WILEY Core and metro networks User Manual

CORE AND METRO NETWORKS
WILEY SERIES IN COMMUNICATIONS NETWORKING
& DISTRIBUTED SYSTEMS
Series Editors: David Hutchison, Lancaster University, Lancaster, UK
Serge Fdida, Universite´Pierre et Marie Curie, Paris, France Joe Sventek, University of Glasgow, Glasgow, UK
The ‘Wiley Series in Communications Networking & Distributed Systems’ is a series of expert-level, technically detailed books covering cutting-edge research, and brand new developments as well as tutorial-style treatments in networking, middleware and software technologies for communications and distributed systems. The books will provide timely and reliable information about the state-of-the-art to researchers, advanced students and development engineers in the Telecommunications and the Computing sectors.
Other titles in the series:
Wright: Voice over Pac ket Networks 0-471-49516-6 (February 2001) Jepsen: Java for Telecommunications 0-471-49826-2 (July 2001) Sutton: Secure Communications 0-471-49904-8 (December 2001) Stajano: Security for Ubiquitous Computing 0-470-84493-0 (February 2002) Martin-Flatin: Web-Based Management of IP Networks and Systems 0-471-48702-3 (September 2002) Berman, Fox, Hey: Grid Computing. Making the Global Infrastructure a Reality 0-470-85319-0 (March 2003) Turner, Magill, Marples: Service Provision. Technologies for Next Generation Communications 0-470-85066-3 (April 2004) Welzl: Network Congestion Control: Managing Internet Traffic 0-470-02528-X (July 2005) Raz, Juhola, Serrat-Fernandez, Galis: Fast and Efficient Context-Aware Services 0-470-01668-X (April 2006) Heckmann: The Competitive Internet Service Provider 0-470-01293-5 (April 2006) Dressler: Self-Organization in Sensor and Actor Networks 0-470-02820-3 (November 2007) Berndt: Towards 4G Technologies: Services with Initiative 0-470-0 1031-2 (Ma rch 2008) Jacquenet, Bourdon, Boucadair: Service Automation and Dynamic Provisioning Techniques in IP/MPLS Environments 0-470-01829-1 (March 2008) Minei/Lucek: MPLS-Enabled Applications: Emerging Developments and New Technologies, Second Edition 0-470-98644-1 (April 2008) Gurtov: Host Identity Protocol (HIP): Towards the Secure Mobile Internet 0-470-99790-7 (June 2008) Boucadair: Inter-Asterisk Exchange (IAX): Deployment Scenarios in SIP-enabled Networks 0-470-77072-4 (January 2009) Fitzek: Mobile Peer to Peer (P2P): A Tutorial Guide 0-470-69992-2 (June 2009) Shelby: 6LoWPAN: The Wireless Embedded Internet 0-470-74799-4 (November 2009)
CORE AND METRO NETWORKS
Editor
Alexandros Stavdas
University of Peloponnese, Greece
This edition first published 2010 Ó 2010 John Wiley & Sons Ltd.,
Except for: Chapter 1, ‘The Emerging Core and Metropolitan Networks’ Ó 2009 Angel Ferreiro and Telecom Italia S.p.A
Chapter 4, Section 4.5.1–4.5.5 and 4.5.7 Ó 2009 Telecom Italia S.p.A
Chapter 5, Section 5.2–5.6 Ó 2009 Telecom Italia S.p.A
Registered office
John Wiley & Sons Ltd, The Atrium, Southern Gate, Chichester, West Sussex, PO19 8SQ, United Kingdom
For details of our global editorial offices, for customer services and for information about how to apply for permission to reuse the copyright material in this book please see our website at www.wiley.com.
The right of the author to be identified as the author of this work has been asserted in accordance with the Copyright, Designs and Patents Act 1988.
All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, except as permitted by the UK Copyright, Designs and Patents Act 1988, without the prior permission of the publisher.
Wiley also publishes its books in a variety of electronic formats. Some content that appears in print may not be available in electronic books.
Designations used by companies to distinguish their products are often claimed as trademarks. All brand names and product names used in this book are trade names, service marks, trademarks or registered trademarks of their respective owners. The publisher is not associated with any product or vendor mentioned in this book. This publication is designed to provide accurate and authoritative information in regard to the subject matter covered. It is sold on the understanding that the publisher is not engaged in rendering professional services. If professional advice or other expert assistance is required, the services of a competent professional should be sought.
Library of Congress Cataloging-in-Publication Data
Core and metro networks / edited by Alexandros Stavdas.
p. cm. Includes bibliographical references and index. ISBN 978-0-470-51274-6 (cloth)
1. Metropolitan area networks (Computer networks) I. Stavdas, Alexandros A. TK5105.85.C678 2010
004.67–dc22
A catalogue record for this book is available from the British Library.
ISBN 9780470512746 (H/B)
Set in 10/12 pt Times Roman by Thomson Digital, Noida, India Printed and Bound in Singapore by Markono Pte.
2009044665

Contents

Preface ix
1 The Emerging Core and Metropolitan Networks 1
Andrea Di Giglio, Angel Ferreiro and Marco Schiano
1.1 Introduction 1
1.1.1 Chapter’s Scope and Objectives 1
1.2 General Characteristics of Transport Networks 1
1.2.1 Circuit- and Packet-Based Network Paradigms 2
1.2.2 Network Layering 3
1.2.3 Data Plane, Control Plane, Management Plane 4
1.2.4 Users’ Applications and Network Services 4
1.2.5 Resilience 5
1.2.6 Quality of Service 7
1.2.7 Traffic Engineering 8
1.2.8 Virtual Private Networks 10
1.2.9 Packet Transport Technologies 11
1.3 Future Networks Challenges 12
1.3.1 Network Evolution Drivers 12
1.3.2 Characteristics of Applications and Related Traffic 12
1.3.3 Network Architectural Requirements 17
1.3.4 Data Plane, Control Plane, and Management Plane Requirements 24
1.4 New Transport Networks Architectures 31
1.4.1 Metropolitan Area Network 33
1.4.2 Core Network 36
1.4.3 Metro and Core Network (Ultra-long-term Scenario) 38
1.5 Transport Networks Economics 39
1.5.1 Capital Expenditure Models 39
1.5.2 Operational Expenditure Models 42
1.5.3 New Business Opportunities 44
Acronyms 52 References 54
vi Contents
2 The Advances in Control and Management for Transport Networks 55
Dominique Verchere and Bela Be rde
2.1 Drivers Towards More Uniform Management and Control Networks 55
2.2 Control Plane as Main Enabler to Autonomic Network Integration 58
2.2.1 Generalized Multi-Protocol Label Switching 59
2.2.2 Evolution in Integrated Architectures 71
2.3 Multilayer Interactions and Network Models 74
2.3.1 Introduction 74
2.3.2 Vertical Integration and Models 78
2.3.3 Horizontal Integration and Models 79
2.3.4 Conclusions on UNI Definitions from ITU-T, OIF, IETF, and OIF UNI: GMPLS UNI Interoperability Issues 104
2.4 Evolution of Connection Services and Special Cases of Optical Networks 105
2.4.1 Evolution in Network Services 105
2.4.2 Virtual Private Networks 106
2.4.3 Layer 1 VPN 109
2.4.4 Layer 2 VPN 118
2.4.5 Layer 3 VPN 122
2.5 Conclusion 123 References 124
3 Elements from Telecommunications Engineering 127
Chris Matrakidis, John Mitchell and Benn Thomsen
3.1 Digital Optical Communication Systems 127
3.1.1 Description of Signals in the Time and Frequency Domains 127
3.1.2 Digital Signal Formats 132
3.2 Performance Estimation 135
3.2.1 Introduction 136
3.2.2 Modeling 141
3.2.3 Comparison of Techniques 146
3.2.4 Standard Experimental Measurement Procedures 149
References 158
4 Enabling Technologies 161
Stefano Santoni, Roberto Cigliutti, Massimo Giltrelli, Pasquale Donadio, Chris Matrakidis, Andrea Paparella, Tanya Politi, Marcello Potenza, Erwan Pincemin and Alexandros Stavdas
4.1 Introduction 161
4.2 Transmitters 161
4.2.1 Introduction 161
4.2.2 Overview of Light Sources for Optical Communications 167
4.2.3 Transmitters for High Data-Rate Wavel ength-Division Multiplexing Systems 178
Contents vii
4.3 Receiver 202
4.3.1 Overview of Common Receiver Compone nts 202
4.4 The Optical Fiber 212
4.4.1 Short Introduction to the Waveguide Principle 213
4.4.2 Description of Optical Single-Mode Fibers 216
4.4.3 Special Fiber Types 222
4.5 Optical Amplifiers 223
4.5.1 Introduction to Optical Amplifiers 225
4.5.2 Principle of Operation 229
4.5.3 Gain Saturation 231
4.5.4 Noise 234
4.5.5 Gain Dynamics 235
4.5.6 Optical Fiber and Semiconductor Optical Amplifiers 236
4.5.7 Raman Amplifiers 239
4.5.8 Lasers and Amplifiers 243
4.6 Optical Filters and Multiplexers 245
4.6.1 Introduction 245
4.6.2 Optical (De-)Multiplexing Devices 246
4.6.3 Overall Assessment of (De-)Multiplexing Techniques 256
4.6.4 Optical Filters 257
4.6.5 Tunable Filters 260
References 263
5 Assessing Physical Layer Degradations 267
Andrew Lord, Marcello Potenza, Marco Forzati and Erwan Pincemin
5.1 Introduction and Scope 267
5.2 Optical Power Budgets, Part I 268
5.2.1 Optical Signal-to-Noise Ratio and Q Factor 268
5.2.2 Noise 273
5.2.3 Performance Param eters. Light Path Evaluation Rules 290
5.2.4 Transmission Impairments and Enhancements: Simple Power Budgets 295
5.3 System Bandwidth 334
5.3.1 System Bandwidth, Signal Distortion, Intersymbol Interference 334
5.3.2 Fiber-Optical Nonlinear Effects 346
5.3.3 Optical Transients 356
5.4 Comments on Budgets for Nonlinear Effects and Optical Transients 362
5.4.1 Compensators/Equalizers 363
5.4.2 CD Equalization 363
5.4.3 PMD Equalization 364
5.4.4 Simultaneous Presence of Distortions, Electronic Equalization, and Cumulative Filtering 364
5.4.5 General Features of Different Modulation Formats 368
5.5 Semianalytical Models for Penalties 370
5.6 Translucent or Hybrid Networks 370
5.6.1 Design Rules for Hybrid Networks 371
viii Contents
5.7 Appendix 372
5.7.1 Dispersion Managed Links 372
5.7.2 Intrachannel Nonlinear Effects 374
References 378
6 Combating Physical Layer Degradations 381
Herbert Haunstein, Harald Rohde, Marco Forzati, Erwan Pincemin, Jonas Martensson, Anders Djupsj€obacka and Tanya Politi
6.1 Introduction 381
6.2 Dispersion-Compensating Components and Methods for CD and PMD 382
6.2.1 Introduction on Optical CD and PMD Compensator Technology 382
6.2.2 Optical Compensation Schemes 383
6.2.3 Key Parameters of Optical Compensators 387
6.2.4 Compensators Suitable for Translucent Networks 389
6.2.5 Impact of Group-Delay Ripple in Fiber Gratings 391
6.3 Modulation Formats 396
6.3.1 On–Off Keying Modulation Formats 397
6.3.2 Comparison of Basic OOK Modulation Formats: NRZ, RZ, and CSRZ for 40 Gbit/s Transmission 400
6.3.3 A Power-Tolerant Modulation Format: APRZ-OOK 408
6.3.4 DPSK Modulation Formats 412
6.3.5 Spectrally Efficient Modulation Formats 414
6.4 Electronic Equalization of Optical Transmission Impairments 416
6.4.1 Electronic Equalization Concepts 416
6.4.2 Static Performance Characterization 420
6.4.3 Dynamic Adaptation of FFE- and DFE-Structures 420
6.4.4 General Remarks 423
6.5 FEC in Lightwave Systems 424
6.5.1 Application of FEC in Lightwave Systems 424
6.5.2 Standards for FEC in Lightwave Systems 425
6.5.3 FEC Performance Characterization 426
6.5.4 FEC Application in System Design 429
6.6 Appendix: Experimental Configuration and Measurement Procedure for Evaluation and Comparison for Different Modulation Formats for 40 Gbit/s Transmission 431
6.6.1 Simulation Setup 434
Acknowledgments 435 References 435
Dictionary of Optical Networking 441
Didier Colle, Chris Matrakidis and Josep Sole-Pareta
Acronyms 465
Index 477

Preface

It is commonly accepted today that optical fiber communications have revolutionized telecommunications. Indeed, dramatic changes have been induced in the way we interact with our relatives, friends, and colleagues: we retrieve information, we entertain and educate ourselves, we buy and sell, we organize our activities, and so on, in a long list of activities. Optical fiber systems initially allowed for a significant curb in the cost of transmission and later on they sparked the process of a major rethinking regarding some, generation-old, telecommunication concepts like the (OSI)-layer definition, the lack of cross-layer dependency, the oversegmentation and overfragmentation of telecommunica­tions networks, and so on.
Traditionally, telecommunications are classified based on the physical properties of the channel; that is, fixed-line/wired-communications and wireless/radio communications. Following this classification, it can be safely a rgued that today’s core networks and metro­politan area networks (metro networks for simplicity) are almost entirely based on optical fiber systems. Moreover, the penetration of optical fiber communications in the access segment is progressing at an astonishing rate, although, quite often, it is the competition between providers, the quest for higher profits based on the established technological framework, and the legislative gridlocks that prevent an even faster adoption of this technology. Thus, a full­scale deployment of optical fiber systems in the access networks, through fixed/wireless convergence, could further reduce the role of wireless technology in transporting bandwidth over a reasonably long distance. Evidently, optical-fiber-based networks are the dominant technology, literally the backbone, of the future Internet. The fields of this technology are diverse and its engineering requires knowledge that extends from layer 1 to layer 3.
Many excellent basic text and specialized books are available today aiming to educate and/or inform scientists, engineers and technicians on the essentials in the field of optical technology. However, there is a pressing need for books presenting both comprehensive guidelines for designing fiber-optic systems and core/metro network architectures and, simultaneously, illustrating the advances in the state of the art in the respective fields. IST-NOBEL (I and II) was a large-scale research project funded from the Framework Programme 6 of the European Commission, incorporating major operators, system vendors and leading European universities. Employing a large number of experts in several fields, the project decided to collectively produce such a book as part of the disseminating activities. Thus, a considerable part of this book is based on the deliverables of IST-NOBEL with significant effort made to provide the necessary introduction of concepts and notions. The objective was to make it readable for a non-highly specialized audience, as well as to demystify the necessity behind the introduction of this or that novelty by clearly stating the underlying “need.” It is left to the readers to decide whether we have succeeded in our goals.
x Preface
The contributors to this book would like to acknowledge the immense help and support of their colleagues in the IST-NOBEL project that contributed to the preparation of the respective d eliverables. A separate, specia l, acknowledgment is for the IST-NOBEL I and II project leaders and colleagues from Telecom Italia, Antonio Manzalini, Marco Schiano, and Giuseppe Ferraris. Also, the editor is extremely grateful to Andreas Drakos and Penny Papageorgopoulou, PhD candidates in the University of Peloponnese, for their help in preparing the final manuscript.
Alexandros Stavdas
Department of Telecommunications Science and Technology
University of Peloponnese, Greece
1

The Emerging Core and Metropolitan Networks

Andrea Di Giglio, Angel Ferreiro and Marco Schiano

1.1 Introduction

1.1.1 Chapter’s Scope and Objectives

The study of transport networks is a vast and highly multidisciplinary field in the modern telecommunication world. The beginner who starts studying this technical subject may remain astonished by the variety and complexity of network architectures and technologies that have proliferated in the last decade. Even an expert in the field may get disoriented in the huge variety of networks’ functions and characteristics.
This introductory chapter is devoted to the definition of transport networks’ fundamentals representing the very basic “toolbox” of any expert in the field. Furthermore, it investigates transport network architectural evolution in terms of new network services supporting emerging users’ applications.
The chapter is structured as follows. Section 1.2 contains the definitions of the basic network concepts used throughout the book. Sections 1.3 and 1.4 describe the requirements and the architectural evolution roadmap of transport networks based on emerging users’ applications. Finally, Section 1.5 shows the economic models and analysis techniques that enable the design and realization of economically sustainable transport services.

1.2 General Characteristics of Transport Networks

For more than a century, the traditional vision of telecommunication networks has been a smart combinationof transmissionand switching technologies. Even if transmissionand switchingare still the basic building blocks of any network, telecommunication networks fundamentals cover a much broader scope nowadays. This new vision is primarily due to the introduction of digital
Chapter 1, ‘The Emerging Core and Metropolitan Networks’, Ó 2009 Angel Ferreiro and Telecom Italia S.p.A from Core and Metro Networks, edited by A. Stavdas, 2009
2 The Emerging Core and Metropolitan Networks
technologies paving the way to packet-based networks. In contrast to old analog networks, packet-based digital networks can be either connectionless or connection oriented, can have a control plane for the automation of some functions, can implement various resilience schemes, can perform a number of network services supporting users’ applications, and so on.
The essential ideas are explained in this section as a background for the entire chapter.

1.2.1 Circuit- and Packet-Based Network Paradigms

Digital networks can transfer information between nodes by means of two fundamental paradigms: circuit switching or packet switching.
.
In circuit-switched networks, data are organized in continuous, uninterrupted bit streams. In this mode of operation, a dedicated physical link between a couple of nodes is established. Before starting the data transfer on a specific connection, the connection itself must be “provisioned”; that is, the network switching nodes must be configured to provide the required physical link. This implies an exclusive allocation of network resources for the whole duration of the connection. Such a task is usually performed by dedicated elements belonging to the network control system; network resources are released when the connection ends.
This is the way that the plain old telephony service (POTS) has been working so far. The private reservation of network resources prevents other connections from using them while the first one is working, and may lead to inefficient network use.
.
In packet-switched networks, data are organized in packets of finite length that are processed one by one in network nodes and forwarded based on the packet header information. In this network scenario, each packet exploits switching and transmission devices just for the time of its duration, and these network resources are shared by all packets. This process of packet forwarding and aggregation is called statistical multiplexing and represents the major benefit of packet-switched networks with respect to the circuit-switched networks in terms of network exploitation efficiency.
Typical examples of circuit-switching and packet-switching technologies are synchronous
digital hierarchy (SDH) and Ethernet resp ectively.
Packet-switched networks can, in turn, work in connectionless or connection-oriented
network modes.
.
In the connectionless networ k mode, packets are forwarded hop by hop from source node to destination node according to packet header information only, and no transfer negotiation is performed in advance between the network nodes involved in the connection; that is, the source node, optionally the intermediate node(s) and the destination node.
.
In the connection-oriented network mode, packet transfer from source node to dest ination node is performed through defined resource negotiation and reservation schemes between the network nodes; that is, it is preceded by a connection set-up phase and a connection usage phase, followed by a connection tear-down phase.
Typical examples of packet-switched connectionless and connection-oriented network
protocols are Internet protocol (IP) and asynchronous transfer mode (ATM) respectively.
General Characteristics of Transport Networks 3
The main characteristic of the connectionless network mode is that packets are routed throughout the network solely on the base of the forwarding algorithms working in each node; hence, packet routes may vary due to the network status. For instance, cable faults or traffic overloads are possible causes of traffic reroute: in the connectionless network mode, the new route of a packet connection is not planned in advance and, in general, is unpredictable.
On the contrary, in the connection-oriented network mode, the route of any connection is planned in advance and, in the case of faults, traffic is rerouted on a new path that can be determined in advance.
Since route and rerouting have strong impacts on the quality of a packet connection, the two network modes are used for different network services depending on the required quality and the related cost.

1.2.2 Network Layering

The functions of a telecommunication network have become increasingly complex. They include information transfer, traffic integrity and survivability aspects, and network manage­ment and performance monitoring, just to mention the main ones. To keep this growing complexity under control and to maintain a clear vision of the network structure, layered network models have been developed. According to these models, network functions are subdivided into a hierarchical structure of layers. Each layer encompasses a set of homoge­neous network functions duly organized for providing defined services to the upper layer, while using the services provided by the lower layer. For example, in an Ethernet network, the physical layer provides data transmission services to the data link layer.
To define transport network architectures, it is essential to start from the description of the lowest three layers [1]: network, data link, and physical layers:
.
Network layer. The main task of the network layer is to provide routing functions. It also provides fragmentation and reassembly of data at the endpoints. The most common layer 3 technology is the IP. It manages the connectionless transfer of data across a router-based network.
.
Data-link layer. This provides frames, synchronization, and flow control. The data link layer also performs transfer of data coming from the network layer. Typical examples of data­link layers are point-to-point protocol and Ethernet MAC (medium/media access control) (IEEE 802.1xx).
.
Physical layer. The physical layer defines the transmission media used to connect devices operating at the upper layer (e.g., data link). Physical media can be, for example, copper-wire pairs, coaxial cables or, more frequently, single-mode or multimode optical fibers. The physical layer also defines modulation encoding (e.g., Manchester, 8B/10B) or topology (e.g., ring, mesh) [2]. Most common technologies implementing layer 1 functionalities are Ethernet (physical layer, IEEE 802.3xx), SDH and optical transport network (OTN).
It is commonly agreed that the Open System Interconnection (OSI) model is an excellent place to begin the study of network architecture. Nevertheless, the network technologies commercially available do not map exactly with the levels described in the OSI basic model.
4 The Emerging Core and Metropolitan Networks

1.2.3 Data Plane, Control Plane, Management Plane

The layered network models encompass all network functions related to data transfer. However, modern transport networks are often provided with additional functions devoted to network management and automatic network control. Hence, the totality of network functions can be classified into three groups named planes: the data plane, the management plane and the control plane.
The functions that characterize each plane are summarized below.
.
Data plane. The data plane aims at framing and carrying out the physical transportation of data blocks to the final destination. This operation includes all transmission and switching functions.
.
Control plane. The control plane performs the basic functions of signaling, routing and resource discovery. These are essential operations to introduce automation on high level network functions such as: connection establishment (i.e., path computation, resource availability verification and connection signaling set-up and tear-down), reconfiguration of signaled connections and connection restoration in case of network faults.
.
Management plane. The management plane performs management functions like alarm reporting, systems configuration and connection provisioning for data and control planes. The complexity of the management plane depends strongly on the availability of a control plane. For example, the management plane of traditional circuit-switched public switched telephone networks is more cumbersome than transport networks with a control plane, since, in the latter case, certain tasks (e.g., connection provisioning and restoration) are carried out by the control plane itself.

1.2.4 Users’ Applications and Network Services

The current challenge of evolving telephony-dedicated transport networks towards enhanced communication architectures is set by two fundamental trends.
First, services offered today to final users are much richer than simple telephony. User services like video telephony, video on demand, and Web browsing require an advanced terminal, typically a personal computer with dedicated software; for this reason, they will be called “user applications” or simpl y “applications” from now on.
Second, to convey these end-user applications, transport networks are relying on “network services,” which effectively refer to a number of transfer modes.
As an example, a point-to-point unprotected circuit connection at 2 Mbit/s represents a specific transfer mode. Other examples of network services are connections based on packet paradigms; for example, IP/multi-protocol label switching (MPLS), ATM or Ethernet. Today, all modern applications make reference to packet-based network services.
The idea of a transport network able to provide many different services is one of the most challenging of recent years and it will be analyzed in detail in the following chapt ers.
Network services and user applications can be provided by different actors. Network operators that own and manage the networks are typical providers of network services. Service providers sell and support user applications by means of network services supplied by network operators.
General Characteristics of Transport Networks 5
Important user application categories are:
.
multimedia triple play – voice, video and high-speed Internet;
.
data storage for disaster recovery and business continuity;
.
grid computing; that is, computing services delivered from distributed computer networks.
The last two categories, storage and grid computing, are dedicated to business company customers and research institutions. On the contrary, multimedia applications address resi­dential customers and the small office, home office.
Examples of network services are:
.
time-division multiplexing (TDM) connections and wavelength connections (e.g., leased lines);
.
Ethernet point-to-point, point-to-multipoint (p2mp) or rooted multipoint connections;
.
virtual private networks (Section 1.2.8).
Each user application is enabled by a network service characterized by specific attributes. A list of the most important ones is shown below.
.
Protocols: Ethernet and IP are the most common.
.
Bandwidth: committed peak, committed average bit-rate, excess peak and excess bit­rate [3].
.
Quality of service (QoS): regarding transport networks, this is defined by mea ns of the maximum allowed packet loss rate (PLR), the packet latency (i.e., the packet transmission delay), and jitter (latency variation); see Section 1.2.6.
.
Resilience: required connection availability (Section 1.2.5).
These service attributes are the main inputs for a network provider to design a multi-service network, in support of a number of defined applications.

