each node pair. The fibre
usage of the different solutions is also shown in Figure 57.

12
Number of wavelengths required

10

Optimised
8
Demux at every node

6 Wavelength for every demand

0
0 2 4 6 8 10 12 14 16 18
Traffic between every node pair (VC4s)

Figure 57. Numbers of wavelengths required for full mesh connectivity in a 5-


node ring network, versus the traffic volume, assuming SPRing protection

1999 EURESCOM Participants in Project P709 page 57 (62)


Volume 5: Annex D Interaction between optical and clinet layer Deliverable 2

Again, attempting to minimise terminal equipment cost can sometimes result in higher
wavelength counts. It is also interesting to note that even when traffic demands are
large enough to completely fill a 2-node ring, deploying 2-node rings consumes more
wavelengths, even though the number of ADMs is the same as for the optimised
design.
An example of the configuration produced by the simulated annealing tool is shown in
Figure 58 for a full mesh demand of 8 VC4s between each pair of nodes. 10 ADMs
and 3 wavelengths are required. The optimised solution for transporting 16 VC4s
between each pair of nodes is simply obtained by doubling the configuration shown in
Figure 58 . It uses only 6 wavelengths, compared to 10 if 2-node rings are stacked.

1 2 3

Figure 58. SPRing ADM placements for a design optimised to carry 8 VC4s between
each pair of nodes. The purple lines represent the routings for the demands

6.3.3 Comparison between 1+1 and SPRing protection

If 1+1 protection is used, pairs of terminal multiplexers are required at a node to


access a wavelength. An SDH ADM can equivalently be used instead of a pair of
terminal multiplexers, provided the protection is turned off in the ADM (protection is
assumed to be provided by the cross-connect). If SPRing protection is used, an ADM
is required to access a wavelength at a node. The numbers of terminal equipment
required for the two different types of protection are shown in Figure 59.

25
Number of TM pairs/ADMs required

20

15

10
SPRing protection
1+1 protection
5

0
0 2 4 6 8 10 12 14 16 18
Traffic between every node pair (VC4s)

Figure 59. Optimised numbers of terminal multiplex pairs/ADMs required for either
1+1 or SPRing protection, to provide full mesh connectivity, versus traffic volume

page 58 (62) 1999 EURESCOM Participants in Project P709


Deliverable 2 Volume 5: Annex D Interaction between optical and clinet layer

It is interesting to note that at high traffic levels, 1+1 protection requires no more
terminal equipment than SPRing protection. The corresponding wavelength
requirements are shown in Figure 60. SPRing protection requires 40% fewer
wavelengths for high traffic.

12
Number of wavelengths required

10

SPRing protection
2
1+1 protection

0
0 5 10 15 20
Traffic between every node pair (VC4s)

Figure 60. Wavelength requirements for SPRing and 1+1 protection for a 5-node ring
The dimensioning process of a 2-layer SDH/WDM network has been considered, in
the case where the SDH layer protects the traffic against cable breaks, either using
SNC-P or MS-SPRing, and where a WDM ring with OADMs is used to pass through-
traffic transparently across nodes when required.
When traffic demands are given in VC4s, and each wavelength can carry up to 16
VC4s, several approaches to the dimensioning process are possible:
Allocate a wavelength to every demand end-to-end over the ring: this can
wasteful when individual demands of less than about 8 VC4s are present.
Terminate every wavelength at every node, and use SDH grooming to fill every
wavelength efficiently: this becomes wasteful when the demand volumes are
large.
Optimise the VC4 routings in the network to minimise the number of expensive
terminal equipment.
The third approach was studied here, and a dimensioning tool based on simulated
annealing was developed. The tool attempts to minimise the numbers of terminal
multiplexers, transponders, or SDH ADMs, while at the same time using the minimum
number of wavelengths and avoiding wavelength conflicts.
The tool was applied to example scenarios where a network was dimensioned to
support a uniform, fully meshed demand pattern of varying size, for the case of 1+1
and SPRing protection. A number of observations were made after examining the
results:
Optimising the VC4 routing and the placement of terminal equipment can result is
large savings, both for 1+1 and SPRing protection.