1.2.5 Resilience

One of the most important features of transport networks is their ability to preserve live traffic even when faults occur. This feature is generally referred to as “resilience.”
In transport networks, resilience is usually achieved by duplication of network resources. For example, a fiber-optic link between a couple of nodes can be duplicated to assure survivability to cable breaks. Similarly, the switching matrix of an exchange node can be duplicated to guarantee service continuity in the case of electronics faults.
The way these extra resources are used depends strongly on networ k topology (rings or meshed network configurations), equipment technology (packet or circuit switching, network mode, optical transmission), and traffic protection requirements. However, the following general definitions help understanding the fundamental resilience schemes.
6 The Emerging Core and Metropolitan Networks
1. If the connections for traffic protection are organized in advance, the resilience mechanism is called “protection.” a. 1 þ1 protection (also called dedicated protection). The whole traffic of a connection
is duplicated and transmitted through two disjoint paths: the working and the protection path simultaneously. The receiving node switches between the two signals in the case of failure. The trigger of 1 þ1 protection is the received signal quality; for example, the received power level or the bit error rate (BER). Since no complex network protocols are needed, 1 þ1 protectio n works very quickly, typically within 50 ms. The drawback of this protection scheme is duplication of network resources.
b. 1: 1 protection (also called protection with extra traffic). The working connection is
protected with one backup connection using a disjoint path. The working traffic is sent over only one of the connections at a time; this is in contrast to dedicated protection, where traffic is always bridged onto two connections simultaneously. Under normal conditions (no network failure) the protecting connection is either idle or is carrying some extra traffic (typically best-effort traffic). Configuring 1: 1 protection depends on the control plane’s ability to handle extra traffic, that is, whether it supports the preemption of network resources for allocating them to the working traffic once it has been affected by the failure. The ingress node then feeds the working traffic on the protecting connection in the case of failure. The trigger of 1: 1 protection is the reception of network failure notification messages. Protection with extra traffic has two main drawbacks: the need to duplica te working traffic resources onto the protection path and, in the case of resource contention, the possibility that extra traffic may be interrupted without effective need.
c. M: N protection (also called shared protection). M working connections are protected
by N backup connections on a disjoint path (N M). The traffic is no longer duplicated because backup connections can carry traffic initially transported by any one of the working connections in the case of fault. Thus, switching to backup connections requires first knowing their availability and then performing traffic switching. Signaling is needed for failure notification and backup connection activation. Once failure has been repaired, traffic is reassigned to the working connection and the resources of the backup connection are available again for protection. In any case, this protection mechanism allows resource savings with respect to 1 þ1 protection.
Both protection mechanisms, dedicated and shared, are used in rings and meshed network configurations. The main advantage of protection is its quick operation, since the backup path is predefined and network resources are pre-allocated.
2. Alternatively to protection, restoration is the resilience mechanism that sets up new backup
connections after failure eventsby discovering, routing, and setting up new links “on the fly” among the network resources still available after the failure. This is achieved by the extension of signaling, routing, and discovery para digms typical of IP networks. In fact, to restore a connection, switching nodes need to discover the network topology not affected by the failure, thus allowing one to compute a set of candidate routes, then to select a new route, and to set up the backup connections. Discovery, routing algorithms, and signaling functions embedded in commercial IP/MPLS routers can quite easily implement restora­tion. On the other hand, transport network equipment needs a dedicated control plane to perform such functions.
General Characteristics of Transport Networks 7
Table 1.1 Indicative figures for network availability
Availability (%) N-Nines Downtime time (minutes/year)
99 2-Nines 5000
99.9 3-Nines 500
99.99 4-Nines 50
99.999 5-Nines 5
99.9999 6-Nines 0.5
Usually, the resilience level of a network service (e.g., a leased line or an Ethernet connection, as defined in Section 1.2.4) is made precise through a number of parameters; the most important are:
.
Mean time to failure (MTTF): the reciprocal of the failure rate, for systems being replaced after a failure.
.
Mean time to repair (MTTR): this depends on the repair time of a network fault.
.
Mean time between failures (MTBF): this is the sum of MTTF and MTTR and defines the mean time interval between successive failures of a repairable system ; it is a measure of network component reliability.
.
Maximum recovery time: this is the maximum delay between a failure injuring a network service and the restoration of the service over another path; in other words, the maximum time during which the network service is not available. It accounts for MTTR and all other possible delays affecting complete system recovery (signaling, rerouting).
The same concept can be given a different flavor, insisting on network status instead of
duration:
.
Unavailability: the probability that the network service is not working at a given time and under specified conditions; it is the ratio MTTR/MTBF. Some indicative numbers for network availability are illustrated in Table 1.1.

1.2.6 Quality of Service

Network services are characterized by a set of parameters that define their quality (QoS).
.
BER: this is a physical-layer parameter, manifesting the fraction of erroneous bits over the total number of transmitted bits. It is closely related to design rules applied to the physical layer transport network. It is studied in detail in Chapter 3.
.
PLR: in packet-switched services, this is the fraction of data packets lost out of the total number of transmitted packets. Packets can be dropped due to congestion, or due to transmission errors or faults.
.
Latency: the time needed for carrying data from the source node to the destination node. Latency is caused by the combination of signal propagation delay, data processing delays, and queuing delays at the intermediate nodes on the connection [3].
.
Latency variation: the range of variation of the latency mainly due to variable queuing delays in network nodes or due to data segmentation and routing of data blocks, via different physical paths (a feature readily available in next-generation (NG)-synchronous optical network (SONET)/SDH). Also, queuing delay variations may occur in the case of traffic
8 The Emerging Core and Metropolitan Networks
overload in nodes or links. An excess of latency variation can cause quality degradation in some real-time or interactive applications such as voice over IP (VoIP) and video over IP (IP television (IPTV)).
.
Service unavailability: this has already been defined in Section 1.2.5.
For connection-oriented network services, the definition of QoS also includes:
.
Blocking probability: the ratio between blocking events (failure of a network to establish a connection requested by the user, because of lack of resources) and the number of attempts.
.
Set-up time: delay between the user application request time and the network service actual delivery time.
Current packet-based networks are designed to satisfy the appropriate level of QoS for different network services. Table 1.2 shows suitable values of QoS parameters for the main users’ applications. As an example, applications like voice or videoconference need tight values of latency and latency variation. Video distribution is more tolerant to latency variation, but it needs low packet loss, since lost packets are not retransmitted. File transfer (e.g., backup) does not have strong requirements about any QoS parameters, since the only requirement is to transfer a pre-established amount of data in a fixed time interval.
1.2.7 Traffic Engineering
In complex meshed networks, careful traffic engineering (TE) and resource optimization is a mandatory requirement providing network management and operation functions at reasonable capital expenditure (CAPEX) and operational expenditure (OPEX). Towards this end, the use of conventional algorithms to set up the working and protection (backup) paths and for traffic routing within the network is insufficient. Toaddress this problem, use is made of TE, which is a network engineering mechanism allowing for network performance optimization by means of leveraging traffic allocation in conjunction with the available network resources.
The purpose of TE is to optimize the use of network resources and facilitate reliable network operations. The latter aspect is pursued with mechanisms enhancing network integrity and by embracing policies supporting network survivability. The overall operation leads to the minimizations of network vulnerability, service outages due to errors, and congestions and failures occurring during daily network operations. TE makes it possible to transport traffic via reliable network resources, minimizing the risk of losing any fraction of this traffic.
TE leverages on some instruments that are independent of the network layer and technology:
.
A set of policies, objectives, and requirements (which may be context dependent) for network performance evaluation and performance optimization.
.
A collection of mechanisms and tools for measuring, characterizing, modeling, and efficiently handling the traffic. These tools allow the allocation and control of network resources where these are needed and/or the allocation of traffic chunks to the appropriate resources.
.
A set of administrative control parameters, necessary to manage the connec tions for reactive reconfigurations.
General Characteristics of Transport Networks 9
Table 1.2 QoS characterization of users’ applications
User application QoS
Max.
latency
(ms)
Storage
Backup/restore N.A. N.A. 0.1 min 99.990 Storage on demand 10 1 0.1 s 99.999 Asyncrhonous mirroring 100 10 0.1 s 99.999 Synchronous mirroring 3 1 min 99.999
Grid computing
Compute grid 100 20 0.0 s 99.990 Data grid 500 100 0.1 s 99.990 Utility grid 200 50 0.0 s 99.999
Multimedia
Video on demand (enter-
tainment quality, similar to DVD)
Video broadcast (IP-TV),
entertainment quality
similar to DVD Video download 2–20 s 1000 1.0 s 99.990 Video chat (SIF quality, no
real-time coding penalty) Narrowband voice, data
(VoIP, ...) Telemedicine (diagnostic) 40–250 5-40 0.5 ms 99.999 Gaming 50–75 10 5.0 s 99.500 Digital distribution, digital
cinema Video conference (PAL
broadcast quality 2.0
real-time coding penalty)
Note: latency is expressed in milliseconds with the exception of video on demand, video broadcast, and video download, where seconds are the unit.
2–20 s 50 0.5 s 99.500%
2–20 s 50 0.5 s 99.500
400 10 5.0 s 99.500
100–400 10 0.5 ms 99.999
120 80 0.5 s 99.990
100 10 0.5 99.990
Max. latency
variation (ms)
Packet loss
(layer 3)
(%)
Max.
set-up
time
Min.
availability
(%)
The process of TE can be divided into four phases that may be applied both in core and in metropolitan area networks, as described by the Internet Engineering Task Force (IETF) in RFC 2702 [4]:
.
Definition of a relevant control policy that governs network operations (depe nding on many factors like business model, network cost structure, operating constraints, etc.).
.
Monitoring mechanism, involving the acquisition of measurement data from the actual network.
.
Evaluation and classification of network status and traffic load. The performance analysis may be either proactive (i.e., based on estimates and predictions for the traffic load, scenarios
10 The Emerging Core and Metropolitan Networks
for the scheduling of network resources in order to prevent network disruptions like congestion) or reactive (a set of measures to be taken to handle unforese en circumstances; e.g., in- progress congestion).
.
Performance optimization of the network. The performance optimization phase involves a decision process, which selects and implements a set of actions from a set of alternatives.

1.2.8 Virtual Private Networks

Avirtual private network (VPN) is a logical representation of the connections that makes use of a physical telecommunication infrastructure shared with other VPNs or services, but main­taining privacy through the use of tunneling protocols (Section 1.2.9) and security procedures. The idea of the VPN is to give a user the same services accessible in a totally independent network, but at much lower cost, thanks to the use of a shar ed infrastructure, rather than a dedicated one [5].
In fact, a common VPN application is to segregate the traffic from different user communi­ties over the public Internet, or to separate the traffic of different service providers sharing the same physical infrastructure of a unique network provider.
VPNs are a hot topic also in the discussion within standardization bodies: different views exist on what a VPN truly is.
According to ITU-T recommendation Y.1311 [6] a VPN “provides connectivity amongst a limited and specific subset of the total set of users served by the network provider. A VPN has the appearance of a network that is dedicated specifically to the users within the subset.” The restricted group of network users that can exploit the VPN services is called a closed user group.
The other standardization approach, used by the IETF, is to define a VPN’s components and related functions (RFC 4026, [7]):
.
Customer edge (CE) device: this is the node that provides access to the VPN service, physically located at the customer’s premises.
.
Provider edge (PE) device: a device (or set of devices) at the edge of the provider network that makes available the provider’s view of the customer site. PEs are usually aware of the VPNs, and do maintain a VPN state.
.
Provider (P) device: a device inside the provider’s core network; it does not directly interface to any customer endpoint, but it can be used to provide routing for many provider­operated tunnels belonging to different customers’ VPNs.
Standardization bodies specified VPNs for different network layers. For example, a transport layer based on SDH can be used to provide a layer 1 VPN [8, 9]. Layer 2, (e.g., Ethernet) allows the possibility to implement L2-VPN, also called virtual LAN (VLAN). Layer 3 VPNs are very often based on IP, and this is the first and the most common VPN concept.
In some situations, adaptation funct ions between the bit-stream that is provided from the “source” (of the applications) and the VPN are required. An example of an adaptation data protocol function is the mapping of Ethernet frames in NG-SDH containers.
General Characteristics of Transport Networks 11

1.2.9 Packet Transport Technologies

Packet technologies have been dominating the local area network (LAN) scenario for more than 25 years, and nowadays they are widely used also in transport networks, where many network services are based on packet paradigms. The main reason for this success is twofold: first, the superior efficiency of packet networks in traffic grooming due to the statistical aggregation of packet-based traffic; second, the inherent flexibility of packet networks that can support an unlimited variety of users’ applications with a few fundamental network services, as shown in Section 1.2.4.
However, until now, the transport of packe t traffic has been based on the underlying circuit­switch ed technology already available for telephony. A typical example is represented by Ethernet transport over NG-SDH networks. This solution is justified by the widespread availability of SDH equipment in already-installed transport networks, and by the excellent operation, administration, and maintenance (OAM) features of such technology. These features are fundamental for provisioning packet network services with the quality required for most users’ applications, but they are not supported by the LAN packet technologies.
This situation is changi ng rapidly, because a new generation of packet-based network technologies is emerging. These new scenarios combine the efficiency and flexibility of packet networks with the effective network control and management features of circuit-based networks. These new technologies are referred to as packet transport technologies packe t transport technology (PTTs).
There are proposals for introducing tunnels engineering features rendering it into a connection-oriented platform. These developments are currently under standardization at IEEE and ITU-T where is known as Provider Backbone Bridge with Traffic Engineering (or simply PBB-TE).
An alternative approach under standardization at the ITU-T and IETF is to evolve the IP/MPLS protocol suites to integrate OAM functions for carrier-grade packet transport networks.
This PTT, known as MPLS-TP (MPLS transport profile) includes features traditionally associated with transport networks, such as protection switching and operation and mainte­nance (OAM) functions, in order to provide a common operation, control and management paradigm with other transport technologies (e.g., SDH, optical transport hierarchy (OTH), wavelength-division multiplexing (WDM)).
The trend imposed by the dramatic increase of packet traffic and the obvious advantages in evolving existing circuit-switched networks into advanced packet-switched networks is going to make PTTs a viable solution to building a unified transport infrastructure, as depicted in Figure 1.1. Incumbent network operators that have already deployed a versatile NG-SDH network for aggregated traffic may follow conservative migration guidelines for their core networks and keep circuit solutions based on optical technologies. These plausible solutions are discussed in Section 1.4.
1
facilitating to allow Ethernet attaining traffic
1
A tunnel is a method of communication between a couple of network nodes via a channel passing through intermediate
nodes with no changes in its information content.
12 The Emerging Core and Metropolitan Networks
Figure 1.1 Unified transport network

1.3 Future Networks Challenges

1.3.1 Network Evolution Drivers

In the past decade, the proliferation of electronic and fiber-optic technologies has allowed network services to evolve from the exclusive support of plain telephony to an abundance of services which are transported based on the IP. These advances have had a major impact on the drivers for network evolution.
Nowadays, network design and planning is the outcome of the interplay between different
technological, legal, and economic drivers:
.
Introduction of new services. A network operator or a service provider can decide to offer new services based on customers’ requests or market trends.
.
Traffic growth. The growing penetration and the intensive use of new services increase the network load.
.
Availability of new technologies. Electronic, optical, and software technologies keep on offering new advances in transmission, switching, and control of information flows based on circuits and packets.
.
Degree of standardization and interoperability of new network equipment. Modern networks are very complex systems, requiring interaction of various kinds of equipment by means of dedicated protocols. Standardization and interoperability are key requirements for a proper integration of many different network elements.
.
Laws and regulati ons. National laws and government regulations may set limitations and opportunities defining new business actors for network deployment and usage.
.
Market potential and amount of investments. The financial resource availability and the potentialof the telecommunicationmarket are thekeyeconomic drivers for networkdevelopment.
1.3.2 Characteristics of Applications and Related Traffic
In this section, the association between applications and network services is presented. The starting point of the analysis is the bandwidth requirement (traffic) of the various applications
Future Networks Challenges 13
and the subsequent classification of this traffic into classes. Figure 1.2, illustrates a classifica­tion of user applications based on the following traffic characteristics:
.
elasticity
.
interactivity
.
degree of resilience (availability)
.
symmetry
.
bandwidth.
Elastic
Traffic Classes
Inelastic
non interactive
interactive
Non interactive
interactive
standard availability
high availability
standard availability
high availability
standard availability
high availability
standard availability
high availability
asymmetrical
symmetrical
asymmetrical
symmetrical
asymmetrical
symmetrical
asymmetrical
symmetrical
asymmetrical
symmetrical
asymmetrical
symmetrical
asymmetrical
symmetrical
downloding
remote backup
P2P file exchange mail
tele-diagnostics medical data storage network supervision
web browsing compute grid telnet
data grid utility grid
gambling
network control
low bandwidth
high bandwidth
low bandwidth
high bandwidth
low bandwidth
high bandwidth
low bandwidth
high bandwidth
radio broadcast Live radio
video on demand video broadcast live TV asynchronous mirroring storage on demand tele-vigilance
voice over IP
video chat
remote surgery synchronous mirroring real time compute grid digital distribution
digital cinema distribution telephony IP telephony gaming
video-conference paramedic communications emergency communications
Figure 1.2 Classification of traffic generated by reference applications
14 The Emerging Core and Metropolitan Networks
Table 1.3 Qualitative classification of traffic types
Elastic Inelastic
Interactive Transactional Real time Noninteractive Best effort Streaming
Elasticity refers to the level up to which the original traffic shape can be modified; the two main categories are as follows:
.
Inelastic traffic (or stream traffic) is generated by applications whose temporal integrity overwhelms data integrity because they try to emulate virtual presence.
.
Elastic traffic is generated by applications where data integrity overwhelms temporal integrity, therefore being rather tolerant to delays and being able to adapt their data generation rate to network conditions.
The term interactivity refers to a mode of operation characterized by constant feedback and an interrelated traffic exchange between the two endpoints of the connection.
To map users’ applications traffic into the appropriate network services, it is essential to define a few classes of traffic patterns that share the main characteristics. For this purpose, Table 1.3 defines four kinds of traffic patterns in terms of QoS requirements.
Another important task is to assign QoS parameters quantitatively to the traffic classes. Table 1.4 sets the values of QoS parameters used to define four basic classes as:
.
real-time traffic
.
streaming traffic
.
transactional traffic
.
best-effort traffic.
In connection with Table 1.4, the term dynamicity refers to the ability of a user to modify the parameters of an existing connection. It is the only important parameter not described in Section 1.2.6, since it is not addressed directly by the classic QoS definition, but it is anyway an important quantity for application classification. The dynamicity refers to the time variation of the following connection characteristics:
.
bandwidth (bit-rate);
.
QoS parameters (latency, availability, data integrity);
.
connectivity (the end-points of the connection).
The level of dynamicity is quantified on a three-state base:
– “none” (it is not possible to modify any parameters of an existing connection); – “bit-rate and QoS” (when only these two parameters can be altered); – “full” (bit-rate, QoS parameters, and connectivity modifications are allowed).
As seen in connectio n with Table 1.4, four traffic categories are defined based only on QoS parameters. Table 1.5 shows examples of applications belonging to each one of the four classes identified above, having different bandwidth requirements.
Table 1.4 Quantitative classification of QoS for traffic classes
Blocking
probability (%)
Real time <0.1 >99.995 <1 <50 ms
Streaming <0.1 >99.99 <1 <1s Transactional <1 >99.9 <3 <1s <200 Bit-rate and QoS <1 E-2 Best effort

Network
availability (%)
Set-up
time (s)
Max.
latency
Mean
latency
(ms)
Max.
latency
variation
Dynamicity Packet loss rate
Bit-rate, QoS and
<5 E-5
connectivity
None <1 E-3
Not applicable
Future Networks Challenges 15
16 The Emerging Core and Metropolitan Networks
Table 1.5 Traffic characterization based on bandwidth (BW) and QoS parameters and map of users’
applications
QoS Low BW (tens of kbit/s) Medium BW (<2 Mbit/s) High BW (>2 Mbit/s)
Real time Legacy and IP telephony Gaming Video conference, grid
computing
Streaming UMTS Remote backup, network
supervision Transactional E-commerce Telnet SAN Best effort E-mail, domotic, VoIP p2p file exchange, data
acquisition
a
Video on demand.
b
Storage area network.
TV and video broadcast,
a
Vo D
b
p2p file exchange, data
acquisition
Table 1.5 is useful to map most common users’ applications into the four traffic classes (real-
time, streaming, transactional, best-effort), taking also the bandwidth use into account.
Similar to the classification of user applications, network services are classified into five categories in association with the network services reported in Section 1.2.4. Thus, the network service map looks as follows:
.
L1 VPN, which provides a physical-layer service between customer sites belonging to the same closed user group. These VPN connections can be based on physical ports, optical wavelengths, or TDM timeslots.
.
L2 VPN, which provides a service between customer terminals belonging to the VPN at the data link layer. Data packet forwarding is based on the information contained in the packets’ data link layer headers (e.g., frame relay data link circuit identifier, ATM virtual circuit identifier/virtual path identifier, or Ethernet MAC addresses).
.
L3 VPN, which provides a network layer service between customer devices belonging to the same VPN. Packets are forwarded based on the information contained in the layer 3 headers (e.g., IPv4 or IPv6 destination address).
.
Public IP, which is considered as the paradigm of best-effort network services. Namely, it is a generalized L3 VPN without restrictions to the user group, but with a consequently poor QoS.
.
Business IP, which is included as a higher priority class that, for instance, can efficiently handle latency
2
in ti me-sensitive applications.
On top of this classification, further “orthogonal” categorizations are often introduced. VPN services are further subdivided into:
permanent VPNs, to be provided on a permanent basis by the network service
provider;
on-demand VPNs, which could be controlled dynamically by the client user/network.
2
See latency and other QoS defining parameters later in this section.
Future Networks Challenges 17
Table 1.6 Mapping network services groups to some applications (BW: bandwidth)
L1 and L2 VPN services are also classified into high- and low-availability services. Table 1.6 provides a mapping between “user applications” and “network services”: in this context, a stippled box means that that particular application may run over on this network service, but not very efficiently. The most efficient support to that application is designated with horizontal rows, whereas a white box should be interpreted as no support at all from this service to that application.

1.3.3 Network Architectural Requirements

This section gives an overview of the architectural requirements for transport networks supporting the services described above.
1.3.3.1 Network Functional Requirements
From an architectural point of view, data services have been traditionally transported over a wide set of protocols and technologies. For example, IP services are transported over the core network usually relying on SDH, ATM, or Ethernet transmission networks. A widespread alternative used in current convergent transport networks is to go towards a meshed network of IP/MPLS routers, interconnected through direct fiber or lambda connections and without any multilayer interaction.
This “IP-based for everything” approach was proved to be valid for the last decade, but with current traffic trends it would lead to scalability problems. Currently, backbone nodes need switching capacities of several terabits per second, and this need is predicted to double every 2 years. Routers are also very expensive and they are not optimized for high-bandwidth traffic transportation, while transport technologies such as SONET/SDH are not efficient enough for packet transport, due to a very coarse and not flexible bandwidth granularity.
18 The Emerging Core and Metropolitan Networks
On the other hand, a number of emerging services (e.g., new multimedia applications served over the Internet; i.e., real-time high-bandwidth video services) are imposing new require­ments on the current “IP-based for everything” architecture in terms of bandwidth and QoS (end-to-end delay and availability). Moreover, mobility of users and devices and new traffic profiles (due to, for example, flash crowds and streaming services) require a network with an unprecedented dynamicity that is able to support unpredictable traffic patterns.
1.3.3.2 Network Scalability
The term scalability is a feature of a network architecture designating the ability to accommodate higher traffic load without requiring large-scale redesign and/or major deploy­ment of resources. A typical (negative) example manifesting lack of scalability is an SDH ring where additional resources and manual configurations are mandatory in order to increase the capacity between two nodes. Thus, future transport networks should be scalable in order to support existing or yet-unknown clients and traffic volumes.
The lack of scalability is demonstrated in two distinctive ways. First, by means of an excessive deployment of network resources to accommodate higher traffic volumes. This inefficiency is leading to higher CAPEX and OPEX that are mainly attributed to the enduring very high cost of switching. Solving this issue requires the deployment of technologies able to transport traffic with a lower cost per bit. Second, it is associated with architectural and/or control plane scalability restrictions due to the excessive number of network elements to control (e.g., the number of paths in the network). To address this issue requires the adoption of layered architectures and aggregation hierarchies.
1.3.3.3 Network Reconfiguration Ability
Network reconfiguration ability refers to the ability of the network to change the status of some or all of the established connections, to modify the parameters of these connections (e.g., modify the amount of allocated bandwidth) or to modify the way the services are provided (for instance, changing the routing of a given connection to allow more efficient grooming on a different route or improve spare capacity sharing).
The interest in having a reconfigurable network comes from the fact that traffic profiles change very frequently, may be fostered by symmetrical traffic patterns, unexpected traffic growth, possible mobile data/multimedia services, varied geographic connectivity (e.g., home, work), andemerging services, suchas user-generated content. All these facts make it reasonable to think in the future about a highly varying traffic profile in a network, thus meaning that reconfigurability would be a highly advantageous characteristic in data architectures.
1.3.3.4 Cost Effectiveness
Taking into account the fierce competition and the pressure upon network operators in the telecommunications market, as well as the descending cost per bit charged to the final user, the only solution for service providers to keep competitive is to reduce traffic transport costs. Therefore, cost effectiveness is the obvious requirement for any new technology. Basic
Future Networks Challenges 19
approaches to achieve this cost reduction are to build networks upon cheap scale-economy technologies, adapted to the applications’ bursty data traffic and specifically designed to keep functional complexity to a minimum. To facilitate this cost per bit reduction even in presence of unp redictable traffic growth, modular solutions are of paramount importance.
1.3.3.5 Standardized Solutions
Standardization of solutions is a key point, because it assures interoperability of equipment from different manufacturers and, as a consequence, it allows a multi-vendor environment.
This leads to the achievement of economies of scale that lower costs, since a higher number of suppliers use the same technology. Besides, standardization allows network operators to deploy networks with components from different suppliers, therefore avoiding dependence on a single manufacturer, both from a technological and an economical point of view.
1.3.3.6 Quality of Service Differentiation
As specified in Sections 1.2.6 and 1.3.2, a differentiating feature between the various applications consists in their dissimilar transport network requirements (e.g., minimum/ maximum bandwidth, availability, security, delay, jitter, loss, error rate, priority, and buffer­ing). For this reason, networks have to support QoS differentiation because their main goal is to assure a proper multi-service delivery to different applications. The intention of QoS specifications is to utilize network mechanisms for classifying and managing network traffic or bandwidth reservation, in order to deliver predictable service levels such that service requirements can be fulfilled.
1.3.3.7 Resilience Mechanisms
As reported in Section 1.2.5, an important aspect that characterizes services offered by telecommunication networks is service availability. Resilience mechanisms must be present in order to react to network failures, providing backup solutions to restore the connections affected by the failure. Typical resilience mechanisms provide full protection against all single failures; they distinguish in terms of how fast restoration is provided and on the amount of backup capacity required for protection, to fully suppor t this single-failure event. Resilience schemes can also be characterized depending on their ability to provide various level of protection (e.g., full protection against single failures, best effort protection, no-protection, and preemption in the case of failure) and on their capability to provide very high availability services (e.g., full protection against multiple failures). For transport network clients, the important aspect is the resulting service availability, measured in terms of average service availability over a given period of time (e.g., 1 year) and of maximum service interruption time.
1.3.3.8 Operation and Maintenance
A fundamental requirement is to keep a proper control over the networking infrastructure: easy monitoring, alarm management, and configuration tools are required. The current trend for
20 The Emerging Core and Metropolitan Networks
OPEX reduction and maintenance simplification leads towards automated distributed cont rol maintenance and operations.
Transport technologies or carrier-grade switching and transmission solutions differ from other technologies in the OAM features: it is important not only in adm inistrating and managing the network, but also to provide services and to deal with its customers. Efficient operation tools and mechanisms must also be implemented within the transport networks.
Finally, it is important to consider the interoperability between different network layers that requires mutual layer independence; for this reason, the transport technology needs to be self­sufficient to provide its own OAM, independently of its client and server layers.
1.3.3.9 Traffic Multicast Support
A multicast transfer pattern allows transmission of data to multiple recipients in the network at the same time over one transmission stream to the switches.
A network with multicast capability must guarantee the communication between a single sender and multiple receivers on a network by delivering a single stream of information to multiple recipients, duplicating data only when the multiple path follows different routes. The network (not the customer devices) has to be able to duplicate data flows. There are only two degrees for the ability to support multicast transfer: able or unable (multicast is an on/off property).
Multicast distribution is considered a useful tool for transport technologies when dealing with IPTV and similar applications. However, it is pointed out that layer 2 multicasting is not the only solution to distribute IPTV.
1.3.3.10 Multiplicity of Client Signals
Previous sections highlighted that metro-core networks are supporting traffic from many different applications, such as business data, Web browsing, peer to peer, e-Business, storage networking, utility computing, and new applications such as video streaming, video confer­ence, VoIP, and tele-medicine applications. The prevalence of multimedia services and the expansion of triple-play has an important role in traffic load and distribution in metro and core networks. A strong increase of broadband access penetration, based on a combination of different fixed and mobile access technologies, is expected for the next years, favoring the increase of terminal nomadism, which might introduce a more variable and unpredictable traffic, especially in the metro area. On the other side, corporate VPN services ranging from MPLS-based VPNs [10] to legacy services cope with the business telecom market.
From a technological standpoint, most services are migrating to packet-based Ethernet framing. This trend makes it mandatory for Core/Metro networks to support Ethernet client services. Nevertheless, many legacy networks are still based on other standards, such as SDH and ATM, and they still need to suppor t these kinds of technology.
A transport infrastructure that can carry traffic generated by both mobile and fixed access is an important challenge for future transport networks.
Fixed and mobile applications present similar QoS requirements, and can be classified according to the four classes previously defined in Section 1.2.4. (i.e., best-effort, streaming,
Future Networks Challenges 21
real-time, and transactional). However, current bandwidth requirements are lower for mobile applications than for fixed applications due to limitations in wireless access bandwidth and terminal screen size and resolution.
1.3.3.11 Transport Network Service Models and Client Interactions
Telecom networks have been upgraded with different network layer technologies, each providing its own set of service functionality based on its own switching paradigm and framing architecture. The GMPLS (Generalized Multi-Protocol Label Switching) protocol architecture paves the way for a convergence between transport and client networks reducing, thus, the overall control and management complexity. GMPLS can be configured to handle networks with dissimilar switching paradigms (on data plane) and different network manage­ment platforms (on control and management plane). This is made feasible by means of LSPs (Label Switched Paths) that are established between two end points. i.e. under the GMPLS protocol architecture the resources of the optical transport network are reserved based on the connectivity requests from a client packet-switched network.
The Overlay Model
The overlay model refers to a business model in which carriers or optical backbone (bandwidth) providers lease their network facilities to Internet service providers (ISPs). This model is based on a client–server relationship with well-defined network interfaces (or automatic switched optical networ k (ASON) reference points) between the transport network involved and client networks. The overlay model mandates a complete separation of the data client network control (that could be IP/MPLS based) and the transport network control plane (e.g., wavelength-switched optical networks/GMPLS). A controlled amount of signaling and restricted amount of routing messages may be exchanged; as a consequence, the overlay model is a very opaque paradigm. The IP/MPLS routing and signaling controllers are independent of the routing and signaling controllers within the transport domain, enabling the different networks to operate independently. The independent control planes interact through a user­to-network interface (UNI), defining a client–server relationship between the IP/MPLS data network and the wavelength-switched optical network (WSON)/GMPLS transport network.
Overlay network service models support different business and administrative classes (as developed in Section 1.5.3.) and preserve confidentiality between network operators. The connection services are requested from client networks to the transport network across distinct UNIs. When a connection is established in the transport network for a given client network, this connection can be used as a nested LSP or a stitched LSP to support the requirements of the client network.
The service interface in the overlay network model can be configured according to the level of trust of the two interacting structures. The interface can be based on a mediation entity such as an operation service support (OSS) or it can use the northbound interface of the network management system. Further, the interface between client network (higher layer network) and transport network (lower layer network) can operate a GMPLS signaling protocol, such reservation protocol with TE (RSVP-TE).
22 The Emerging Core and Metropolitan Networks
Peer Model
Compared with the overlay model, the peer model is built on a unified service representation, not restricting any control information exchanged between the transport network and the clients. This model is relevant and represents an optimal solution when a transport network operator is both an optical bandwidth provider and an ISP. In this case, the operator can optimally align the virtual topologies of its transport network with the network services required by its data network. The IP/MPLS control plane acts as a peer of the GMPLS transport network control plane, implying that a dual instance of the control plane is running over the data network (say, an IP/MPLS network) and optical network (say, a WSON/GMPLS network). The peer model entails the tightest coupling between IP/MPLS and WSON/GMPLS compo­nents. The differentnodes are distinguished by their switching capabilities; for example, packet for IP routers interconnected to photonic cross-connects (PXCs).
Integrated Model
Compared with the peer model, the integrated model does not require different service interfaces between the different networks. The integrated model proposes the full convergence of data network control plane and transport network control plane. All nodes are label-switched routers (LSRs) all supporting several switching capabilities; say, wavelength, SDH and Ethernet. Each LSR is also able to handle several orders of the same switching capability as it happens; for example, with SDH. An LSR embeds one GMPLS control plane instance and is able to control simultaneously different switching capability interfaces. Only this model can handle a complete and global optimization of network resource usages through transport and client networks.
Augmented Model
The augmented model considers that the network separation offered by the overlay model provides a necessary division betwee n the administrative domains of different network service providers, but also considers that a certain level of routing information should be exchanged between the transport network and the client networks. In a competitive environment, a complete peer network service model is not suitable because of the full exchange of topology information and network resource status between client and server optical networks, imposing on a network operator to control the resources of the client data networks and triggering the scalability issues of the management functions.
The augmented model provides an excellent support for the delivery of advanced connec­tivity services as might be offered from a multilayer network (MLN)/multiregion network (MRN). The capability, such as wavelength service on demand, integrated TE or optical VPN services, can require controlled sharing of routing information between client networks and the optical transport network.
User-to-network Interface
The UNI is a logical network interface (i.e., reference point) recommended in the “Requirements for Automatic Switched Transport Network” specification ITU-T G.807/Y.1302. The UNI defines the set of signaling messages that can be exchanged between a client node and a server node; for instance, an IP router and an SDH optical cross-connect (OXC) respectively. The server node provides a connection service to the client node; for example, the IP router can request TDM LSPs from its packet over SONET (PoS) interfaces. The UNI supports the exchange of auth entication, authorization, and connection admission
Future Networks Challenges 23
control messages, and provides the address space set of the reachable nodes to the client network. Different versions of the implementation agreement for a UNI have been produced by the Optical Internetworking Forum (OIF) since October 2001 as OIF UNI 1.0. The different OIF implementation agreement versions are supporting the overlay service model as well as the augmented service model. The signaling messages exchanged between the client node and the server node are focused on the LSP connection request, activation, deactivation, and tear down. The IETF specifies a GMPLS UNI that is also applicable for a peer model. Fully compliant with RSVP-TE, GMPLS UNI allows the end-to-end LSP handling from the ingress customer edge equipment to the egress customer edge equipment and at each intermediate LSR involved in the signaling sessions.
Network-to-network Interface
The network-to-network interface (NNI) is a logical network interface (i.e., reference point) recommended in the “Requirements for Automatic Switched TransportNetwork” specification ITU-T G.807/Y.1302.The NNI defines the set of both signaling messages and routing messages that can be exchanged between two network server nodes; for example, SONET OXC and SDH OXC. There are two types of NNI, one for intranetwork domains and one for internetwork domains: an external NNI (E-NNI) and an internal NNI (I-NNI) respectively.
.
The E-NNI assumes an untrusted relationship between the two network domains. The routing information exchanged between the two nodes located at the edge of the transport
Figure 1.3 Customer nodes to public network link through server nodes by the UNI as defined in ITU architecture
24 The Emerging Core and Metropolitan Networks
network specified within the E-NNI is restricted. The control messages exchanged include reachable network addresses that are usually translated, authentication and connection admission control messages, and a restricted set of connection requests of signaling messages.
.
The I-NNI assumes a trusted relationship between two network domains. The control information specified within the I-NNI is not restricted. The routing control messages exchanged include topology, TE link state, and address discovery. The signaling messages can allow controlling of the resources end to end between several network elements and for each LSP and its protection path.

1.3.4 Data Plane, Control Plane, and Management Plane Requirements

1.3.4.1 Data Plane Requirements
The challenges of the physical layer part of the data plane are covered in Chapters 3–6. In this section, two conceptual challenges of the data plane are addressed, namely the quest for transparency and the search for novel transport formats.
Transparency
During the last 20 years the cornerstone of transport network evolution has been the notion of “transparency.” Today, there are two distinctive understandings of the term “transparency”: bit-level transparency and optical transparency.
In the original idea, the two meanings were synonymous and they were based on the following simple concept. The advent of the erbium-doped fiber amplifier facilitated the proliferation of WDM (these topics are discussed in Chapters 3–6), which increased the product “bandwidth times length” to about two to three orders of magnitude, an event that eventually led to a significant curb of the transmission cost. As a result of this evolution, the cost of switching started dominating (and still does) the cost of a transport network. At the same time, there was a clear disparity between the data throughput that the fiber-optic systems could transmit and the amount of data that synchronous systems (SDH/SONET) could process, a phenomenon that was termed an “optoelectronic bottleneck.” For these reasons, every effort was made to minimize electronic switching wherever possible, making use of optical­bypassing concepts for the transit traffic, namely avoiding transport schemes requiring frequent aggregation and grooming of the client signals through electronic switches.
The widespread deployment of applications based on the IP and the emergence of a “zoo” of other protocols (like Ethernet, ESCON, Fiber Channel, etc.) gave a renewed impetus to incentivize transparency in transport networks.
In mor e recent times, transparency went colored of different shades.
Bit-level Transparency
The transport network (OTN in particular) should convey client signals with no processing of the information content. This will minimize the aggregation/grooming used throughout the network, whilst it will provide a client/provider agnostic transportation (transparent bit­mapping into the transport frame). Here, transparency mainly indicates service transparency; that is, the minimization of bit-by-bit processing regardless of the technological platform that is used. This definition is shifting the interest from technologies to functions; hence, both
Future Networks Challenges 25
all-optical and optoelectronic subsystems are of equal interest in building service-transparent networks.
Optical Transparency
Nevertheless, in conjunction with bit-level transparency, the initial notion of a transparent network is still of interest, where the optical/electrical/optical (O/E/O) conversions are minimized so the signal stays in the optical domain. The benefits from the reduc tion in the number O/E/O conversions include:
.
reduction of a major cost mass by minimizing the number of transponders and large (and expensive) switching machinery;
.
improved network reliability with a few in number of electronic systems;
.
significant reduction in power consumption (from switching fabrics to cooling requirements).
In an optically transparent network, the routing of the sign al is based on the wavelength and/or on the physical port of the signal. Framing takes place at the ingress to the optically transparent domain and it adds overhead information that makes it possible to detect errors – possibly occurring during transmission – at the egress node. Each standardized format has a specific frame, and several different frames are possible at the ingress to an optically transparent domain; for example, Ethernet frames, synchronous transport module frames, and G.709 [11] frames.
Ethernet as an Alternat ive Transport Platform
Ethernet is a frame-based technology that was defined in 1970s. It was originally designed for computer communications and for broadcasting, and since then it has been widely adopted. This was made possible thanks to two main competitive advantages. First, a successfully implemented switching capability besides the original broadcasting LAN technology. Second, because all generations of Ethernet share the same frame formats, making it feasible to support interfaces from 10 Mbit/s over copper to 100 Gbit/s over fiber (the latter still under standardi­zation), thus ensuring seamless upgra deability.
Nowadays, Ethernet represents the most successful and widely installed LAN technology and is progressively becoming the preferred switching technology in metropolitan area networks (MANs). In the latter scenario, it is used as a pure layer 2 transport mechanism, for offering VPN services, or as broadband technology for delivering new services to residential and business users. Today, Ethernet traffic is rapidly growing and apparently it has surpassed SDH traffic. As mentioned in the previous section, bit transparent mapping is essential for a cost-effective transportation of data and it is a feature provided by Ethernet thanks to its framing format. As Ethernet is becoming the dominant technology for service provider networks and as 40/100 GbE interfaces will be standardized in the years to come, it is essential to keep Ethernet transport attractive and simple. Currently, under the Ethernet umbrella, three network layers are considered:
.
Network layer, based on Metro Ethernet Forum (MEF) documents. The network services include E-Line, defining point-to-point connections, and E-LAN, defining multipoint­to-multipoint connections and rooted multipoint connections.
26 The Emerging Core and Metropolitan Networks
.
Layer 2, which is also called the MAC layer. This provides network architectures, frame format, addressing mechanisms, and link security (based on IEEE 802.1 and IEEE 802.3).
.
Physical layer, which includes the transmission medium (e.g., coaxial cable, optical fiber), the modulation format, and network basic topologies (based on IEEE 802.3).
The numerous networks where Ethernet is installed confirm that it is a valuable frame-based technology, capable of assuring an inexpensive physical stratum, providing high bit-rates, and allowing network architectures to offer emerging network services for distributing both point­to-point and p2mp variable bit-rate traffic efficiently.
There are several initiatives at standardization bodies that aim at a revision of Ethernet to make it valuable and deployable in transport networks. The work carried out at IEEE, IETF,and ITU-T is improving Ethernet with faster and more efficient resilience mechanisms and valuable OAM tools for fault localization and measurement of quality parameters to verify customers’ service level agreements (SLAs).
Given this success in the access area and MAN and the simplicity and transparency it offers, Ethernet is stepping forward to the core network segment under the definition of carrier Ethernet which is investigated by the MEF as “a ubiquitous, standardized, carrier-class Service and Network.” Carrier Ethernet improves standard Ethernet technology facing scalability issues (assuring a granular bandwidth increment from 1 Mbit/s to 10 Gbit/s). It also assures hard QoS mechanisms (allowing the transport on the same lambda of different traffic categories) and reliability (the network is able to detect and recover from failures with minimum impact on users). Carrier Ethernet aims to achieve the same level of quality, robustness, and OAM functions typical of circuit technologies (think of SDH or OTN) while retaining the Ethernet advantage in offering a cost-effective statistical aggregation.
1.3.4.2 Control Plane Requirements
The control plane is studied in detail in Chapter . Here, some important issues are highlighted.
Provisioning of End-to-end Connections over the Entire Network
The main function of the control plan e is to set up, tear down, and maintain an end-to-end connection, on a hop-by-hop basis, between any two end-points. The applications supported from the transport network have specific QoS require ments (Section 1.2.6), which the control plane must uphold.
Unified Control Plane
In the quest to upgrade or build new integrated network infrastructures, a paradigm shift has been witnessed in network design principles. The focus has shifted from a layered-network model involving the management of network elements individually at each layer, to one of an integrated infrastructure able to provide a seamless management of packets, circuits, and light paths. The reasons for this industry trend towards a unified set of mechanisms (the unified control plane), enabling service providers to manage separate network elements in a uniform way, can be traced to the historical evolution of transport and packet networks. The IP became the uncontested platform for supporting all types of application and the associated, IP-based, GMPLS provides a single, unified control plane for multiple switching layers [12].
Future Networks Challenges 27
Horizontal Integration (Unified Inter-domain Control)
This issue refers to the way of multidomain interconnection at control plane level. Horizontal integration refers to the situation where, in the data plane, there is at least one common switching facility between the domains, whilst the control plane topology extends over several domains. For instance, the control plane interconnection between lambda-switching-capable areas defines a horizontal integration.
Control Plane and Manage ment Plane Robustness
In the emerging optical network architectures, the interplay between the control plane and management plane is essential to ensure fast network reconfiguration, while maintaining the existing features of SDH/SONET like robustness against failures, which is essential for the preservation of traffic continuity.
Network Autodiscovery and Control Plane Resilience
Automated network discovery refers to the ability of the network to discover autonomously the entrance of new equipment or any changes to the status of existing equipment. This task is assigned to the control plane. Additional functions of the control plane are the automated assessment of link and network load and path computation process needed to substantially reduce the service provision time and the changes invoked in the network infrastructure to support these services. Moreover, the automation is essential to reallocate resources: as customers cancel, disconnect, or change orders, the network resources can be readily made available to other customers.
The term control plane resilience refers to the ability of the control plane to discover the existing cross-connect topology and port mapping after recovering from a failure of itself. For example, when only control plane failures occur within one network element, the optical cross-connects will still be in place, carrying data traffic. After recovery of the control plane, the network element should automatically assess the data plane (i.e., optical cross­connects), and reconfigure its control plane so that it can synchronize with other control plane entities.
Appropriate Network Visibility among Different Administrative Domains Belonging to Different Operators
Administrative domains may have multiple points of interconnections. All relevant interface functions, such as routing, information exchanges about reachable nodes, and interconnection topology discovery, must be recognized at the interfaces between those domains. According to ASON policy, the control plane should provide the reference points to establish appropriate visibility among different administrative domains.
Fast Provisioning
As part of the reliable optical network design, fast provisioning of optical network connections contributes to efficient service delivery and OPEX reduction, and helps reaching new customers with broadband services.
Automatic Provisioning
To achieve greater efficiencies, optical service providers must streamline their operations by reducing the number of people required to deliver these services, and reducing the time required to activate and to troubleshoot network problems. To accomplish these objectives,
28 The Emerging Core and Metropolitan Networks
providers are focusing on automated provisioning through a distributed control plane, which is designed to enable multi-vendor and multilayer provisioning in an automated way. Therefore, requests for services in the data network that may require connectivity or reconfiguration at the optical layer can happen in a more automated fashion. In addition, instead of provisioning on a site-by-site basis, the control plane creates a homogeneous network where provisioning is performed network-wide.
Towards Bandwidth On-demand Services
Providers can also set up services where the network dynamically and automatically increases/ decreases bandwidth as traffic volumes/patterns change. If the demand for bandwidth increases unexpectedly, then additional bandwidth can be dynamically provisioned for that connection. This includes overflow bandwidth or bandwidth over the stated contract amount. Triggering parameters for the change may be utilization thresholds, time of day, day of month, per­application volumes, and so on.
Bandwidth on demand (BoD) provides connectivity between two access points in a non­preplanned, fast, and automatic way using signaling. This also means dynamic reconfiguring of the data-carrying capacity within the network; restoration is also consider ed here to be a bandwidth on-demand service.
Flexibility: Reconfigurable Transport/Optical Layer
A network operator may have many reasons for wanting to reconfigure the network, primarily motivated by who is paying for what. Flexibility of the transport layers means a fair allocation of bandwidth between competing routes dealing with bursts of activity over many timescales. Reconfigurability increases network flexibility and responsiveness to dynamic traffic demands/ changes.
1.3.4.3 Interoperability and Interworking Requirements
Multidomain Interoperability
In many of today’s complex networks, it is impossible to engineer end-to-end efficiencies in a multidomain environment, provision services quickly, or provide services based on real-time traffic patterns without the ability to manage the interactions between the IP-layer function­ality of packet networks and that of the optical layer. Accordi ng to proponents of ASON/ GMPLS, an optical control plane is the most advanced and far-reaching means to control these interactions.
Another important issue is that of translating resilience classes from one domain to another. The ASON reference points UNI and I-NNI/E-NNI are abstracted functional interfaces that can resolve that topic by partitioning the transport network into sub-networks and defining accurately the exchanges of control information between these partitions. As recommended in Ref. [13], the UNI is positioned at the edge of the transport network as a signaling interface used by the customer edge nodes to request end-to-end connection services between client networks, with the explicit level of availability. Routing and signaling messages exchanged at the I-NNI concern only the establishment of connections within a network domain or across the subnetwork. The E-NNI is placed between network domains or sub-networks to carry the control message exchanges between these regions of different administration.
Future Networks Challenges 29
Multi-vendor Interoperability
The multi-vendor interoperability of metro and core solutions maximizes carrier performance and ensures the interoperability of legacy with emerging network architectures. One of the most important objectives of the development of a standardized ASON/GMPLS control plane is to contribute to interoperability, which validates the speed and ease of provisioning enabled by ASON/GMPLS in a live, multi-vendor network.
Seamless Boundary in between Networks
Given the vast amount of legacy SONET/SDH equipment, there is a clear need for an efficient interworking between traditional circuit-oriented networks and IP networks based on the packet-switching paradigm. For example, efficient control plane interworking between IP/MPLS and SONET/SDH GMPLS layers is indispensable and requires the specification of their coordination.
1.3.4.4 Management Plane Requirements
Easy-to-use Network
Emerging standards and technologies for optical networks allow for a significantly simplified architecture, easy and quick provision of services, more effective management, better interoperability and integration, and overall lower cost. In addition, it will be possible to provision services on these future networks such that global applications will be much more location independent.
Transparent for Applications: Hide Network Technology to Users
There are multiple separate service, technology, and technical considerations for networks depending on location, at the metro edge, metro core, aggregation points, long haul, and ultra­long haul. Next-generation optical networking has the potential to reduce significantly or eliminate all of these barriers, especia lly with regard to application and end users.
To some degree, one of the key goals in this development is to create network services with a high degree of transparency; that is, allow networ k technical elements to become “invisible” while providing precise levels of required resources to applications and services. To allow an optimal use of the optical network infrastructure interconnecting different types of application, network service management functions are required to establish automatically connection services with adequate amount of allocated network resources. The network service man age­ment layer can rely on the routing and signaling control functions.
Monitoring of End-to-end Quality of Service and Quality of Resilience
The requirement of integrated monitoring of the (optical) performance of connections, QoS, and fault management speeds up system installation and wavelength turn-up and simplifies ongoing maintenance. Furthermore, the management plane should be able to monitor end­to-end quality of resilience. That means the end-to-end type of transport plane resilience parameters (such as recovery time, unavailability, etc.) should be monitored and adhered according to the SLAs).
Connectivity and Network Performance Supervision
As networks run faster and become more complex, infrastructure, links, and devices must operate to precise levels in a tighter performance. As a result, a huge number of network
30 The Emerging Core and Metropolitan Networks
problems stem from simple wiring and connection issues. Connectivity and performance supervision is at the heart of an efficient network management.
Network Monitoring
A monitoring system is dedicated to the supervision of the physical and optical layers of a network. Optical-layer monitoring should provide valuable, accurate information about the deterioration or drift with slow and small signal variations, helping to detect problems before they may become so serious to affect the QoS. It helps maintain the system from a lower layer’s perspective.
Policy-based Management (Network and Local-basis)
Today’s optical network architectures lack the proper control mechanisms that would interact with the management layer to provide fast reconfiguration. The problem of accurate intra­domain provisioning in an automated manner allows satisfying the contracts with customers while optimizing the use of network resources. It is required that a policy-based management system dynamically guides the behavior of such an automated provisioning through the control plane in order to be able to meet high-level business objectives. Therefore, the emerging policy­based management paradigm is the adequate means to achieve this requirement.
End-to-end Traffic Management (Connection Admission Control, Bandwidth Management, Policing)
Traffic managem ent features are designed to minimize congestion while maximizing the efficiency of traffic. Applications have precise service requirements on throughput, maximum delay, variance of delays, loss probability and so on. The network has to guarantee the required QoS. For instance, the prima ry function of the connection admission control is to accept a new connection request only if its stated QoS can be maintained without influencing the QoS of already-accepted connections. Traffic man agement features are key elements in efficient networking.
Multi-vendor Interoperability
In the near future, network element management interfaces and OSS interfaces will be pre­integrated by control plane vendors. Indeed, independent control planes increase the perfor­mance of network elements and OSS, and reduce carriers’ reliance on any single network element or OSS application. This eliminates the task of integrating new network elements into a mass of OSS applications.
Connection services (respectively, connectivity services) are described from the network infrastructure operator (respectively, the service customer) point of view, which is comple­mentary for the connections implemented through the control functions at customer edge (CE) nodes. Provider VPN services offer secure and dedicated data communications over telecom networks, through the use of standard tunneling, encryption, and authentication functions. To reconfigure automatically the provisioning of VPNs, automated OSS functions are required to enhance existing network infrastructures for supporting networked applications sharing the optical infrastructures.
Network service functions can automatically trigger addition, deletion, move, and/or change of access among user sites. The description of each connection service includes the UNI corresponding to the reference point between the provider edge (PE) node and CE node. At a
New Transport Networks Architectures 31
given UNI, more than one connection can be provisioned from the network management systems or automa tically signaled from the control plane functions according to multiplexing capabilities. GMPLS controllers enable signaling of the connection establishment on demand, by communicating connectivity service end-points to the PE node. This operation can be assigned to an embedded controller to exchange the protocol messages in the form of RSVP-TE messages.
Support Fixed–Mobile Convergence
Fixed-mobile convergence means alliance of wired and wireless services and it is referring to single solutions for session control, security, QoS, charging and service provisioning for both fixed and mobile users. Fixed-mobile convergence is clearly on the roadmap of operators that want to create additional revenue streams from new value-added services.