1999 EURESCOM Participants in Project P709 page 59 (62)


Volume 5: Annex D Interaction between optical and clinet layer Deliverable 2

For 1+1 protection, assigning a whole wavelength to a demand becomes economical a


soon as the demand is larger than half the capacity of the wavelength.
Using 2-node rings is wasteful of wavelengths compared to SPRing protection, even
when each demand is large enough to completely fill a 2-node ring.
For large traffic volumes, the number of terminal equipment required for 1+1
protection is very similar to that required when SPRing protection is used.
These observations may be converted into guidelines for designing SDH/WDM
networks capable of provisioning a certain traffic level of fully meshed demands at the
lowest possible cost. In particular, certain designs produced automatically by the
simulated annealing tool are of interest, and can be picked out and applied where
required.

page 60 (62) 1999 EURESCOM Participants in Project P709


Deliverable 2 Volume 5: Annex D Interaction between optical and clinet layer

7 Conclusion
The aim of this Annex concerns the interactions between the optical and electrical
layers. Its objectives consists in identifying different optical architectures.
It is shown through this document that the main area of optical/electrical networks
planning is correctly covered. Indeed, electrical layer network planing aspects are
briefly recalled in the first chapter as they have already been detailed in Eurescom
P615. References are given for both P615 project and ITU-T recommendations.
Moreover, SDH architectures and protection policies are also recalled and compared.
The third chapter can be considered as the starting point for this Annex. Indeed, the
concept of optical architectures and classification methods are recalled according to
Deliverable 1 of P709.
The optical layers are defined and both traffic routing and traffic protection are
provided. Moreover, the capabilities and the limitations of such networks are stressed.
Finally, general information, functional models, implementations and design rules are
described for all the three architectures selected in P709 within Deliverable 1.
Once the architectures are defined and their models are described under their own
constraints, dimensioning methods for solving and evaluating such architectures are
given in the third chapter. Such methods are described according to different
architectures i.e. rings or meshed networks.
Firstly, multiple ring and optical MS-SPRing dimensioning methods are depicted. The
underlayed theoretical methods are stressed for ring based networks. Moreover,
routing problems in multiple ring network is provided
Secondly, for the optical networks, the dimensioning methods are based on resource
allocation algorithms described in Annexes A and B. The dimensioning methods takes
into account wavelength conversion and an application to network example described
in Annex C is given. The underlayed method is based upon graph coloring methods.
Before dimensioning electrical/optical networks, the management aspects are
depicted. They are based on TMN data and OAM parameters. The constraints and
limitations are also given. Interactions and feasibility problem are presented.
Moreover, the importance of electrical processing for optical is stressed.
Finally, concerning the dimensioning methods for electrical/optical layers, the
assumption of the model chosen are described and two problems are presented.
Firstly, the traffic grooming problem which consists in minimizing the number of
terminal equipments. Different protections such as SNC-P and MS-SPRing are
handled. Such methods are clearly illustrated by several examples. Secondly, a
simulated annealing based algorithm is given for grooming and routing VC_4s in
order to minimize the terminal equipment costs over WDM rings. Additional
information to WDM routing described in Annex C is given. Moreover, numerical
results are depicted for 5 node architecture networks with different protection
schemes.
Finally, this document provides guidelines for optical network planing covering
different aspects going from pure electrical architectures to pure optical counterpart
through electrical/optical architectures. Network planners may find information both
from management point of view and from dimensioning point view.

1999 EURESCOM Participants in Project P709 page 61 (62)


Volume 5: Annex D Interaction between optical and clinet layer Deliverable 2

References
[1] ITU-T Recommendation G.841 Types and Characteristics of SDH Network
Protection Architectures, July 95
[2] H. Okamura and P. D. Seymour, Multicommodity flows in Planar Graphs,
Jnl. Combin. Theory, Series B 31, pp.75-81 (1981).
[3] A. Frank, Edge-Disjoint Paths in Planar Graphs, Jnl. Combin. Theory,
Series B 39, pp. 164-178 (1985).
[4] D. Marcenac, C. Felicite, N. Stewart, EURESCOM P709, PIR3.3, contribution
BT33-05a, Techniques for resource allocation and dimensioning for optical
mesh networks with 1+1 protected traffic and their comparison, November
1998.
[5] D. Marcenac, EURESCOM P709, PIR3.3, contribution BT33-02a,
Identification of criteria for resource allocation and dimensioning techniques
for optical rings, July 1998.
[6] Project P615: "Evolution towards an Optical Network Layer", Deliverable 1,
Annex B : OAM requirements to the optical network layer.

page 62 (62) 1999 EURESCOM Participants in Project P709

Das könnte Ihnen auch gefallen