1.4 New Transport Networks Architectures

Today’s telecommunication networks have evolved substantially since the days of plain­telephony services. Nowadays, a wide variety of technologies are deployed, withstanding a substantial number of failures, supporting a broad range of applications based on diversified edge-user devices; they span an enormous gamut of bit-rates, and they are scaling to a large number of nodes.
In parallel, new services and networking modes (e.g., peer-to-peer) are emerging and proliferating very rapidly, modifying the temporal and spatial traffic profile in rather unpre­dictable ways. As has been discussed in the previous sections, it is widely recognized that the existing mind-set for the transport network architecture largely fails to accommodate the new requirements. However, the bottleneck is not only on the technology front. Architectural evolution presupposes a consensus between the many providers which, quite often, is hard to reach. This situation is exacerbating interoperability issues that, potentially, negate any competitive advantage stemming from architectural innovation. Market protectionism could stall technological advances.
Nevertheless, a major rethinking on network architectures is mandatory in the quest for a cost-effective, secure, and reliable telecommunications network. The research today is pivoted around notions on how network dynamicity can be significantly enhanced, how the cost of ownership can be reduced, and how the industrial cost of network services can be decreased. The scope of this section is to present plausible scenarios for the evolution of the core and the metropolitan transport networks, taking into account the data plane as well as the control/ management planes. It is organized so as to provide snapshots of the current situation in both segments and for three discrete time plans:
.
short term (2010)
.
medium term (2012)
.
long term (2020).
Figure 1.4 depicts the existing network architecture for metro/regional and core/backbone segments, which will be the starting point in the network evolution scenario. Today, the functionality requirements are dissimilar in the two network segments, leading to the adoption of different solutions, as was shown in Figure 1.3:
32 The Emerging Core and Metropolitan Networks
Figure 1.4 Existing metro and core network architecture
.
In an MAN, the client traffic is transported from a “zoo” of protocols (IP, Ethernet, ATM, SDH/SONET, ESCON, Fiber channel, etc., to mention the most import ant instances only) whilst it is characterized from a low level of aggregation and grooming; this a problem exacerbated by the coexistence of unicast (video on demand, high-speed Internet and voice services) and multicast traffic (i.e., mainly IPTV). This environment postulates a highly dynamic networking, making packet-oriented solu tions a necessity.
.
In the core network, on the other hand, the efficient aggregation and grooming out of the MAN indicates a smoothed-out, slowly varying traffic profile, so that a circuit-switched solution is a good candidate for a cheaper switching per bit. These developments, for the core, are beefed up by the past and current developments of dense WDM (DWDM) and OTN technologies, to enhance the “bit-rate times distance” product significantly, by two to three orders of magnitude, compared to what was feasible in late 1980s, by shifting to transmission the balance for a lower cost per bit transportation.
Regarding the scenarios presented in the rest of this section, it is pointed out that their
common denominator is progress in the following enablers:
.
Packet technologies (in particular IP/MPLS and Ethernet) for a more efficient use of bandwidth due to the subsequent statistical multiplexing gains; that is, advanced aggregation and grooming.
.
Control plane (currently dominated by ASON and GMPLS, which are further discussed in Chapter ) to decrease the cost of provisioning dramatically and the possibility to have on­the-fly resilience mechanisms.
.
Optical transparency, which, as explained in Section 1.3.4.1, aims at minimizing the level of bit-by-bit processing, simplifying client signal encapsulation, leading to transparent
New Transport Networks Architectures 33
bit-mapping in the transport frame and providing optical bypassing for the transit traffic. These are key functions for a reduction in CAPEX and OPEX.
As it emerges from simple inspection of the existing network architecture paradigm, efficiency and robustness today are achieved from the interplay between two, rather mutually exclusive, technologies: packets (mainly IP/MPLS and Ethernet) and circuits (SDH/SONET, OTN and WDM) do coexist in transport networks with a low level of interoperability and significant functionality duplication. Apparently, for an overall optimization, it is fundamental to increase the synergy between the layers and reduce the unnecessary functionalityduplication. Thus, the emerging technologies (PTTs, see Section 1.2.9), which are in the standardization process, aim at combining the best features from both circuit and packet worlds. Therefore, features like OAM and control plane and resilience mechanisms are inherited from the circuit transportnetwork, whileframe format, statistical aggregation, and QoS supportare similar to the corresponding features of packet technologies. Within the standardization bodies, two main technologies are currently under discussion: PBB-TE (IEEE802.1Qay [14] based on Ethernet) and MPLS-TP (developed in the ITU-T and IETF, starting from MPLS).

1.4.1 Metropolitan Area Network

The introduction of triple-play applications (voice, video, and high-speed Internet) has a strong impact on Metropolitan Area Network (MAN) traffic. The advances include improve­ments in residential access networks (whose traffic aggregates upwards to the MAN), multimedia distribution (which is using MANs in an intensive way) from points of presence point of presence (PoPs) to the home, and finally VPNs that are used for business and residential applications. The necessity to provide multicast services (i.e., to carry IPTV) and to add/release users to multicast groups very quickly is a strong driver towards packet solutions (IP, Ethernet, ATM, ESCON, Fiber channel, etc.).
In the MAN segment, the main advantages of cir cuits (low cost per switched bit per second, strong OAM, and efficient resilience mechanisms) are not essential: the bandwidth at stake is not really huge and the distances of the cables interconnecting the nodes are not very long, so that the probability of failure due to fiber cut is not that high to mandate a circuit-switched level of resilience.
However, using IP over Ethernet or pure Ethernet over WDM systems (architectural examples are available in Ref. [15]) presents some problems in terms of resilience, bandwidth guarantee, and OAM, because packet technologies currently do not have efficient mechanisms to cope with such functions. For these reasons, technologies with the ambition to couple circuit­like OAM and resilience mechanisms with packet-like flexibility, dynamicity, and granularity might represent the right candidates for next-generation networks.
Both the emerging PTTs (PBB-TE and MPLS-TP) currently have a lack of multicast traffic that, at the date of writing (July 2008), is still under study.
Figure 1.5 shows the possible evolution of the architecture for networks in the MAN or regional segment. The following sections describe in depth the concepts illustrated in this picture. At the moment, the most plausible scenario is a migrations towards technologies that, from one side, assure packet granularity and, from the other side, have “circuit-like” OAM
34 The Emerging Core and Metropolitan Networks
current scenario
short term scenario
mid-long term scenario
DeviceServices
data
voice
IPTV
ETH
e2e
PTT
services
G707 G709
leased line
lambda
IEEE802.1AD
SDH
IEEE802.1q
ETH int
IP/MPLS
ETH int
IEEE802 1ah
xWDM
dark fiber
PBB-TE
ETH ETH
T-MPLS
OTN
IP/MPLS router
Ethemet switch
PTT device
circuit device
optical system
Figure 1.5 Evolution scenario for metropolitan/regional network architecture
and resilience mechanisms (capabl e of switching in times shorter than 50 ms after failure, a performance requested by the majority of applications).
1.4.1.1 Short Term
In the short term, there will probably be a progressive migration of fixed and mobile services to IP. This migration will speed up the increasing interest towards the Ethernet technology [16]. So, the roll out of native Ethernet platforms in the metro space is likely to start in the next time frame. Metro network solutions in the short term are expected to be mainly based on the Ethernet technology in star or ring topologies.
Nevertheless, in this phase, both packet (IP and Ethernet) and circuit (for the most part SDH) will coexist. In most cases, different kinds of traffic (with different quality requirements) will be carried on the appropriate platform (e.g., voice or valuable traffic) on SDH and the remainder on packet platforms.
No unified control plane is available for the transport layers. Further, the control plane is restricted to a single network domain and in most cases to a single layer within the network.
1.4.1.2 Medium Term
In the metro network, Ethernet will probably be the dominant technology in the medium-term scenario. The utilization of Ethernet is mainly driven by Ethernet conformal clients; however, non-Ethernet payloads, such as TDM, ATM, and IP/MPLS, will still exist for a long time and have to be adapted into an Ethernet MAC frame.
New Transport Networks Architectures 35
Therefore, any incoming non-Ethernet payload behaves as an Ethernet payload from a network perspective; the reverse operation is performed at the outgoing interface of the egress network node.
Since Ethernet networks are growing in dimension, moving from simple LAN or basic switched networks (e.g., in a campus behavior) towards a situatio n where an entire metropoli­tan (or regional) area is interconnected by an Ethernet platform, hundreds of thousands (or even millions) of MAC addresses would have to be learned by the switches belonging to the metro networks. To prevent this severe scalability problem, IEEE 802.1ah (PBB or MACinMAC) might be adopted. This evolution of the classical Ethernet allows layering of the Ethernet network into customer and provider domains with complete isolation amo ng their MAC addresses.
Leaving the SDH technology, the main problems that still remain are related to the lack of efficient resilience mechanisms, present in SDH, but not mature with IP or Ethernet. In fact, traditional Ethernet (802.1q, 802.1ad, and 802.1ah) bases resilience on the “spanning tree” mechanism and its evolutions (for instance, VLAN spanning tree), which are inefficient for carrying traffic that has strong requirements in terms of unavailability.
The same argument might be argued if resilience is demanded at the IP level. In this case, routing protocols (e.g., open short path first, intermediate system to intermediate system) after a failure rearrange routing tables on surviving resources; also, this process assures stability after some seconds, a time that is often too long for voice or some video applications.
Innovative solutions to this problem might be represented by resilient packet ring (RPR) or optical circuit switching (OCS) rings. OCS rings are specially adapted to metro–core scenarios, as well as to metro access characterized by high-capacity flows between nodes (e.g., business applications and video distribution services), while RPR and dual bus optical ring network solutions fit better in scenarios with higher granularity and lower capacity requirements per access node.
At control plane level, the most important aspects that are expected for the medium-term scenario are the implementation of interfaces to make possible the exchange of information (routing and signaling) between control planes even between different domains and finally the vertical integration of the control planes of layer 1 and layer 2 technologies.
1.4.1.3 Long Term
The metro segment is composed of metro PoPs (GMPLS-capable LSR), some of which link the metropolitan network to the IP/optics core backbone (core PoP).
In this phase, the solutions based on Ethernet technology (that is, on 802.1ah (MACinMAC) and the IP/MPLS routing) will probably be replaced by innovative PTTs.
These technologies (MPLS-TP and PBB-TE) are connection-oriented transport technolo­gies based on packet frames, enabling carrier-class OAM and fast protection. IP/MPLS should remain at the edge of the network (e.g., in the access), while the metro-core will be dominated by packet transport.
The reasons for a migration towards packet transport are a very efficient use of the bandwidth (due to the packet behavior of the connection) joint to OAM and resilience mechanisms comparable in efficiency to what is standardized in circuit-based networks.
36 The Emerging Core and Metropolitan Networks
In addition, PTTs keep the door open to the introduction of a fully integrated ASON/GMPLS network solution, which seems to be one of the most interesting approaches to meet network emerging requirements, not only overcoming the four fundamental network problems (band­width, latency, packet loss, and jitter) for providing real-time multimedia applications over networks, but also enabling flexible and fast provisioning of connections, automatic discovery, multilayer TE, and multilayer resilience, all based on an overall view of the network status.

1.4.2 Core Network

Consistent with the metro/regional description, Figure 1.6 depicts a possible migration trend for the architecture of the backbone network.
The current network architecture, depicted in the left side of the figure, is influenced by the long-distance traffic relationships that are currently covered by two networks: an IP/MPLS network (based on routers) and a transmission network based on optical digital cross-connects (in SDH technology).
In a first phase, the evolution is represented by the migration from legacy SDH to an OTN (based on ITU-T G.709 and its evolution); in parallel, Ethernet interfaces as routers’ ports will substitute PoS ports. The next phase will be characterized by the adoption of PTTs for providing connectivity and substituting pass-through routers.
The following sections describe more deeply the concepts summarized in the figure.
1.4.2.1 Short Term
In the backbone segment, for an incumbent operator, the dominance of SDH carried on DWDM systems will be confirmed in the near future.
mid-long term scenario
PBB-TE
ETH ETH
T-MPLS
OTN
DeviceServices
IP/MPLS router
Ethemet switch
PTT device
circuit device
optical system
data
voice
IPTV
ETH
e2e
PTT
services
G707 G709
leased line
lambda
fiber
current scenario
Pos
SDH OT N
short term scenario
IP/MPLS
ETH int
ETH
xWDM also (flexible)
dark fiber
Figure 1.6 Evolution scenario for core/backbone network architecture
New Transport Networks Architectures 37
No unified control plane is available for the transport layers, yet. Furthermore, the control plane is restricted to a single network domain and in most cases to a single layer within the network.
Single-layer TE and resilience mechanisms will be still in use for quite a while. The full standardization of the ASON/GMPLS control plan e is not yet complete. However, some vendors already provide optical switches equipped with a standard or proprietary implemen­tation of the GMPLS control plane and make feasible control-plane-driven networking using the overlay network model. This suggests that the automatic switched transport network architecture using GMPLS protocols is being implemented and deployed together with different releases of the UNI and NNI.
1.4.2.2 Medium Term
In the core network, standard SDH (and to some extent OTH) is being introduced. The support for Ethernet-based services is increased. The use of the native Ethernet physical layer as layer 1 transport in the core will most likely depend on the availability of OAM functionality.
With the introduction of an intelligent SDH and OTH network layer, service providers can achieve significant cost savings in their backbones. IP over static L1 networks (e.g., IP over peer-to-peer links) should cope with a high amount of transit traffic in the core routers. As traffic increases, there comes a point where the savings in IP layer expenses realized by end-to-end grooming – where the bypass traffic is sent on the L1 layer without going back to the IP layer – compensate the extra expenses of introducing the intelligent layer 1 (SDH/OTH) switches needed.
The vertically integrated control plane refers to the underlying concepts that are called MLN/MRN at the IETF [17] and next-generation networks (NGNs) at the ITU-T. On the other hand, horizontal integration refers to the ability of control planes to provide support for service creation across multiple network domains.
The control plane will be aware of the physical layer constraints, which are important to consider, for instance, during routing in transparent/hybrid networks. Indeed, topology and resource information at wavelength level, as well as simplified signal quality/degradation information on links/wavelengths and nodes, is needed to allow the routing and wavelength assignment algorithm to place feasible paths efficiently into the network. In opaque networks, routing is based only on the overall path length constraint.
TE over domain borders between two or more domains will be a crucial topic.
1.4.2.3 Long Term
PTTs will also probably dominate the long-term scenario for the backbone segment, even if at later times than their adoption in the metro/regional network segment.
A probable architecture will consider some edge routers aggregating traffic and a network consisting of packet transport switches that will connect the edge routers.
The task of this packet transport network is to connect routers with dynamic connectivity (thanks to a control plane, probably of GMPLS type) and to assure multilayer resilience.
As shown in Figure 1.6, only some relationships might be confined at packet transport level, not the whole traffic.
38 The Emerging Core and Metropolitan Networks
For CAPEX reasons, the deployment of packet transport devices is more similar to that of an L2 switch than to L3 routers; as a consequence, the cost of switching (normalized per bits per second) is expected to be much lower than the current cost of switching in IP routers. For this reason, large bandwidth relationships (say, larger than 2 Gbit/s) should be carried more conveniently in connection-oriented mode. In a first phase, these circuit networks should be represented by the G.709 technology that, as said before, will probably dominate the medium­term scenario. Successively, the architecture of the backbone network will see the coexistence of G.709 and packet transport networks.
In general, integrated equipment can be assumed for the core network in the long-term scenario; this means that, within the core network, each equipment (LSR) would integrate multiple-type switching capabilities such as packet-switching capability (PSC) and TDM (utilizing SDH/SONET or OTH fabrics); or, in an even more evolutionary scenario, a solution where PSC and lambda switching capability (LSC) or where LSC and fiber switching capability coexist may be available.
The introduction of an integrated ASON/GMPLS network control plane solution might represent one of the most interesting approaches to meet network emerging requirements, both to overcome the four fundamental network problems (bandwidth, latency, packet loss, and jitter), to provide real-time multimedia applications over networks, and to enable flexible and fast provisioning of connections, automatic discovery, multilayer TE, and multilayer resil­ience, all based on an overall view of the network status.
As mentioned before, the control plane model considered for the long-term scenario is a fully integrated (horizontal and vertical) GMPLS paradigm, allowing a peer-to-peer interconnection mode between network operators, as well as network domains. Specifically, full integration means that one control plane instanc e performs the control for all the switching capabilities present in the network.

1.4.3 Metro and Core Network (Ultra-long-term Scenario)

Optical burst and/or packet switching might represent important technologies to face the flexibility demand of bandwidth in future networks. It is still unclear when some implementa­tions of these technologies will be available, even if they could be developed (both at standard level and commercially) before 2012 to 2015, which seems very unlikely.
However, a dramatic increase of the traffic amount and the necessity of end-to-end QoS, in particular for packet-based network services, may open the door to PTTs as a new layer 1/layer 2 network solution that can overcome existing shortcomings.
The further evolution of PTTs may be represented by innovative solutions based on burst/ packet switching, which would offer the following functionalities:
.
Burst/packet switching will have the required dynamicity and flexibility already in layer 2, since an appropriate size of bursts/packets eliminates the grooming gap by offering a fine granularity, with less processing effort compared with short IP or Ethernet packets/frames.
.
Reliability, monitoring, and QoS functionalities will be provided at layer 2, offering a solid carrier-class network service supporting higher layer network services at low cost.
.
Hybrid circuit/burst/packet switching capabilities will be fully integrated into the GMPLS control plane philosophy (full vertical integration).
Transport Networks Economics 39
Specifically for core networks, layer 2 network technologies could consist of a hybrid circuit/ burst/packet solution based on large containers carrying TDM, packet transport, and data traffic.
In metro networks – currently being dominated by Ethernet transport – the future architec­ture may be represented by the adoption of carrier-grade Ethernet protocols endowed with extensions on control, monitoring, and QoS. This implementation should also fit into the vertical integration strategy based on a GMPLS control plane and the horizontal integration with domain interwor king providing end-to-end QoS.
From the control plane point of view in the ultra-long-term scenario, the ASON architecture based on GMPLS protocols is the most promising solution to integrate the TDM-based optical layer transport technologies (i.e., G.709) and the dominating packet-based data traffic using IP and IP over Ethernet protocols.

1.5 Transport Networks Economics

There is no unique infrastructure to support the required network services for the expected traffic; furthermore, not all plausible migration scenarios are cost effective for any of the network operators or within different market regulations. To analyze these differences, network operators use models that help them to evaluate how much a given network service imple­mentation is going to cost, both in terms of CAPEX (the initial infrastructures roll out) and management and operation of the service (OPEX).

1.5.1 Capital Expenditure Models

CAPEX creates future benefits. CAPEX is incurred when a company spends money either to buy fixed assets or to add to the value of an existing fixed asset, with a useful life that extends beyond the taxable period (usually one financial year). In the case of telecommunications, operators buy network equipment to transmit, switch, and control/manage their infrastructures; this is part of CAPEX, but it also includes some more items:
.
rights of way and civil works needed to set the equipment and deploy the lines;
.
software systems (or licens es);
.
buildings and furniture to house personnel and equipment;
.
financial costs, including amortization and interest over loans (used to buy any of the former items).
A reasonable comparison among different solutions need not take all those items into account. However, two limiting approaches should be considered in evaluating investment and amortization:
.
a “green field” situation, where a network operator starts to build their new network (or a part of it, as may happen for the access segment);
.
upgrading the deployed network with new equipment or to add new functionality to the existing infrastructure.
40 The Emerging Core and Metropolitan Networks
traffic related to service “X”
dimensioning
elements’ price list
X
investment
useful life
amortization
Figure 1.7 Process to evaluate the investment and amortization of a telecommunication network
As for the equipment deployment, the starting point naturally consists of dimensioning the
requirements as a resu lt of traffic estimation (Figure 1.7).
Amortization is the process of allocating one lump sum (CAPEX) to different time periods. Amortization can be calculated by different methods, but its concept describes the total expenses due to an asset over the time of its economic usefulness. As a coarse rule, the following list can be used to estimate the useful lifetime:
.
network infrastructures (excavation, buildings, ...) 30 years
.
optical fibers and copper wires 10 years
.
network equipment 5 years
.
software 3 years.
Different systems and technologies give rise to different components and, thus, different CAPEX analyses.
Figure 1.8 represents a general block model for most switching systems, including interface components, software, power supply, and common hardware elements. The cost mode l obviously arises from adding all individual prices for each component.
Sometimes, network operators prefer to establish a certain redundancy for some (common hardware) components so as to ensure service reliability.
Software
Software (libraries)
Power
Power
supply
supply
Hardware: Switches,
Hardware: Switchs,
Rx, Tx, etc.
Rx, Tx, etc.
(libraries)
Interfacing cards for clients
Figure 1.8 Simplified block model for a switching system
Transport Networks Economics 41
XC
XC
XC
XC
Conmutador
Conmutador
DWDM
DWDM
DWDM
DWDM
nλ nλ
Electrical
Conmutador
Conmutador
Eléctrico
Eléctrico
Matrix
Eléctrico
Eléctrico
OXC
OXC
OXC
OXC
Conmutador
Conmutador
Optical
Conmutador
Conmutador
Matrix
Optico
Optico
Optico
Optico
DWDM
DWDM
DWDM
DWDM
λ1... λn
Figure 1.9 Hybrid SDH/lambda switching node
On the other hand, manufacturers also may introduce hybrid system s taking into account that common elements may be used by different technological solutions to accomplish a given function (switching packets, for instance).
Figure 1.9 shows the case for a transparent and opaque hybrid solution for a node where some lambdas are switched at the optical level, requiring no electrical regeneration (most lambdas are bypassed and transponders are only used to insert/extract lambda on/off the transmission line), whereas opaque switching (for low-granularity lines and to deal with the ingress/egress lambda of the transparent module) require electro-optical conversion. In fact, modular equipment is offered in a “pay as you grow” model to keep upgrading systems according to traffic growth for different (and interoperable) capacities like opaque/transparent hybrid nodes or L2/L3 multilevel solutions.
Aside from the amortization time schedule, network operators plan network infrastructures several years in advance. That is the reason why the introduction of new technologies must consider not only current prices, but also somewhat different ones for the future, taking into account certain conditions.
.
Vendors offer discounts to network operators; discounts are very frequent because this is a way for vendors position their products as the de facto standard by massive installations.
.
Equipment prices get lower after standardization agreements.
.
Learning curves finally represent a significant price reduction as technologies mature. The empirical rule that states “as the total production of a given equipment doubles, the unit cost decreases in a constant percentage” can be combined with an initial estimate of market penetration to allow for predicting such a price reduction due to the maturity process.
These kinds of techno-economical prediction are usually carried out in combination with more general strategic considerations: whether network operators expect to expand their business outside present geographical limits or not, whether they will be able to reuse equipment for other purposes or places (from core networks to metropolitan ones, for example), or simply if it is possible to buy equipment from other companies, and so on.
On the other hand, cost models are always performed with a sensitivity analysis that highlights which elements are the most important to define a trend in the price evolution of
42 The Emerging Core and Metropolitan Networks
a system and, as a consequence, to provide a tool for benchmarking it. This task must be done in combination with a general vision of the network architecture, since it is not straightforward to compare different topologies and multilevel interoperation.

1.5.2 Operational Expenditure Models

OPEX is not directly part of the infrastructure and, thus, is not subject to depreciation; it represents the cost of keeping the network infrastructure operational and includes costs for technical and commercial operations, administration, and so on. Personnel wages form an important part of the OPEX, in addition to rent infrastructure, its maintenance, interconnection (with other network operators’ facilities) cost s, power consumption, and so on.
Only considering specific OPEX derived from network services provision, the following list
can be used as a guide to analyze its components:
.
Costs to maintain the network in a failure-free situation. This includes current exploitation expenditures, like paying rents for infrastructures, power for cooling and systems operations, and so on.
.
Operational costs to keep track of alarms and to prevent failures. This involves the main control activities to ensure QoS, namely surveying systems and their performance versus traffic behavior.
.
Costs derived from failures, including not only their repair, but also economic penalties (in case an SLA states them for service failures).
.
Costs for authentication, authorization, and accounting (AAA) and general management of the network.
.
Planning, optimization, and continuous network upgrading, including software updating and QoS improvement.
.
Commercial activities to enhance network usage, including new service offers.
Several approaches to analyzing OPEX can be used. To compare technologies and network architectures or different services implementations, a differential approach may be sufficient instead of considering all OPEX parts. However, if a business case-study requires knowledge of all costs and revenues, then a total OPEX calculation must be performed.
On the other hand, OPEX calculation is different for green-field and migration situations; for instance, bulk migration of customers or removal of old equipment will not be taken into account for a green-field scenario. Furthermore, bottom-up or top-down approaches can be used to calculate OPEX: the top-down method fits well to get a rough estimation of costs, as a starting point for a finer analysis of relative costs; the bottom-up approach is based on a detailed knowledge of operational processes.
Various approaches can be combined when dealing with OPEX calculations. In addition, OPEX and CAPEX may be balanced for accounting exploitation costs (e.g., buying or renting infrastructures) and some OPEX concepts can also be included in different items: salaries, for instance, can be considered as an independent subject or inside maintenance, commercial activities, and so on. A deep evaluation of OPEX is really important; in fact, it is possible that some technologies may offer high performances and perhaps at rel ative low CAPEX, but their
Transport Networks Economics 43
maintenance
reparation
charging billing/ marketing
netw. spec. infrastr. cost
network planning
provisioning and service mgmt
availability
number of users
average link length
network dimensions
number of boxes
network techonology
number of connections
Figure 1.10 Network characteristics as cost drivers for OPEX calculations
complexity, software updating, power consumption, or personnel specialization may render them unaffordable.
Figure 1.10 shows, in arbitrary units, the dependence between OPEX components and network characteristics, considered as cost drivers so as to appreciate the impact of network technology on operational expenditure. It is clearly observed that network technologies determine the cost for maintenance, reparation, provisioning, service management, and network planning. They have less impact on charging/billing and commercial activities.
A more detailed analysis of Figure 1.10 gives the following information:
.
The numb er of network elements (network components – e.g., routers, OXCs, PXCs) has an important impact on the cost for maintenance/reparation.
.
Network technology determines not only network performance, but also some specific cost of infrastructure (more or less floor space and need of energy).
.
Network dimension is important for all considered OPEX subparts, except AAA (run in centralized scheme) and marketing.
.
The number of connections strongly influences the cost for provisioning and service operation and management (each connection needs to be set up), but it is less important for network plan ning.
44 The Emerging Core and Metropolitan Networks
.
The number of users determines the cost for provisioning, network planning, charging/ billing, and marketing, but has a small impact on the cost of maintenance and reparation.
.
The average link length has little impact on maintenance, but may be significant for reparation cost when failures require a technician to go on site.

1.5.3 New Business Opportunities

1.5.3.1 The Business Template
Techno-economic drivers must let business progress, since network services are no longer of public strategic interest, covered by national monopolies. Such a situation is not new, but it is still evolving in accordance with clients’ demands and technical improvements. Just to complete the vision and help in understanding telecommunication network evolution, some ideas about market agents and their driving actions are presented here.
The technological evolution of network infrastructure leads it to becoming multifunctional. Hence, the old scheme of one network for one purpose and kind of client, in a vertical structure (see Figure 1.11), must be changed into a matrix scheme of services based on a unique transport infrastructure: all tasks related to network service provisioning are no longer repeated for every telecommunication business; for instance, AAA is common for IPTV, teleconferencing, or POTS, as well as QoS assurance systems, network configuration, or customer commercial issues. This scheme of cross-management can also lead to telecommunication companies exploding into several specialized companies that cover a part of the business (for more than one single network operator, perhaps). In addition, the market agents involved in the telecommunication business, from content providers to end custom ers, play their role freely in a cross-relational model (see Figure 1.12), where commercial interests should prevail.
So, companies have to redesign their business model along the guidelines discussed so far: a method of doing business by which telecom companies generate revenues, in order to set strategies and assess business opportunities, create or get profit of synergies and so align business operations not only for financial benefits, but also to build a strong position in the
Figure 1.11 Network operator business model scheme adapted to NGN concept, from a pyramid structure to a matrix one
Transport Networks Economics 45
C ontent provi der
C ontent provi der
C ontent provi der
Content provider
C ontent provider
C ontent provider
Content provider
Service
Service
Service
Service
provider
provider
provider
provider
(VASP)
(VASP)
(VASP)
(VASP)
Network
Network
Network
Network
Service
Service
Service
Service
provider
provider
provider
provider
Service
Service
Service
Service
provider
pr ovider
pr ovider
pr ovider
(VASP)
(VASP)
(VASP)
(VASP)
End users
C ontent provi der
C ontent provi derC ontent provider
C ontent provi der
Content provider
Network
Network
Network
Network
Service
Service
Service
Service provider
provider
provider
provider
Service
Service
Service
Service
provider
pr ovider
pr ovider
pr ovider
(VASP)
(VASP)
(VASP)
(VASP)
Figure 1.12 Telecommunication market agents and their cross-relationship just governed by commer­cial interests
market. In fact, network technology and market environment evolution affect all components of a business model, namely market segment, value proposition, value chain, revenue generation, and competitive environment. For these reasons, new business opportunities arise and the business model template has to be updated.
For the market segment, defined as the target customers group, it is clear that incumbent network operators may find virtual network operators as new clients and so the market liberalization generates new market segmentation. Each segment has different needs and expectations (e.g., business customers, banking) and companies create services for a specific types of customer;
3
in addition to customer types, geographical issues must also be considered
for market segmentation.
The segmentation of the market, in turn, can be based on various factors, depending on the analysis to be performed; for example, different parts of a network can be shared by different market segments; every market segment will have its own behavior, reflected in its demand
4
model.
Conversely, the way of using applications/services depends on the type of market segment to which they are focused (the same applications have different characteristics in residential and business areas, for example). Therefore, a proper market segmentation analysis should not only aim to map traffic demands onto transport services, but also tackle the heart of the network operator’s business model: the question of how a network operator is designing
3
Some broadband services target roughly the same segments, such as big enterprises (for instance, VPN services, virtual network operators, or regional licensed operators) and Internet service providers. But residential, business and public administration market segments are normally offered network services that specifically support applications for storage, grid computing, multimedia content distribution, and so on.
4
This aspect includes the applications and the way they are used by those customers: frequency, time-of-the-day distribution, holding time, and so on.
46 The Emerging Core and Metropolitan Networks
Web interface
UNI
User Network
Business Plane/ Service Plane
Management Plane
Business and
Business and
management plane
management plane
interworking
Management
Management and control plane
and control plane interworking
Control Plane
Data Plane
Figure 1.13 Interfaces user-network provider that may generate value proposition to different market segments
its access segment, which kind of alliance should be formed between network operators and content providers or virtual access services platform (VASP) taking into account traffic models derived from new customer demands like IPTV or new customer approaches to telec ommuni­cation issues like Web2.0. Finally, as far as network operators contract SLAs with their clients (availability and QoS), their profit depends on sharing their network infrastructure optimally.
The importance of properly segmenting the target market is critical in the development of a value proposition. The value proposition is identified by the combination of the customer needs, the services required to fulfill their needs, and the virtual value of the product.
5
NGNs allow network operators to design new products/services that can have value to some customers, like those derived from VPNs and temporary bandwidth leasing (BoD). In addition, customers are going to be able to access network services using new service interfaces and procedures. These interfaces can be provided on different planes (or any combination of them):
.
Service plane. VASP may offer their products with (or without) a middleware to their customers using this interface to access network resources, either via previous reservation or in a dial-in mode (with a certain service availability).
.
Management plane. This is the old way. Carriers establish SLAs and reserve network resources either for dedicated connections or by statistical knowledge of traffic. Internally, management/control plane interworking functions get the transport system (data plane) and set the required connections.
.
Control plane. Network operators let some users directly enter into their network cont rol system through the UNI. In this way (as established in an SLA), those customers can set up, modify, and tear down their own VPN once the required AAA process has been performed (by the management plane).
Thus, network operators’ migration activities should take into account that deployment of new services and functionalities may attract new users and increase the operator position in the market. Such an approach is worth it even if the net cash balance remains unaffected due to extra costs for implementing the new services.
5
The real value of the product is formed once customers select the service among available alternatives, according to its
functionality and price.
Transport Networks Economics 47
Figure 1.14 Market agents’ roles mapped to the value chain by network layersto show service interfaces
On the other hand, virtual network operators (VNOs) may find easier ways of creating specialized services (TV distribution, for instance) without any handicap derived from the necessity of maintaining old services, QoS systems, or complex AAA mechanisms: their value proposition is clearly identified by a market segment. This is just the opposite case of incumbent telecommunication companies, composed of a number of departments that offer differently valued proposition services; then, synergies, market position, and customer fidelity are their main assets.
The creation of new services also modifies the value chain that represents a virtual description of the environment in which the market agents offer their services. If market desegregation implies unbundling the value chain to exploit a set of cross-relationships (Figure 1.14); also, tradeoffs and vertical alliances can be formed. These alliances are interesting whenever they produce added value to end customers because of technical improvement or service enhancement (Figure 1.14).
In fact, each box of the value chain represents an activity (Figure 1.15), and groups of activities are generally covered by inde pendent market agents.
6
This value chain scheme illustrates the connections between the telecom network layers and the corresponding telecom services; layer handovers can be commercial transactions between different organizations or enterprises, or they can be internal (technical) interfaces without any real commercial transaction. For example, there may be fiber availability issues (with underlying ducts and rights of way); the access to this dark fiber is then a service (something to sell), and the seller is known as a “dark fiber provider.” The organization that is purchasing this service – the buyer­has also to place the necessary equipment infrastructure (WDM, SDH) in order to become a Network Operator.
7
Other buyer–seller market profiles can be identified through the value
chain in similar ways.
6
Here is a key to the identification of some roles: customer, packager, connectivity provider, access network provider,
loop provider, network service provider, application service provider, and content provider.
7
The light boxes denote the seller and the dark boxes the buyer of the service at that interface.
48 The Emerging Core and Metropolitan Networks
Figure 1.15 Impact of migration factors to business model components
In addition, the initial roles of the market agents can evolve according to commercial interests; for example, equipment vendors can develop network management activities, since their knowledge of the systems (hardware and software) allow them easily to propose value­added functions; in the same way, big media companies (content providers) may become service providers or, conversely, VASP may get exclusive rights on certain contents. Network operators can split old functionalities and even outsource some of them in order to concentrate their operations in either connectivity services (“bit transmission”) or services of higher quality closer to the end customer (with higher margins); on the other hand, the FMC can have different consequences in the role that network operators would like to assume in the future.
The role of network operators within the value chain scheme given in Figure 1.15 is going to comprise other activities, within the value chain, apart from selling pure network services; for instance:
.
Purchasing multimedia contents and distributing them throughout a nationwide platform for other (VASP) agents to deal with them.
.
Leasing or buying underlying network infrastructure (right of way, network equipment, fiber) from utility companies or other vendors and renting part of it to other network operators as their network migration is carried on.
.
Renting or selling buildings and complementary equipment (for cooling, for example) when they are no longer needed by new switching and transmission systems.
.
Service provisioning to other VNOs.
The revenue amount is the main assessment to determine the network operator’s ability to translate the value of the product into the money received. A typical revenue model is based on monthly subscription fees and costs of transactions, commissions, and services used by customers. A supplementary income can be achieved by selling or leasing a product/service to other, nonresidential companies. In each situation a network operator should consider an appropriate pricing model in order to maximize the revenue.
Transport Networks Economics 49
Moreover, the network migration strategy will be planned by a network operator, as long as this is possible, choosing the smallest cost of investment that may lead to greater profitability. Providing new services, after an initial investment, is a risk for several years to create new revenue-generation streams. Thus, increasing these margins is always an objective for any market agent before extending its role in the value chain or capturing new market. From the network operator point of view, it is always crucial to monitor the margins, since these are constantly change, and to plan network migration and to update the business plan, accordingly. This is not a straightforward operation; some puzzling considerations are described below:
.
New services typically mean more profits from higher income; however, the higher the number ofservices, the more complex the management of them, aside from the effort required to search newcustomers. Then, the higher the diversificationof services for value proposition and market segment capture, the bigger the problem with accounting and operating.
.
Higher network quality and closer matching between customers’ demands and network capabilities mean potentially more subscribed services and higher income. More subscribers, attracted by emerging, pioneer services lead to increasing income, provided that transport, control, and management do not overload network capabilities.
.
In general, however, a higher number of customers means a faster return of investment, thus allowing reduc tions in the prices for services; and customizing existing services allows proposing more services and making them used more frequently, thus generating higher incomes.
.
In addition, higher accessibility to the network services (through geographical expansion, higher availability derived from FMC or via interconnection facilities) and higher penetra­tion due to combined service proposals increase the number of new subscribers and help to keep their fidelity.
.
On the other hand, technical modifications affecting the traffic model will have consequences on the revenue streams. For example, the volume of metropolitan network traffic will be affected by the growth of social networks; network operators and VASP should then study carefully how IP networks are going to work in coordination with L1 and L2 transport layers and also analyze the importance of developing multicast-capable transmission systems, as well as the placement of service nodes in order to avoid bottlenecks. However, sometimes it can be proven that an extra initial investment (CAPEX) will be compensated by a consequent OPEX reduction: introduction of GMPLS for broadband standard networks will surely compensate the extra provision of bandwidth for some services unless network operators are obliged to share their transport capacity or extend it for the rest of their networks by the regulatory authority.
.
Incumbent network operators must finally face the challenge of designing their network migration, not only to reach more revenue generation and gain market position, but also to keep backward compatibility. This decreases the risk of the investment and allows one to perform the infrastructu re modernization in reasonable time schedules with the interesting possibility of reusing equipment.
In general, a higher value of the network and a higher position on the market allows financial advantages and new revenue streams for the telecommunication core business, like leasing offices and selling auxiliary equipment, since the eventually selected winning technology has lower requirements in terms of cooling, footprint and smaller real estate for the housing of the equipment.
50 The Emerging Core and Metropolitan Networks
Figure 1.15 summarizes the possible impact of the already analysed components on network migration factors and to NO’s business model. Perhaps all factors involved in a migration strategy will have some influence on the company competitiveness, and so a migration strategy is always designed having as a goal to increase revenues and/or to gain a better position in the competitive environment. In the latter case, a network operator can deploy new serv ices and functionalities to develop niche markets, thus attracting new customers, and achieve diversifi­cation of services so as to extend their market segment, attracting customers that prefer dealing with a unique network operator and get advantages of synergies for both network services and AAA tasks. The impact of regulatory framework on the migration strategies has, amongst other things, the following consequences from a NO point of view:
.
Changes in the regulatory framework taken from the governments may either accelerate or discourage new entrant NOs, FMC or inter company (vertical) alliances.
.
Vertical or horizontal alliances, even if they do not end up in one company’s takeover,modify the competition environment (for suppliers and clients, respectively). On the other hand, the advancement in standardization, in contrast to exclusive vertical (in the value chain) dependencies, also modifies the competitive environment of network operators and other telecommunication market agents.
Network operator clients are also taking advantage of new network service implementations for end-to-end broadband communication, to be based on the NGN concept; thus, the competitive environment is becoming harder, as network operators’ customers can trade off different network operator service propos als and use UNI to different transport infrastructures, so as to build up their own customized VPN (see Figure 1.16) or lease part of their bandwidth capacity, thus acting as VNOs.
1.5.3.2 Business Opportunities
VPN services, regardless to the association with FMC issues, seams to be the most important driver in telecommunication business plans in the short term. Currently, VPNs are mainly supporting legacy data services in circuit switched (CS) networks. However, the advances in
Figure 1.16 (a) Tradeoff network services from different network operators. (b) VPN concatenation to get extended services under a unified management. (CE: customer edge.)
Transport Networks Economics 51
Figure 1.17 VPN cascade either to get dynamic network resources allocation or to be permanently used by VNOs (CE: customer edge; PE: provider edge; P: provider)
network technologies for VPNs are making it possible for NOs to operate a packet switched (PS) transport infrastructure like a CS one. But there is more than simulating CS networks over PS networks with the economical interest derived from statistical multiplexing and resource sharing. The currently perceived image of a “dummy” IP network will be reconsidered with the advances made possible in VPN technology: a NO may take advantage of the VPN technology to provision restricted connectivity using diffserv
8
-instead of over-provisioning network capacity- offering, thus, differentiated SLAs. Hence, ISPs and other VASPs could implement their preferred mechanisms to ensure application quality (for information integrity and the speed at which it reaches their customers).
AVPN may simply be used to implement VNO network resources (Figure 1.17). This fact, and the possibility of part-time network resource renting as BoD for any network operator client, is produc ing a radical change in the telecommunication business.
Moreover, incumbent network operators can evidently profit from operating their transport resources by means of a new developed TE approach. This policy can be based on the VPN technology,together with a network operator’s ability to manage their capabilities and Ethernet network resources by reusing network elements (for the “virtuality” of the VPN paradigm) and make more scalable the management of the whole infrastructure by (virtually) dividing it and operating it as usual with the old pyramid scheme (Figure 1.17).
The expectations opened by new technologies and the market environment go beyond VPN-related topics. Improvements in network exploitation and new network service avail­ability also give network operators new opportunities derived from:
.
Cheaper and faster network upgrade (according to traffic demand) allowed by NGN. Furthermore, an NGN control plane makes it possible to adapt network resources quickly to client demands (a few minutes instead of days to reconfigure transport resources).
8
For packet-switched networks, diffserv concepts of QoS can easily be understood as a mechanism of setting transmission preferences (at edge or core nodes) for determined sorts of packets (over a simple best-effort approach): those of a specific kind of application (real-time class, for instance) or those of a privileged customer, VPN, and so on. The concept is not new, and examples of its implementation may be found in standards of ATM services or RPR architecture. For IP networks, this non-neutral network operator performance, beyond a plain flux control, is a hot issue actively opposed by ISPs that promote ruling a net neutrality.
52 The Emerging Core and Metropolitan Networks
.
Reduction of time needed to provide connections and new services.
.
Multicast traffic carried over p2mp connection solutions, implemented in new switching equipment, allow the introduction of specific architectures for triple-play services (namely IPTV or other multim edia applications).
.
Developing Ethernet solutions for extended networks allows carriers to reduce CAPEX for wide area networks and specific-purpose networks.
.
Migration from ring to mesh topologies happens to allow more efficient resilient strategies for packet-switching networks: one can think of restoration versus protection, for example.
.
Complete transparent optical networks (instead of hybrid solutions) for OSI level 1 are also expected to reduce CAPEX by eliminating undue O/E/O conversions. The possibility of directly managing wavelengths also has some OPEX advantages (wavelength-switched­oriented networks).
.
Parallel development of the service layer, over the control plane, is often addressed to Ethernet, new applications like storage area networks, video on demand, grid computing, and so on.
Aside from these improvements in network exploitation and in the new services arising from NGN architecture and deployment of new technologies, some collateral new business must be mentioned too, since their economical impact is not negligible.
VoIP and migration to the IPv6 protocol are making possible a complete quadruple-play offer, as well as easier ways of managing networks to provide any kind of networ k service, including those for real-time applications, by means of the same packet-switched-oriented equipment.
As new equipment is smaller and need less cooling and DC generators, a considerable amount of real estate is made redundant in central offices, thus allowing NOs to get extra revenues from their selling or renting. This new real estate business opportunity should be perceived as in an integral part of the current trend in extending the role of NO’s towards a value added service provider which will eventually turn them to their current antithetical pole i.e. the VASP could become a (V)NO offering specialized network services. Finally, the aforemen­tioned SLA diversification supporting dynamic network service opportunities, as well as a complex trade off between the factors affecting it, may spawn the appearance of a new market agent; that of the extended service broker. This is an agent serving as the “middle-man” between suppliers and customers, operating not only in a direct NO-client scheme but also under a dynamic and multi-step (VPN cascade, for instance) network service leasing pattern.

Acronyms

AAA authorization, authentication, and accounting ASON automatic switched optical network ATM asynchronous transfer mode BER bit error rate BoD bandwidth on demand CAPEX capital expenditure E-NNI external NNI
Acronyms 53
FMC fixed–mobile convergence GMPLS generalized multi-protocol label switching IETF Internet Engineering Task Force I-NNI internal NNI IPTV Internet protocol television ITU International Telecommunications Union LSC lambda switching capability LSR label-switched router MAN metropolitan area network MEF Metro Ethernet Forum NGN next-generation networks NNI network-to-network interface (see ASON) OCS optical circuit switching O/E/O optical/electrical/optical OIF Optical Internetworking Forum OPEX Operational Expenditures OSI open systems interconnection OSS operation service support OTH optical transport hierarchy OTN optical transport network OXC optical cros s-connect p2mp point-to-multipoint PBB provider backbone bridge PBB-TE provider backbone bridge traffic engineering PLR packet loss rate PoP point-of-presence PoS packet over SONET POTS plain old telephone service PSC packet-switching capability PXC Photonic Cross-connect QoS quality of service RPR resilient packet rings (refer to IEEE 802.17) RSPV-TE reservation protocol with TE SDH synchronous digital hierarchy (refer to the ITU-T framework) SLA service level agreement SONET synchronous optical network (refer to the ANSI framework) TDM time-division multiplexing (see SONET and SDH) UMTS Universal Mobile Telecommunication System UNI user-to-network interface (see ASON) VASP virtual access services platform VNO virtual network operator VoIP voice over IP VPN virtual private network WDM wavelength-division multiplexing WSON wavelength-switched optical network
54 The Emerging Core and Metropolitan Networks

References

[1] Spargins, J.D., Hammond, J., and Pawlikowski, K. (1991) Telecommunications: Protocols and Design, Addison-
Wesley.
[2] Dorf, R.C. (ed.) (1997) Electrical Engineering Handbook, 2nd edn, CRC Press. [3] MEF 10.1 Technical specifications (November 2006) Ethernet services attributes Phase 2. [4] IETF RFC 2702 (09/1999) Requirements for traffic engineering over MPLS. [5] Davie, B. and Rekhter, Y. (2000) MPLS Technology and Applications, Morgan Kaufmann Phublishers. [6] ITU-T Recommendation Y.1311 (03/2002) Network-based VPNs – generic architecture and service requirements. [7] IETF RFC4026 (03/2005) Provider provisioned virtual private network (VPN) terminology. [8] ITU-T Recommendation Y.1312 (09/2003) Layer 1 virtual private network generic requirements and architecture
elements.
[9] ITU-T Recommendation Y.1313 (07/2004), Layer 1 virtual private network service and network architectures.
[10] Tomsu, P. and Wieser, G. (2002) MPLS-based VPNs, Prentice Hall. [11] ITU-T Recommendation G.709 (03/2003) Interfaces for optical transport network (OTN). [12] Comer, D.E. (2003) Internetworking with TCP/IP Principles, Protocols and Architectures, Prentice Hall. [13] ITU-T Recommendations G.8080/Y.1304 (November 2001) Architecture for the automatically switched optical
network (ASON).
[14] IEEE 802.1Qay (2007) Standard provider backbone bridge traffic engineering. [15] Serrat, J. and Galis, A. (2003) IP Over WDM Networks, Artech house. [16] Kadambi, J. Crayford, I., and Kalkunte, M. (2003) Gigabit Ethernet Migrating to High Bandwidth LANs, Prentice
Hall.
[17] IETF RCF 5212 (07/2008) Requirements for GMPLS-Based Multi-Region and Multi-Layer Networks
(MRN/MLN).
2

The Advances in Control and Management for Transport Networks

Dominique Verchere and Bela Berde

2.1 Drivers Towards More Uniform Managem ent and Control Networks

We observe the convergence at different levelson the network given the application and service convergence as presented in Chapter 1 (and might be highlighted /summarized here); the convergence and the related Internet protocol “(IP)-orientation” in network services seems inevitable.
What is the network convergence? The converged IP-based service platform, despite the wide variety of network technologies, requires network control and management functions that tend to be uniform at the different switching layers in order to reduce operational expenditure (OPEX). The tran sition from “a network per a service” to “network integration of multi­services and multilayer with support for end-to-end quality of service (QoS)” concept enables one to get higher network resource utilization coupled with higher resiliency.
This section further elaborates the new role of the network control functions and network management functions relying on this integration at different transport layers by considering the network layer (i.e., IP layer or layer 3) with the convergence to IP-based services using the essentials of traffic engineering (TE) and QoS support, and how these requirements can be illustrated in IP or IP/multi-protocol label switching (MPLS) networks. Then we consider the data layer (i.e., layer 2) as essentially based on Ethernet with the transport MPLS (T-MPLS)
1
1
At the time this text was produced, key agreements have been defined between ITU-T SG15 and IETF leadership concerning T-MPLS/MPLS-TP evolution in December 2008. The agreements can be summarized in three statements: (i) there is no agreement, or proposal to cancel or deprecate the T-MPLS recommendations from ITU-T currently in force; (ii) ITU-Twill not undertake any further work to progress T-MPLS;and (iii) it is possible that the ITU-T will have a future role in MPLS-TP standardization.
Core and Metro Networks Edited by Alexandros Stavdas Ó 2010 John Wiley & Sons, Ltd
56 The Advances in Control and Management for Transport Networks
extensions, resilient packet ring empowered with TE and operation, administration, and maintenance functions. Finally, we consider the physical layer (i.e., layer 1) as essentially based on optical transmission networks with synchronous optical network (SONET)/synchro­nous digital hierarchy (SDH) and its next-generation SONET/SDH enhanced with data protocols such as generic framing procedures, virtual concatenation and link capacity adjustment schemes, G.709, and optical transmission technologies [33].
In circuit-switched layer transport networks, a layer (L1) path is constructed with physical links and ports, one or many optical wavelengths, or time-division multiplexing (TDM) timeslots, and an L1 path is established and dedicated to a single connection between the access points (APs) of the transport network for the duration of the applications using the connections.
In packet-switched networks, packets are forwarded hop by hop at each network element involved in the connection (e.g., IP routers or Ethernet switches) based on information in the packet header. An IP-switching-based network provides connectivity services while making efficient use of network resources by sharing them with many connections. It is based on the assumption that not all the provisioned connections need to use the resource all of the time. Packet-switched networks (PSNs) can be connectionless or connection oriented. PSNs can be based on the IP, such as IP networks (connectionless) or IP/MPLS networks (connection oriented), or can be based on the Ethernet protocol, such as Ethernet network (connectionless) or Ethernet/MPLS transport profile (MPLS-TP) networks (connection oriented). IP-based networks are said to be layer 3 and Ethernet-based networks are said to be layer 2 in reference to the Open Systems Interconnection (OSI) basic reference model [31].
In connectionless packet-switched networks, once the data packet is sent on the interface, the connection is available until further information is either sent or received at this same interface.
In connection-oriented PSNs, connections are established and maintained until connectivity is no longer required, regardless of whether data packet has been transmitted or not.
Network services are classified according to three assigned groups taking into account their switching capabilities, features, relations, and differences. The three groups identified are virtual private networks (VPNs) on transport layer 3, VPNs on transport layer 2, and VPNs on transport layer 1. Additional network services subclassification can be defined for the layer 3, such as public IP and business IP, extending the network service classification to five groups. The characteristics of each network service are usually described along with their performance parameters.
The different layers’ VPNs are explained from the perspective of connectivity, control, scalability, and flexibility. Mechanisms enabling the network “confidentiality” are described, such as MPLS and tunneling in PSNs, and wavelength services enabled by L1-VPN services from optical networks. How network services should match the connectivity requirements of distributed applications is not developed in this chapter, but the connectivity requirements of on-demand storage, (grid) cloud computing, or multimedia-based distributed applications are developed in [32].
Three classes of network services are identified based on the related switching and transmission layers in which they are provided to the network customers edges (Table 2.1):
1. Layer 1 network services provide “physical layer” services between the network client
and 5the provider network (server). The customer edge (CE) equipment belongs to the same L1-VPN as the other equipment of the provider network (provider edge (PE) nodes,
Drivers Towards More Uniform Management and Control Networks 57
Table 2.1 Network service modes versus switching layers
Network service mode Switch capability layer
Layer 1 Layer 2 Layer 3
Connectionless
packet-switched
Connection-oriented
packet-switched
Circuit oriented OTUx, (OTN)
Optical packet switching
STM-n (SDH), STS-n (SONET)
Ethernet IP
Ethernet/MPLS-TP, ATM,
frame relay
IP/MPLS
provider (P) nodes). The connections can be established based on TDM timeslots (SONET/ SDH), optical wavelengths, optical wavebands, or physical ports, such as Ethernet ports or fiber ports.
2. Layer 2 network services provide “data link layer” connection services to the CE equipment involved in the L2-VPN. At the interface, the forwarding of user data frame s is based on the control information carried in the data link layer headers, such as media access control (MAC) addresses (for Ethernet), virtual circuit identifier/virtual path identifier (for asynchronous transfer mode (ATM)) or data link connection identifier (frame relay). The customer layer 2 data frames are associated with the right L2-VPN by each PE node.
3. Layer 3 network services provide “network layer” services between the CE equipment involved in the L3-VPN. At each user-to-network interface (UNI), the forwarding of data packets is based on the IP address information embedded in the layer 3 header; for example, IPv4 or IPv6 destination address.
The network services have been generically labeled VPN at L1, L2, and L3. For each of these three types of network service, typical performances are defined and usually referenced by network operators to alloca te the connectivity service requests.
The layer 3 network service (such as IP) is, as already mentioned, divided into public and business IPs, where public IP is a “best-effort” service and business IP is a higher priority class of services that, for example, can handle latency-sensitive applications. “Business IP” is also presumed to guarantee higher bandwidth from CE to CE.
The VPN services on all layers, L1, L2, and L3, are divided into either a permanently configured network service (typically provisioned by a network management system (NMS)) or an on-demand service (typically triggered and signaled from a network controller) (Table 2.2). The permanent service is totally managed by the network operators, but the on-demand connectivity service can be triggered dynamically by the CE node through a suitable UNI.
L1 and L2 VPN services are further divided according to their availability; this can be high or low availability. The high-availability network services are normally configured with defined protection/restoration schemes [2], which offer an alternative way for carrying the traffic impacted by network failures. Different types of distributed application are identified and reported in [32], and for each case the application performance requirements are checked against the network services.
58 The Advances in Control and Management for Transport Networks
Table 2.2 Mapping applications into network services (light: application will run on this network
service; dark: more efficient implementation, white: no support for the application)
Public IP Bu sin e ss IP VPN - L3
VPN - L2
VPN - L1
perm anen t on-demand perm anen t, H i avail perm anen t, L o w avail on-demand, Hi avail on-dem an d , Low avai l
perm anen t, H i avail perm anen t, L o w avail on-demand, Hi avail on-dem an d , Low avai l
- Back-Up/ Restore
Storage
- Storage on Demand
- Asyncrhonous Mirroring
- Synchronous Mirroring
Gridcomputing
- UtilityGrid
- Compute Grid
- Data Grid
- VideoBroadcast(IP-TV)
Multimedia
- Video on Demand
- Video Download
- VideoChat
- NarrowbandVoice,data(VoIP,...)
- Digitaldistribution,digitalcinema
- Gambling
- Gaming
- Video conference
Tele-medicine/diagnostic
It should be noted that the same transport layers could be used to provide multiple VPN services. For example, a transport layer based on SDH can be used to provide layer 1 (e.g., TDM), layer 2 (e.g., Ethernet), and layer 3 (e.g., IP) VPN services. For this reason, it is important to distinguish VPN clients and VPN transport networks (server part). Both VPN transport and VPN client each have their own set of connectivity inputs and outputs known as APs.
When the VPN client switching capability layer and VPN transport switching capability layer are different, the VPN client layer information must be adapted for transmission across the transport VPN. Examples of adaptation functions include multiplexing, coding, rate changing, and aligning (which may also include some form of fragmentation and sequencing if the transport layer traffic unit is smaller than the service layer traffic unit). Even if VPN transport layer trail/connectionless trail terminat ion functions are VPN client layer indepen­dent, adaptation functions must exist for each transport–client pair defined [41]: adaptation functions are required between APs in the VPN transport layer and control plane planes/ forwarding planes at the VPN client layer.

2.2 Control Plane as Main Enabler to Autonomic Network Integration

Telecommunications equipment designers face a huge task in making the optical networking dynamically configurable according to the IP traffic demands of customers. Provisioning connections from NMSs or signaling connections from the network control planes based on real-time traffic patterns require the ability to manage the interactions between the IP-layer functionality of packet networks and of the lower transport layers (L2 or L1 networks).
Control Plane as Main Enabler to Autonomic Network Integration 59
End-to-end control issues in transport networks with multiple technology layers and multiple administrative or organizational domains with multi-vendor equipment are becoming common. This is the role of the control plane to bring an answer in the form of tools, algorithms, automated processes, and common management interfaces to these issues, with the automated connection provisioning capability.

2.2.1 Generalized Multi-Protocol Label Switching

Derived from MPLS and driven by the Internet Engineering Task Force (IETF) Common Control and Measurement Plane (CCAMP) working group, generalized MPLS (GMPLS) has drawn a lot of attentio n on enabling different network layers to interact automatically. GMPLS provides uniform control capabilities from the optical transport network (OTN) up to the IP client layer to permit open and scalable expansion for next-generation transport networks.
This section will examine the environment for GMPLS control architecture and describe the following topics:
.
Primary drivers that compelled the developers of GMPLS to come up with a standard architecture.
.
MPLS/GMPLS evolution.
.
GMPLS architecture.
.
Fundamental technology components – protocols (and others, such as path computation element (PCE)).
.
GMPLS functionality – resource discovery, routing, signaling, and link management.
.
Goals of GMPLS – operation automation, flexibility, scalability, restoration, path selection optimization, and TE capabilities.
.
Primary current and future benefits and applications of GMPLS.
.
What are competing or complementary standards to GMPLS and how do they compare?
.
How do GMPLS-related equipment features and functions map to service provider requirements?
.
Specific economic benefits of GMPLS with reference to capital expenditure (CAPEX) and OPEX.
2.2.1.1 Primary Driver s to GMPLS
In multilayer transport networks, where IP, ATM, Ethernet, SONET/SDH, and fiber/port switching devices must constantly cooperate, GMPLS was specified with the objectives to automate the connection service provisioning, and, especially, with reliability and TE capabilities. This allows the combining of different network layers for service delivery: IP and Ethernet (also with GMPLS), and so on, down to optical switching.
Service providers need to reduce the cycle times for service provisioning. Adopting a new architecture including GMPLS leads to a reduced length of provisioning times, which stems primarily from the requirements to manually set up interconnectivity between network partitions, such as SONET/SDH rings, and networks. For instance, the time period to provision in a traditional optical network has been estimated at an average of 15–20 days
60 The Advances in Control and Management for Transport Networks
from the service order through to final test, where the actual provisioning of the physical network takes a major part. With the automated service subscription, one of the primary goals of GMPLS is to automate provisioning with cutting cycle times down to a few seconds [32].
Deploying and building out regional or national fiber infrastructure, or simply deploying expensive switching equipment, require operational cost reduction. Manually provisioned billed services with the incident service support do not fit this figure of expenditure. The reliability and protection guarantees for, especially, voice, TV, and video services over a multilayer network, not only SONET/SDH, pose the demand on operational machinery running at a reduced cost.
2.2.1.2 From MPLS Evolution to GMPLS Consolidation
MPLS, drafted in 1999 by the IETF for the convergence of IP, ATM, and frame relay layer technologies, provides higher speed data forwarding with the network internal options of TE and support for QoS. The predictability and reliability was then brought to IP networks with MPLS, with the efficiency of automation for connection establishment.
The rout ers, at the ingress of a network, inject a fixed-format label positioned at the layer 2 protocol header of a data packet as it enters the network. At every label-switched router (LSR), the pair of values built from the incoming label number and the incoming interface (i.e., port) determines the route. The complete path is actually predetermined at the beginning of the route at the ingress LSR. The data packet flows are carried over the transport network and signaled as what is called the label-switched path (LSP). Given that the LSRs perform a much less time-consuming label examination, and not a packet header-maximal matching forward­ing, the LSPs enable significant speed improvements with reference to a traditional IP forwarding network.
Moreover, extending MPLS to other network technologies first requires separating the label control information from the packet header. Next, MPLS uses in-band control, meaning that the control information is sent with the actual data traffic flows. Also, the IP addressing schemes need to be adapted to other technologies. Another issue comes from the MPLS label formed as a number of up to 32 bits.
All these questions were addressed by the IETF in order to accommodate MPLS into the switching capabilities of other network layer, and that below the IP layer; that is, L2 and L1 networks. The result is the standard GMPLS architecture that allows generalizing the label switching to non-packet-switching technologies, such as time division (e.g., SONET/SDH, PDH, G.709), wavelength (lambdas), and spatial switching (e.g., incoming port or fiber to outgoing port or fiber).
2.2.1.3 Architecture to GMPLS
In the IETF standard [42], GMPLS addresses multiple switching layers and actually covers five groups of switching types. The switching type of a network element defines the data frames that the element can receive, switch, and control; it corresponds to the switching capability layer to which the network element can demultiplex the data signal from an
Control Plane as Main Enabler to Autonomic Network Integration 61
input interface, to switch it and send it to the output interface. The switching types are ordered among switching capabilities as follows:
1. Packet-switch-capable (PSC) interfaces: interfaces that recognize packet boundaries and can forward data based on the content of the packet header. These devices, such as IP or ATM routers, can also receive routing and signaling messages on in-band channels.
2. Layer-2-switch-capable (L2SC) interfaces: interfaces that recognize frame/cell boundaries and can forward data based on the contents of the frame/cell. Examples include ATM, frame relay, Ethernet, and its evolution towards MPLS-TP.
3. TDM-capable interfaces: interfaces that forward data based on the data’s time slot in a repeating synchronous cycle. Examples are SDH/SONET cross-connects and add–drop multiplexers.
4. Lambda-switch-capable (LSC) interfaces: interfaces that forward data based on the wavelengths on which data is received. Examples include photo nic cross-connects (PXC) or optical cross-connects (OXC). These devices can operate either at the level of an individual wavelength or a group of wavelengths, called waveband-switching equipment.
5. Fiber-switch-capable (FSC) interfaces: interfaces that forward data based on the position of the physical interfaces. Examples are PXC or OXC equipment.
The importance of the hierarchy comes from the multi-switching capability that can occur on the same interface, or between different interfaces. As described in [42], a circuit can be established only between, or through, interfaces of the same type. Depending on the particular technology being used for each interface, different circuit names can be configured; for example, SDH circuit or optical trail. In the context of GMPLS, all these circuits are signaled with a common control entity named the LSP.
In a GMPLS-controlled OTN, the labels are physical resources and are used to signal the route of the data stream it carries. Each label is specific between connected nodes, and the values of the labels require to be controlled with the same meaning. In [44–46], labels are given a special encoding so that the referenced resource (SDH, SONET, G.709 or wavelength) can be deduced automatically without the need to be configured and negotiated through the link management protocol engine [47].
The importance of GMPLS, and especially in multilayer networks, comprising multiple but collaborative network technologies, comes from its TE capabilities. TE is defined as that aspect of network engineering dealing with the issue of data and information modeling, performance evaluation, and optimization of operational networks. TE encompasses the application of technology and scientific principles to the measurement, characterization, modeling, and control of Internet traffic [68]. An important objective of TE is to facilitate reliable network operations. These operations can be facilitated by providing mechanisms that enhance network integrity and by embracing policies emphasizing network survivability. This results in a minimization of the vulnerability of the network to service outages arising from errors, faults, and operational failures.
Using the word “multilayer” means that two (or more) switching-capability layers collabo­rate in the network. In particular, with GMPLS-based control functions, the switching­capability layers may be IP, Ethernet, SDH, and optical transport.
62 The Advances in Control and Management for Transport Networks
Figure 2.1 MLTE and service allocation modes
As an illustration for TE-like optimization, consider multiple IP/MPLS services provided to clients through an IP/MPLS over an optical network. The illustration presents three ways for an IP/MPLS services request to be allocated over optical network resources in Figure 2.1. It is considered that the optical connections (e.g., TDM LSP or higher switching capability type) are used as TE links and constitute a virtual network topology (VNT) for an IP/MPLS network. The hierarchy of the switching-capability layers (defined in Table 2.1) is used to define that the IP/ MPLS layer is a client of the optical network layer.
The first allocation mode (case (a) in Figure 2.1) consists of designating separate network control instances for each network service request. This means that each IP/ MPLS-based serv ice request is built separately from a data flow and is routed and allocated through one IP/MPLS control instance. IP/MPLS service requests are sent independently to the optical network layer. The acceptance of one IP/MPLS service request by the optical connections may influence acceptance for the subsequent service requests from IP/MPLS network client.
The second allocation mode (case (b) in Figure 2.1) consists of combining IP/MPLS service requests into a common network control instance. This causes the service requests to be allocated at the IP/MPLS layer first; for example, in IP/MPLS routers, data packet flows associated with different IP/MPLS service requests will be processed in the same logical queues. Note that in this case the optical network server is still not optimally used because some optical connections can remain idle at a time when their capacity can be used in speeding the services of the other IP/MPLS connection requests. By handling the service requests onto a single IP/MPLS control plane, it is possible to recover some wasted capacity (e.g., due to nonoptimal light-path filling) and balance the service requests between the optical layer resource usage. To overcome this inefficiency, the TE at the IP/MPLS layer and at the optical layer are merged to produce a single multilayer TE (MLTE) control instance.
The third allocation mode (case (c) in Figure 2.1) uses MLTE joint optimization. A single MLTE control instance can optimize both IP/MPLS service requests, meaning that, for
Control Plane as Main Enabler to Autonomic Network Integration 63
example, traffic flow measurements from several service requests are collected, combined, and optimized into one IP/MPLS connection. The measurement from this optimal combination IP/ MPLS service is used in routing the service over the optical network. This is different from the previous allocation mode, where measurements (and MLTE actions) are performed separately. Since the traffic flows of several IP/MPLS services must be carried on a single IP/MPLS layer, a joint optimization will improve resources assignment, leading to the removal of idle connection servers in the optical layer. Contention may still exist when the joint IP/MPLS service requests are superior to the global optical network capacity.
2.2.1.4 Fundamental Technology Components
The GMPLS control plane is made of several building blocks. These building blocks are based on discovery, routing, and signaling protocols that have been extended and/or modified to support the TE capabilities of MPLS and then GMPLS. Each network element with a GMPLS controller in general needs to integrate a specialized link protocol for identifying and referring each data channel explicitly that the adjacent nodes can reference accurately. This complexity is due to the separation between the data transport network plane and the control network plane; consequently, there is no direct mapping between the data channels and the control channels. In order to support the network element discovering the data link capabilities and their identifiers, a control protocol for link management, called the link management protocol (LMP) [47] is defined.
However, fundamental technology building blocks to GMPLS include not only protocols (i.e., routing, signaling, and LMP), but also new elements developed to include brand new concepts. It is important, indeed, to describe, at least, two new notions for the GMPLS framework: the TE link and forwarding adjacency (FA) LSP concepts. A TE link is a representation of a physical link. The link state advertisements (LSA) is used by the routing protocol and stored into the link state database, also called the TE database (TEDB), to advertise certain reso urces, and their properties, between two GMPLS nodes.
This logical representation of physical links, called TE Links, is used by the GMPLS control plane in routing for computing and selecting the resources and in signaling parts for establish­ing LSPs. The TE extensions used in GMPLS corresponds, therefore, to the TE (link) information of TE links. GMPLS primarily defines, indeed, additional TE extensions for TDM, LSC, and FSC TE, with a very few technology-specific elements.
The FA-LSP corresponds to an LSP advertised in the routing domain and stored in the TE link state database as a point-to-point TE link; that is, to an FA. That advertised TE link no longer needs to be between two direct neighbors, as the routing protocol adapts the TE link information to that indirect neighborhood. Importantly, when path computation is performed, not only just conventional links, but also FA-LSPs are used.
From the point of view of protocol engines, extensions to traditional routing protocols and algorithms are needed to encode uniformly and carry TE link information. Therefore, for GMPLS, extended routing protocols were developed; that is, intermediate system to intermediate system (IS-IS)-TE and open short path first (OSPF)-TE. In addition, the signaling must be capable of encoding the required circuit (LSP) parameters into an explicit routing object (ERO). For this reason, GMPLS extends the two signaling protocols defined for MPLS-TE signaling; that is, resource reservation protocol (RSVP)-TE [50] and
64 The Advances in Control and Management for Transport Networks
constraint-based routing–label distribution protocol (CR-LDP) [49] described in the signal­ing functional description [48]. GMPLS further extends certain base functions of OSPF-TE and IS-IS-TE and, in some cases, adds functionality for non-packet-switching-capable networks.
2.2.1.5 GMPLS Functionality
The GMPLS framework separates the network control plane containing the signaling and routing protocol. The fundamental functional building blocks of GMPLS can be structured as:
.
resource discovery
.
topology/state information dissemination
.
path selection
.
routing
.
signaling
.
link man agement
.
restoration and protection
.
management information base (MIB) modules
.
other functionality.
GMPLS was designed for integrating MLTE and multilayer recovery mechanisms in the
network. In addition, we are specifically interested in the promising alternatives of:
.
Scalability, with reference to the number of protocol messaging. Practical limitations on information processing may exist in equipment.
.
Stability,or more properly protocol convergence times. For MLTE, especially in upper layers that see logical topology updates, control plane traffic may be bursty, and large networks may see large convergence times.
Given the collaborative aspects in MLTE when running GMPLS protocols, the processing of the vast amount of information, which may potentially be emitted by network devices running GMPLS, the following sections briefly present base functions with reference to the funda­mental building blocks.
Resource Discovery
The use of technologies like dense wavelength-division multiplexing (DWDM) may imply a very large number of parallel links between two directly adjacent nodes (hundreds of wavelengths, or even thousands of wavelengths if multiple fibers are used). Such a large number of links was not originally considered for an IP or MPLS control plane, although it could be done. Moreover, the traditional IP routing model assumes the establishment of a routing adjacency over each link connecting two adjacent nodes. Having such a large number of adjacencies does not scale well. Each node needs to maintain each of its adjacencies one by one, and the link state routing information must be flooded throughout the network.
To solve these issues, the concept of link bundling was introduced. Moreover, the manual configuration and control of these links, even if they are unnumbered, becomes impractical. The LMP was specified to solve the link management issues.
Control Plane as Main Enabler to Autonomic Network Integration 65
Topology/State Information Dissemination
The goal of information dissemination is to provide information on the TE link and its attribute information in order to allow LSRs to select an optimized path based on TE criteria. A primary goal of the GMPLS routing controller in handling this information is to deal with the problems of scalability that are essential to multilayer networks. The impact of control plane limitations is dependent on the “meshedness” of the topology formed in the control plane. This means that, for sparser topologies, greater (relative) changes occur in the GMPLS logical topology when traffic patterns shift.
Since the number of physical ports on an OXC may be large, GMPLS devises the concept of a bundled link, used to minimize, via aggregation, the amount of information that is propagated over the network. Bundle links are parallel links, equivalent for routing purposes, which share common attributes used for path selection.
When distributing topology information to other sub-networks, bundle links can be used and only aggregate information is provided. There is a trade-off between the granularity of the information disseminated and the scalability; balancing this trade-off is not trivial.
Link state information has both static and dynamic components. Examples of static information include neighbor connectivity and logical link attributes, while dynamic ones include available bandwidth or fragmentation data. Similar to the scalability issues regarding topology information, the amount of link state information to be conveyed has to be controlled. The premise is to distribute only what is absolutely necessary. Although all link state data mus t be set up in the initial database, only changing information needs to flood the network. There are rules used to set controls to determine when to send information: static information changes, exceeding of a threshold, or periodic refreshment of topology information are good examples.
Path Computation and Selection
Path selection means a constrained path computation process. Upon the LSP request, the LSR checks out all the request admission rules related to this request. These rules, plus some user­specified rules carried by the LSP request, are parsed into constraints that instruct the LSR only to retrieve the related resources from the TEDB and create a copy of the retrieved resource information in the memory. Constraint shortest path first (CSPF) computation is carried out on this reduced TE information to select a routing path.
Routing
MPLS developed the application of constraint-based routing, which added extensions to existing routing protocols such as OSPF and IS-IS. These extensions were designed to enable nodes to exchange control information about topology, resource availability, and policy constraints [50]. RSVP-TE and CR-LDP establish the label forwarding state used to compute the path. In optical networks, the dynamics of routing are made significantly more complex by the potentially huge number of numbered and unnumbered links (e.g., the impossibility of assigning IP addresses to individual fibers, lambdas, and TDM channels) and the difficulty in identifying physical port connectivity information.
One of the important extensions to MPLS to address routing issues in optical networks is the concept of the LSP hierarchy. LSPs of the same type (i.e., FSC, L2SC, LSC, TDM or PSC) that enter a network on the same node and leave a network on the same node can be bundled together; that is, aggregated and tunneled within a single LSP. This handles the fact that optical networks have discrete bandwidths that are largely wasted if they are transmitting a lower
66 The Advances in Control and Management for Transport Networks
bandwidth stream from, for example, an IP network. Rather than use an entire 2.488 Gbit/s optical link for a 100 Mbit/s LSP data stream, the lower bandwidth stream can be tunneled through the optical LSP leaving the rest of the bandwidth for other data flows. There is a hierarchy that dictates the order in which these LSPs are aggregated or nested. The hierarchy orders the FSC interfaces at the higher order, followed by LSC interfaces, followed by TDM interfaces, with PSC interfaces at the bottom. The other important benefit of this hierachy [34] is the aggregation of information that would otherwise need to be disseminated in its original detail over the network.
Signaling
A generalized label request in GMPLS is signaled with RSVP-TE and it includes three kinds of information: (i) the LSP encoding type, which represents the nature of the LSP and it indicates the way the data are framed in the LSP (values represent packet, Ethernet, SDH or SONET, lambda, fiber or digital wrapper); (ii) the switching type used by the LSP that is being requested on a link, and this value is normally the same across all the links of an LSP; the basic GMPLS switching types are PSC, L2SC, TDM switch capable, LSC, or FSC; (iii) the generalized payload identifier (G-PID) is generally based on standard Ethertypes for Ethernet LSPs or other standard for non-packet payloads such as SONET/SDH, G.709, lambda encodings. The establishment of the LSP itself (i.e., the reservations of physical resources (interfaces) between adjacent nodes) is done through signaling messages [29].
Adjacent network elements have four basic interactions that involve signaling messages: create, delete, modify, and query messaging. Create means to create a hop that is within an end-to-end path. The create request requires information about the hop list, since the upstream node also suggests a label to the next node. Suggesting the label in advance means that the upstream node can begin to configure its hardware in advance, thus saving time and reducing latency. However, the suggested label can be rejected by the next node, in which case the upstream node will have to reconfigure itself. Since lower order LSPs can tunnel through higher granularity LSPs [34], the create request needs to include the ID of the higher order LSP in its hop list. Delete transactions deallocate a link, meaning it must be torn down. The modify request changes the parameters of the link and the query request asks for the status of a link within an LSP.
Link Management
To enable communication between nodes for routing, signaling, and link management, control channels mus t be established between a node pair. In the context of GMPLS, indeed, a pair of nodes (e.g., photonic cross-connects) may be connecte d by tens of fibers, and each fiber may be used to transmit hundreds of wavelengths in the case that DWDM is used. Multiple fibers and/or multiple wavelengths may also be combined into one or more bundled links for routing purposes.
Link management is a collection of procedures between adjacent nodes that provide local services such as:
.
control channel management, which sets up the out-of-band control channel between the nodes;
.
link verification, which certifies that there is connectivity between links;
.
link property correlation, confirming that the mappings of interface IDs and aggregate links are consistent;
Control Plane as Main Enabler to Autonomic Network Integration 67
.
fault management to localize which data link has failed;
.
authentication, which confirms the identity of the neighboring node.
The LMP has been defined to fulfill these operations. The LMP has been initiated in the context of GMPLS, but it is a generic toolbox that can also be used in other contexts. Control channel management and link property correlation procedures are mandatory per LMP. Link connectivity verification and fault management procedures are optional.
The control channels that are used to exchange the GMPLS control information exist independently of the managed links.
Restoration and Protection
GMPLS enables a physically separate control channel from the data bearer channel for the signaling to be done out of band. It is especially relevant for protection and restoration, as one among the most important GMPLS extensions to MPLS. Through signaling to establish back­up paths, GMPLS offers the option of both protection against failed links and protection against a failed path. When a route is being computed, the ingress LSR also determines a back-up path through the network (also termed a secondary path [7,8]). When protecting on an end-to-end basis, the concept of shared risk groups (SRGs) can be used to guarantee that the secondary path does not share any physical links in common with the original one. The terminology used in GMPLS for the types of end-to-end protection are [9]:
.
1 þ1 protection. simultaneous data transmission over two physically disjoint paths and a selector is used at the receiving LSR to choose the best signal.
.
M: N protection. M pre-allocated secondary paths are shared between N primary paths. Data is not replicated on to the back-up path but is assigned to that path only in the case of failure.
.
1: N protection. one pre-allocated secondary path is shared among N primary paths.
.
1: 1 protection. one dedicated secondary path is pre-allocated for one primary path.
GMPLS allows per-segment-based path protection mechanisms [16].
Restoration is a reactive process, involving troubleshooting and diagnostics, dynamic allocation of resources, and route recalculation to determine the cause of a failure and route around it. The stages of restoration include failure detection, failure notification, traffic restoration, and post restoration. Failure detection can differ depending on the interface. In SONET/SDH, the change in performance of a physical signal may trigger an alarm indication signal. When the failure notification phase is complete at the node, the controller then triggers a restoration scheme.
GMPLS provides for line, segment, and path restoration. Paths can be dynamically restored; that is, they can be rerouted using an alternate intermediate route, if a path fails. The route can be between (i) the two node adjacent to the failure in the link restoration case, (ii) two intermediate nodes in the segment restoration case or (iii) the two edge nodes (ingress node, egress node) in the path restoration case.
GMPLS provides some unique options for restoration that allow flexibility. Mesh restora­tion, for example, means that instead of dedicating spare capacity for a back-up connection for each path, the originating node re-establishes the connection using available capacity after the failure event. Optimization algorithms help to restore more efficiently; for example, alternate routes can be precomputed by the originating node and cached in case of need. Restored paths
68 The Advances in Control and Management for Transport Networks
may reuse original path nodes or new ones, but using existing ones is more efficient since the new nodes could be used as elements in other LSPs. In the section on GMPLS applications and benefits, the use of mesh protection in providing tiered services will be discussed. Although SONET/SDH voice channels require restoration in 50 ms or less due to the quality require­ments of voice, some compelling arguments can be made for cost savings using mesh restoration techniques to use bandwidth more efficiently for data services.
MIB Modules
The introduction of GMPLS shifts some provisioning roles from network operators to the network service providers, since GMPLS controllers are managed by the NMS. The service provider should utilize an NMS and standard management protocols, such as the simple network management protocol (SNMP) [69–71] – with the relevantMIB modules – as standard interfaces to configure, monitor,and provision LSRs. The service provider may also wish to use the command line interface (CLI) provided by vendors with their devices.
SNMP MIB modules require additional flexibility, due to the versatility of the GMPLS control plane technology, to manage the entire control plane. Based on the Internet-standard management framework and existing MPLS-TE MIBs [72], various MIBs were developed for representing the GMPLS management information [73], and the work is still ongoing. As an example, the following is a summary of MIB objects for setting up and configuring GMPLS traffic-engineered tunnels:
.
tunnel table (gmplsTunnelTable) for providing GMPLS-specific tunnel configuration parameters;
.
tunnel hop, actual tunnel hop, and computed tunnel hop tables (gmplsTunnelHopTable, gmplsTunnelARHopTable, and gmplsTunnelCHopTable ) for providing additional configu­ration of strict and loose source routed tunnel hops;
.
performance and error reporting tables (gmplsTunnelReversePerfTable and gmplsTunnel­ErrorTable).
Other Functionality
An “LMP adjacency” is formed between two nodes that support the same LMP capabilities. Multiple control channels may be active simultaneously for each adjacency. A control channel can be either explicitly configured or automatically selected; however, LMP currently assumes that control channe ls are explicitly configured while the configuration of the control channel capabilities can be dynamically negotiated.
For the purposes of LMP, the exact implementation of the control channel is left unspecified. The control channel(s) between two adjacent nodes is no longer required to use the same physical medium as the data-bearing links between those nodes. For example, a control channel could use a separate wavelength or fiber, an Ethernet link, or an IP tunnel through a separate management network.
2.2.1.6 Advantages of GMPLS
Scalability
Optional configuration can be used to increase the scalability of GMPLS for large transport networks, and especially in the addressing and the routing. The concepts of unnumbered links
Control Plane as Main Enabler to Autonomic Network Integration 69
and link bundling were introduced with the extensions to signaling (RSVP-TE and CR-LDP) through an ERO and record routing object (RRO) and routing (OSPF-TE and IS-IS-TE) protocols through the TE-link attributes, for com bining the intelligence of IP with the scalability and capacity of optical transport layers. These two mechanisms can also be combined. In addition, the hierarchical LSP concept and the LMP both contribute to the improved scalability and, therefore, to the performance of the control plane, which should not depend on the scale of the network and should be constant regardless of network size.
Flexibility
GMPLS brings the mechanisms and functions to the control plane in the network in a way that it should have flexibility and provide full control and configurability in the task of optimizing the use of network resources by provisioning LSPs. In TE, GMPLS technical components provide and maintain flexibility so that the network-level coordination of the resource is optimal.
Operation Automation
From the operational cost savings due to the automation capability of GMPLS, the control plane relieves operators from unnecessary, painful, and time -consuming manual operations of network service provisioning. Reduction in the provisioning cycle of LSPs due to automation helps service providers – even running proprietary element and NMSs – to provision services at a reduced cost.
Adopting GMPLS means the automation and TE capabilities spanning multiple service provider domains. It means also that equipments from different vendors can be integrated in the same network.
Path Computation, Selection, and Optimization
This complexity of path computation usually exceeds the computational capabilities of ordinary LSR. A specialized network element has been proposed by the IETF to overcome this problem.
The PCE serves as a computing entity, specialized for constraint-based path computation. The network entities (routers, switches, NMS, or other service element) requesting the PCE service is called a path computation client (PCC) [38]. The protocol requirements between both entities are proposed in [37]. The PCC request includes source and destination node and additional path constraints. The PCE responds with a NO-PATH object if no path is found or it includes a list of strict or loose hops if a path has been found.
In this study we compared the newly proposed five different PCE approaches of the newly available IETF draft by Oki et al. [36]:
.
single multilayer PCE
.
PCE/VNT manager cooperation
.
multiple PCEs with inter-PCE communication based on the PCE protocol
.
multiple PCEs without inter-PCE communication
.
multiple multilayer PCE.
Our comparisons are quantitatively in respect to the path setup delay in an SDH over wavelength-division multiplexing (WDM) network scenario. We assumed a recon-
70 The Advances in Control and Management for Transport Networks
figurable wavelength-switched optical network as well as a reconfigurable TDM-switched network. Additionally, the single multilayer PCE approach is also experimentally evaluated in the MLTE demonstration scenario described in Section 4.4.
We found that the amount of multilayer paths, which legitimate the argument of PCEs, is very low in the considered SDH/WDM core network scenario. Thus, in our scenario, PCEs are legitimated more by the complexity of const raint-based path computation requests and by the reduced computation time than by extensive ML path computation.
Because of the small frequency of multilayer paths, the minimum and mean path setup times do not show much difference in our scenario. The expected path setup delay is in the order of tens of milliseconds. The maximum path setup delay (representing multilayer paths) triples the path setup delay in certain scenarios.
W e found that, among all PCE deployment scenarios, one single multilayer PCE performs best. In all cases, path setup delays are far less than a second even in the case of multilayer paths. The small communication overhead and the reduced number of PCEs needed back up this decision.
Multilayer/Multiregion TE Support
The so-called multilayer/multiregion network (MRN)concept is closely linked to GMPLS [51]. A region is a control-plane-related notion and is based on interface switching capability [34]. A GMPLS switching type (PSC, L2 SC TDM, LSC, FSC) describes the ability of a nod e to forward data of a particular data plane technology and uniquely identifies a network region. On the other hand, a network layer is a layer in the data plane; typical ly, based on client–server relationship, the usual layers are PSC over TDM, PSC over LSC, and PSC over PSC; that is, interface switching capabilities.
A network comprised of multiple switching types controlled by a single GMPLS control plane instance is called an MRN. The notion of LSP region is defined in [34]. That is, layers of the same region share the same switching technology and, therefore, need the same set of technology-specific signaling objects. In general, we use the term layer if the mechanism of GMPLS discussed applies equally to layers and regions (e.g., VNT, virtual TE-link), and we specifically use the term region if the mechanism applies only for supporting an MRN.
MLTE in GMPLS networks increases networ k resource efficiency, because all the network resources are taken into account at the same time. However, in GMPLS MRN environments, TE becomes more complex, compared with that in single-region network environments. A set of lower layer FA-LSPs provides a VNT to the upper layer. By reconfiguring the VNT (FA-LSP setup/release) according to traffic demands between source and destination node pairs of a layer, the network performance factors (such as maximum link utilization and residual capacity of the network) can be optimized.
Expectation of Service Provider
With control plane resilience, the network element can discover the existing cross-connects after recovering from the control plane protocol failure. For example, when control plane failure only occurs within one network element, the LSR, such as an OXC, will still be in place carrying data traffic. After recovery of the control plane, the network element should automatically assess the data plane (i.e., the OXCs here) and reconfigure its control plane so that it can synchronize with other control plane entities.
Flexibility of the transport layers means a fair allocation of bandwidth between competing routes dealing with bursts of activity over many timescales. Reconfigurability increases network flexibility and responsiveness to dynamic traffic demands/changes.
Control Plane as Main Enabler to Autonomic Network Integration 71
The service provider can also set up the service where the network dynamically and automatically increases/decreases bandwidth as traffic volumes/patterns change. If the demand for bandwidth increases unexpectedly, then additional bandwidth can be dynamically provisioned for that connection. This includes overflow bandwidth or bandwidth over the stated contract amount. The triggering parameters may be utilization thresholds, time of day, day of month, per-application volumes, and so on.
Bandwidth-on-demand (BoD) provides connectivity between two APs in a non-preplanned, fast, and automatic way using signaling of GMPLS. This also means dynamic reconfiguring of the data-carrying capacity within the network, routing, and signaling, and that restoration is also considered here to be a BoD service.
Economic Models and Benefits of GMPLS
To achieve ever-greater efficiencies, service providers must streamline their operations by reducing the number of people required to deliver services, and reducing the time required to activate and to troubleshoot network problems. To accomplish these objectives, they are focusing on automated provisioning through a distributed GMPLS control plane, which is designed to enable multi-vendor and multilayer provisioning in an automated way. Therefore, requests for services in the data network that may require connectivity or reconfigurations can happen in a more automated fashion. In addition, instead of provisioning on a site-by-site basis, the control plane creates a homogeneous network where provisioning is performed network-wide.

2.2.2 Evolution in Integrated Architectures

2.2.2.1 The Path Computation Engine
The interior gateway protocols (IGP) in IP networks relies on fully distributed routing functions. Each network element that is part of the routing domain has its own view of the network stored inside the IGP routing table (link-state database). For scalability, performance, and security reasons, link-state routing protocols (e.g., IS-IS-TE and OSPF-TE) are used in today’s carrier networks. Constraint-based path computation, typically using CSPF [27], is a fundamental building block for TE systems in MPLS and GMPLS networks. Path computation in large, multidomain, multiregion, or multilayer networks is highly complex, and may require special path computational components and efficient cooperation between the different network domains.
A PCE is an entity that is capable of computing paths in a network graph, applying computational constraints [38]. A PCE supports a network with a distributed control plane as well as network elements without, or with rudimentary, control plane functions. Network nodes can query the PCE for calculating a usable path they want to set up via signaling. The PCE entity can be seen as an application on a network node or component, on a separate out-of­network server. PCEs applied to GMPLS networks are able to compute paths by interfacing with one TEDB fed by a routing controller engine, and considering the bandwidth and other constraints applicable to the TE-LSP service request enhances this definition for GMPLS networks: “In GMPLS networks, PCE provides GMPLS LSP routes and optimal virtual network topology reconfiguration control, and assesses whether a new higher switching capability LSP should be established when a new LSP is triggered. PCE also handles inter-working between GMPLS and IP/MPLS networks, both of which will coexist at some
72 The Advances in Control and Management for Transport Networks
point during the migration process.” The IETF defines a region or an LSP region which refers to a switching technology domain; interface switching capabilities construct LSP regions. The deployment of a dedicated PCE can be motivated under several circumstances:
1. CPU-intensive path computation – for example, considering overall link utilization;
computation of a P2MP tree; multicriteria path computation, such as delay and link utilization; integrated multilayer path calculation tasks.
2. Partial topology knowledge – the node responsible for path computation has limited
visibility of the network topology towards the destination. This limitation may, for example, occur when an ingress router attempts to establish an LSP to a destination that lies in a separate domain, since TE information is not exchanged across the domain boundaries.
3. Absence of the TEDB – the IGPs running within parts of the network are not able to build a
full TEDB; for example, some routers in the network do not support TE extensions of the routing protocol.
4. A node is located outside the routing domain – an LSR might not be part of the routing
domain for administrative reasons.
5. A network element lacks control plane or routing capability – it is common in legacy
transport networks that network elements do not have controller. For migration purposes, the path computation can be performed by the PCE on behalf of the network element. This scenario is important for interw orking between GMPLS-capable and non-GMPLS­capable networks.
6. Backup path computation for bandwidth protection – a PCE can be used to compute
backup paths in the context of fast reroute protection of TE-LSPs.
The main driver for a PCE when it was born at the IETF was to overcome particular problems in the path computation in the inter-area environment; that is, inter-area/autonomous system (AS) optimal path computation with nodes having partial visibility of other domains only and computation of inter-area/AS diverse paths with nodes having partial visibility of other domains only. However, a PCE is seen as suitable for performing complex path computation in single-domain scenarios; for example, for MLTE concepts or in migration scenarios with non-GMPLS-capable nodes.
2.2.2.2 The Role of the Management Plane: Provisioned Connection
NMS functions manage a set of network equipment that is inventoriedin the resource inventory database. This resource inventory database can be compounded with tunable and reconfigur­able optical add–drop multiplexers (R-OADMs), transport service switches, IP/MPLS routers, or other switching network elements. These managed network elements embed controller functions that are classified in different agents, such as GMPLS agents, RSVP-TE subagents, OSPF-TE subagents and LMP subagents. These agents report the information in a repository attached to the NMS that gathers and manages the information about the network devices. The subagents report the network element management information to the NMS periodically or in a burst mode when there are changes on the network infrastructure due to upgrades with new equipment or maintenance periods.
Network management functions are composed of a set of activities, tools, and applications to enable the operation, the administration, the maintenance, and the provisioning of networked
Control Plane as Main Enabler to Autonomic Network Integration 73
systems providing connectivity services to the application users. Administration includes activities such as designing the network, tracking the usages, and planning the infrastructure upgrades. Maintenance includes diagnosis and troubleshooting functions. According to auto­matic switched transport network (ASTN) recommendations [40], connection provisioning concernsthe settings of the proper configurationof the network elements so that the connection is established byconfiguringevery network element alongthe path with the required information to establish an end-to-end connection. Connection provisioning is provided either by means of the NMS or by manual intervention. When an NMS is used, an access to the resource inventory database of the network is usually required first, to establish the most suitable route, and then to send commands to the network elements that support the connection. When a connection is provisioned by the NMS, it is usually referred to as a permanent connection. The ITU-T introduced the fault, configuration, accounting, performance, and security (FCAPS) framework. This framework for network management is part of a bigger model that is the Telecommunica­tions Management Network model of the ITU-T Recommendation series M.3000.
The management plane and control plane functions offer complementary functions, which can be used to construct an optimal provisioning control approach that is efficient, fast, and cost effective.
The functional components of the NMS are:
.
network resource management inventory
.
network resource provisioning – that is, network element configuration
.
network resource performance management
.
support for resource trouble management
.
manage network service inventory
.
network service configuration and activation.
By opposition to permanent connection, signaled connections are established on demand by communicating end points within the control plane using a dynamic protocol message exchange in the form of signaling messages (e.g., RSVP-TE protocol and messages). These messages flow across either the internal network-to-network interface (I-NNI) within the control plane instance or UNI/external network-to-network interface (E-NNI) between two distinct control plane instances. GMPLS-based connection establishment is referred to as a switched connection and it requires network naming and addressing schemes and control protocol functions.
2.2.2.3 Hybrid (Control-Management) Solutions
Architectures in production networks in the short term would essentially combine certain control plane and management plane components into a hybrid scheme. This type of connection establishment exists whereby the transport network provides a permanent connec­tion at the edge to the client network. And it utilizes a signaled connection within the network to provide end-to-end connections between the permanent connections at the network edges; that is, CE node and PE node. Within the transport network, the optical connections are established via control signaling and routing protocols. Permanent provisioning, therefore, is only required on the edge connections and, consequently, there is usually no UNI at the edge nodes. This type of network connection is known as a soft permanent connection (SPC).
74 The Advances in Control and Management for Transport Networks
IP over WDM
IP over WDM
IP over WDM
Layer 3
Layer 3
Layer 3
Layer 3
Layer 2
Layer 2
Layer 2
Layer 2
Layer 1
Layer 1
Layer 0
Layer 0
Layer 0
Layer 0
Figure 2.2 Two examples of multilayer network architecture models
SDL
SDL
SDL
Packet over SDH
Packet over SDH
Packet over SDH
Packet over SDH
MPLS
MPLS
MPLS
MPLS
HDLC
HDLC
HDLC
HDLC
Frame Relay
Frame Relay
Frame Relay
Frame Relay
SONET/SDH
SONET/SDH
SONET/SDH
SONET/SDH
PPP
PPP
PPP
PPP
IP
IP
IP
IP
WDM
WDM
WDM
WDM
ATM
ATM
ATM
ATM
IP over WDM
G.709
G.709
G.709
G.709
ETHERNET
ETHERNET
ETHERNET
ETHERNET
FIBER
FIBER
FIBER
FIBER
CHANNEL
CHANNEL
CHANNEL
CHANNEL
From the perspective of the end points, an SPC appears no different than a management­controlled connection; that is, a permanent connection.

2.3 Multilayer Interactions and Network Models

This section develops on the different options of network control plane configurations over a multilayer transport network from the interactions between several control networks to the integration of one uniformed control network instance.

2.3.1 Introduction

Operator networks have been upgraded with different network technologies, each providing its own set of functionalities, defining its own switching capability and framing. Enabled with the GMPLS protocol architecture, the functional interactions between the different network layers can become more uniform and allow a reduction in the complexity of the global control and management of the network. With GMPLS, the data network (typically L3 or L3 networks) and transport network (typically OTNs) layer convergence can be addressed by providing the end-to-end LSPs (i.e., GMPLS-controlled connections) integrating the requirements expressed by the users of the connections.
The network service layer introduced in Section 2.1 can be referenced with the OSI communications model as illustrated in Figure 2.2. Each network service layer has the property that it only uses the connectivity service offered by the server layer (i.e., the layer below) and only exports functionality to the layer above. The different network service layers and their interaction models express how the transport network layer server should interact with the client network layer to establish connections (permanent connections, soft-permanent connections, or signaled connections [40]) in support of the user-network services.
Two classic ways of transporting IP traffic by optical networks are packet over SONET/SDH (PoS) and IP over WDM. Within a GMPLS-controlled network, each network layer carry another network layer i as it corresponds to the concept of nesting LSPs; that is, LSPs
2
A network layer, also referred as a “switching layer,” is defined as a set of data links with interfaces that havethe same switching and data encoding types, and switching bandwidth granularity. Examples of network layers are SDH VC-4, SONET STS-3, G.709 OTU-1, Ethernet, IP, ATM VP, and so on.
2
i 1 can
Multilayer Interactions and Network Models 75
originated by other LSRs at layer i into that LSP at layer i 1 using the label stack construct, defined by the LSP hierarchy [34]. For example for i ¼2, L2-LSPs (Ethernet) can be carried by TDM LSPs or G.709 LSPs.
Nesting can occur between different network layers within the same TE domain, implying
interfaces switching capability information to be controlled in a hierarchical manner as shown in Figure 2.2. With respect to the switching capability hierarchy, layer i is a lower order switching capability (e.g., Ethernet) and layer i 1 is the higher order switching capability (e.g., SONET/SDH).
Nesting of LSPs can also occur at the same network layer. For example at the layer 1, a lower order SONET/SDH LSP (e.g., VT2/VC-12) can be nested in a higher order SONET/SDH LSP (e.g., STS-3c/VC-4). In the SONET/SDH multiplexing hierarchy, several levels of signal (LSP) nestings are defined.
2.3.1.1 Overlay Model
The overlay model refers to telecom carriers or optical backbone (bandwidth) providers who lease their network infrastructure facilities to Internet service providers (ISPs). This model is based on a well-defined client–server relationship with controlled network interfaces (or reference points) between the provider networks and customer networks involved. The overlay model mandates a complete separation of the client network control (e.g., based on MPLS architecture [27]) and the transport network control plan e (e.g., based on GMPLS [42]). Only a controlled amount of signaling messages may be exchanged. As a consequence, the overlay model is very opaque. The client network routing and signaling controllers are independent of the routing and signaling controllers within the transport network domain. The two independent control planes interact through a UNI [39], defining a client–server relationship between the customer network and the transpor t network.
2.3.1.2 Peer Model
Compared with the overlay model, the peer model does not restrict any control routing information exchanged between the network switching layers. This model is relevant and can be optimal when a carrier network is both a network infrastructure provider (NIP) and an ISP. In this case, the provider networks can align the topological design of their transport network with the service operations of their data network, but they might be in conflict with some national or international policies.
3
The client network control plane acts as a peer of the GMPLS transport network control plane, implying that a dual instance of the control plane is running over the data network (e.g., IP/MPLS) and optical network (e.g., SDH/GMPLS), as illustrated in Figure 2.3. The peer model entails the tightest coupling between the customer networks and the transport network. The different nodes (CE, PE or P) can be distinguished by their switching capabilities; for example, PSC for IP routers interconnected to GMPLS PXCs.
3
The European Commission, in its Green Paper, has regulated that a formal split be made within telecoms in network operating departments and service provider departments with a formal supplier relationship that is equal to an external vendor/buyer relationship. This relationship must be nondiscriminatory. Similarly, requirements of the FCC Ruling on Interconnection in the USA are encouraging companies to formally separate their network provider and service provider business.
76 The Advances in Control and Management for Transport Networks
Figure 2.3 Network models and reference points [39]
2.3.1.3 Integrated Model
Compared with the peer model, the integrated model does not require different control plane interfaces between the network layers and different TE domains. The integrated model proposes one single instance of a control network for the customer networks and the provider network. All the nodes are LSRs and they are not classified in different network domains due to the administration they belong to or their interface switching capabilities. Each LSR is an integrated platform able to handle several orders of switching capabilities: for example, IP, Ethernet, TDM, and wavelength. An LSR embeds one GMPLS control plane instance and is able to control different switching-capability interfaces simultaneously. On the one hand, only this model can handle a global optimization of the network resource usages through the network; for example, packet, layer 2 (Ethernet, ATM, etc.), TDM (SONET, SDH, etc.), lambda (G.709), and fiber switching capabilities; on the other hand, this model has to face the scalability challenges to integrate the control of TE-links belonging to different switching capabilities and to control their states in a very reactive manner.
2.3.1.4 User-to-Network Interface
The UNI is a logical network interface (i.e., reference point) introduced in the requirements for the ASTN specification [40] and recommended in [39]. The UNI defines the set of signaling messages that can be exchanged between a node controller of the client network and a server node of the transport network. The server node provides a connection service to the client node; for example, the IP router can signal TDM LSPs on its PoS interf aces. The UNI supports the exchange of authentication and connection admission control messages and provides to CE
Multilayer Interactions and Network Models 77
nodes the address space set of the reachable nod es. The first implementation agreement for a UNI was produced by the Optical Internetworking Forum (OIF) in October 2001 [57]. This OIF implementation agreement is for an overlay model. The signaling messages exchanged between each client node and server node are restricted to LSP connection request/tear down only. The IETF specifies a GMPLS UNI that is applicable for a peer model [50]. Fully compliant with RSVP-TE, the GMPLS UNI allows the end-to-end LSP handling along LSR signaling paths. Some recent contributions at OIF integrate the alignments of RSVP-TE within OIF UNI 2.0 [58].
2.3.1.5 Network-to-Network Interface
The network-to-network interface (NNI) is a logical network interface (i.e., reference point) recommended in the “Requirements for Automatic Switched TransportNetwork” specification ITU-T G.807/Y.1302. The NNI defines the set of both signaling messages and routing messages that can be exchanged between two server nodes; for example, between two GMPLS­controlled PXCs. There are two types of NNI: one for two different TE domains and one for intra-domain TE: i E-NNI and ii I-NNI respectively.
.
An E-NNI assumes an untrusted relationsh ip between the two network domains. The information exchanged between the two nodes located at the edge of the transport network specified within the E-NNI is restricted. The control messages exchanged include the reachability network addresses that are usually summa rized, authentication and connection admission control messages, and a set restricted to connection requests of signaling messages. Some contributions have been initiated at the OIF conce rning the signaling message exchanges [59], but the project plan for producing an OIF E-NNI Routing 2.0 implementation agreement [60], which started in November 2007, is without significant progress despite some solid contributions from NOBEL2.
.
An I-NNI assumes a trusted relationship between two network domains and is usually implemented in the same TE domain or administrative domain. The control information specified within the I-NNI is not restricted. The routing control messages exchanged include connection service topology, LSA (from IGP), and node discovery. The signaling messages can allow controlling of the resources end to end for the two networks of each LSP and its protection.
Network integration usually removes the restrictions imposed by administrative network domains. It covers interworking capabiliti es and optimized control functions for multiple switching layers running GMPLS with unified signaling and routing approaches for connection (i.e., LSP) provisioning and recovery. Integrated network architectures can include (i) network equipment hosting multiple switching capabilities that are controlled by a single instance of the GMPLS control plane and (ii) seamless collabo ration between network domains (e.g., routing areas, autonomous systems [28]) of the network.
Hence, the multilayer interaction network models have been categorized into a horizontal integration and vertical integration framework (Figure 2.4):
.
Vertical integration Collaborative switching layers are controlled by a single control plane instance: the UNI reference point is coalescing with the I-NNI control interface.
78 The Advances in Control and Management for Transport Networks
Figure 2.4 Horizontal and vertical integration of network domains with reference to the GMPLS
control plane instances (not shown)
.
Horizontal integration Each administrative entity constituting a network domain is controlled by one single control plane instance usually inferring one common (data plane) switching capability and the control plane topology extends over several routing areas or autonomous systems. This horizontal integration includes control functions such as: (i) signaling and routing in multi-domain transparent networks, (ii) interface adaptation between MPLS and GMPLS control domains, and (iii) proposed multi-domain signaling protocol extensions for the end-to-end service.

2.3.2 Vertical Integration and Models

The GMPLS control network is mor e complex when more than one network layer is involved. In order to enable two network layers to be integrated, interface adaptation capabilities are required.
2.3.2.1 Towards the IP/Ethernet Convergence
The Ethernet MAC is one of the main (if not the only) layer 2 technology that is going to remain really attractive in the long run. Together with pure layer 3 IPv4/IPv6 packet forwarding, these are the two fundamental data plane building blocks of any future access, metro, and backbone network. Starting from local area network (LAN) environments, Ethernet is under deployment in metropolitan area networks (MANs), and its extensions to metro–core and core network environments are foreseen.
2.3.2.2 Integrating Several Transport Layers
The role, limitations, and strengths of synchronous networking (SONET/SDH) and its association with WDM. How statistical multiplexing and framing can be further enhanced during the migration process to “packetized” and “flatter” multilayer networks with optimized costs of transpor t equipment.
g
Multilayer Interactions and Network Models 79

2.3.3 Horizontal Integration and Models

When a single instance of a network controller covers several routing areas or autonomous systems, each having its own TE domain, the model is said to be horizontally integrated. Usually, horizontal integration is deployed within a single network layer spanning several administrative network domains [53], whereas vertically integrated models involve multiple network layers.
2.3.3.1 Multi-Domain Interoperability
In many of today’s complex networks, it is impossible to engineer end-to-end efficiencies in a multi-domain environment, provision services quickly, or provide services based on real-time traffic patterns without the ability to manage the interactions between the IP-layer functionality of PSNs and that of the optical transmission layer. According to proponents of automatically switched optical network (ASON)/GMPLS, an optical control plane is the most advanced and far-reaching means of controlling these interactions.
The multi-domain issue and evolution in interconnecting network domains.
Currently, GMPLS is evolving to cover the end-to-end service establishment in multi­domain configuration.
A control plane consists inherently of different control interfaces, one of which is concerned with the reference point called the UNI. From a historical point of view, each transport technology has defined UNIs and associated control mechanisms to be able to automatical ly establish connections. NOBEL WP4 has examined the currently available standardized UNI definitions, and points out the differences between them as illustrated in Figure 2.5.
The UNIs most applicable for ASON- and GMPLS-controlled networks are being actively worked on by the IETF, OIF, and ITU-T. They show a significant protocol reuse for routing and
optical connection service
Transport Network
UNI-N
Source UNI-C
UNI-N
UNI-C
OIF UNI
GMPLS UNI
OIF UNI (Overlay Network)
Path message triggered w/o EXPLICIT_ROUTE:
UNI Session A Tunnel add.: source UNI-N node ID If RSVP-TE: Session Tunnel add.: dest. UNI-N node ID
UNI Session B Tunnel address: dest. UNI-C node ID
Source OXC UNI-N ERO/RRO Processing
ERO: all OXC hops up-to destination UNI-N (strict)
Source UNI-N computes the path to reach dest. UNI-N and
creates an ERO in the outgoing Path messages
RSVP
Session A
RSVP Session A
Figure 2.5 Network models compared in NOBEL: OIF UNI versus GMPLS UNI
Client Network
Destination UNI-C
UNI-N
UNI-N
UNI
UNI
UNI-C
RSVP
Session B
GMPLS UNI (Augmented Network)
Path message triggered with EXPLICIT_ ROUTE:
Session Tunnel address: destination UNI-C node ID
Source LSR UNI-C ERO/RRO Processing:
ERO: ingress UNI-N (strict) and dest. UNI-C (loose)
Source UNI-N computes path to reach dest. UNI-C, updates
the ERO and includes it in the out
oing Path messages
80 The Advances in Control and Management for Transport Networks
signaling in the different standards bodies. Each standards body has initial ly used identical fundamental protocols, which have been further developed and modified to fulfill the standards body’s requirements, including the European carrier network requirements (i.e., telecom network operators). Thus, the initially fundamental protocols have all evolved in slightly different directions, as at the time there were no significant established liaisons requesting cooperation between them.
The differences were driven by user requirements and are significant enough to lead to incompatibility between the UNIs recommended by different standards bodies. Although the IETF and ITU-T/OIF are both based on RSVP-TE, the requirements are different.
This section reports on the analysis and comparison of the currently available UNI definitions of the IETF, OIF, and ITU-T at the time of the NOBEL project analysis. Furthermore, this section provides a fully GMPLS-compliant UNI definition, and the corre­sponding technical description, for the support of end-to-end signaling for fast connection provisioning with different options for connection recovery. This definition, called GMPLS UNI, is one of the contributions of NOBEL to achieve convergence towards future high­performance networks providing high-impact, end-to-end services for optical networks (TDM, wavelength or port switching).
A UNI is a referenc e point over which predefined (i.e. standardized ) messages are exchanged with the objective of requesting services from a network. An UNI reference point is logically associated with a physical interface on the interdomain link between the user network and the provider network. The messages exchanged between the user and the provider require a physic al bearer. The physical network infrastructure designed for that purpose – that is, to relay control and network management information – is called a data communication network [39].
The following information and procedures are typically managed over UNI:
.
endpoint name/address registration – directory service [57]
.
authentication, authorization, and connection admission control (CAC)
.
connection service messages
.
connection request/release þconfirmation
.
resource discovery
.
grade of service (GoS) parameters selection (which are typically captured in service level
4
agreements (SLA)s)
.
handling of error and exception conditions during the recovery process.
.
non-service affecting attribute modification
.
query of attribute values for a given circuit.
There is typically no routing protocol information; that is, link state flooding [28] exchanged over the UNI. This is fundamentally different from other interfaces, such as I-NNI and E-NNI. However, under certain policies, the CE node can request and maintain optical connection services by signaling explicitly an explicit route with, for example, an explicit route object [29,48 ,59].
4
A user’s connection request towards a carrier network can be specified with a specific time of the day and/or day of the week. The -N controller agent verifies whether connection requests are received during the contracted hours. All information required to make the policy decision (time-of-request receipt) is contained in the UNI signaling message.
Multilayer Interactions and Network Models 81
The concept of “user” in this context is considered to be the connection requestor, that is, the entity that requests to utilize from the network particular network infrastructural resources, with the typical aim to exchange information with one or more remote entities. The UNI can be single-ended end-users or corporate LAN interconnections, metro network c onnections, and so on. From the description of “user,” a multitude of scenarios can be envisioned under which the UNI needs to operate. This obviously puts significant pressure on the UNI, as it needs to be able to operate under a broad spectrum of different applications areas, requiring the UNI to be able to support all the different types of connection and physical infrastructures. It equally shows the importance of the UNI functionality evolvingtowards the user requesting automatically the connectivity services offered by the dynamic optical network infrastructure enabled by GMPLS control functions.
There is mainly one view on the recommendations of the UNI that views the UNI as a client–server relation The following nonexclusive list presents some points:
.
there is a client–server relation between the connectivity service user and the NIP;
.
routing information is not exchanged between the parties involved and it is based on an overlay network interaction model;
.
there is no trust relation between the two parties involved, but there is a (commercial) agreement;
.
there is a business relationship between an end-user and a provider;
.
this business relationship is based on an SLA (i.e., a commercial contract);
.
this business relationship is typically transport technology dependent.
In this document we focus solely on the control and management functions and specific extensions required for the UNI client–server relationship.
2.3.3.2 The User-to-Network Interface
The UNI is an asymmetrical interface and is divided into two parts: UNI-C on the client side and UNI-N on the network side, as shown in the Figure 2.5.
The different functional requirements on the UNI client and network sides necessitate the two distinct parts constituting the UNI,; that is, the client-specific part and network-specific part. On UNI-N, an additional routing interface for connection control has, for example, to be provided [39]; that is, the UNI-N has a routing controller that participates in the network­related routing. However, the UNI-N does not distribute that routing information over the UNI towards the UNI-C.
The UNI supports, as a minimum, the following information elements or functions to trigger SPCs or switched connections [39,40]:
.
authentication and admission control
.
endpoint name and addressing
.
connection service messages.
The control plane functions of the UNI-C side are mainly call control and resource discovery. Only limited connection control and connection selection are necessary at UNI-C. The following functions reside at the UNI-N side [39]:
82 The Advances in Control and Management for Transport Networks
.
call control
.
call admission cont rol
.
connection control
.
connection admission control
.
resource discovery and connection selection.
The optical network layer provides transport services to interconnect clients such as IP routers, MPLS LSRs, carrier-grade Ethernet switches, SDH cross-connects, R­OADMs, and so on. In its initial form, the OTN uses SONET/SDH as the interface switching-apable interfaces, and the network is migrating to Ethernet framing clients in the future.
The OIF defines the UNI as the ASON control interface between the transport network and client equipment. Signaling over the UNI is used to invoke connectivity services that the transport networ k offers to clients.
The purpose o f the OIF UNI is to define int eroperable procedures for requesting, configuring, and signaling dynamic connectivity between network equipment clients (e.g., Ethernet switches or IP routers) connected to the transport network. The develop­ment of such procedures requires the definition of a logical interfaces between clients and the transport network, the connectivity services (specified as a call in Ref. [39]) offered by the transport network, the signaling protocols used to invoke the services, the mechanisms used to transport signaling messages, and the autodiscovery procedures that aid signaling.
We have the following definitions:
.
Client/user: network equipment that is connected to the transport network for utilizing optical connection.
.
Transport network: an abstract representation, which is defined by a set of APs (ingress/ egress) and a set of network services.
.
Connection: a circuit connecting an ingress transport network element (TNE) port and an egress TNE port across the transpor t network for transporting user signals. The connection may be unidirectional or bidirectional [61].
.
UNI: The logical control interface between a client device and the transport network.
.
UNI-C: The logical entity that terminates UNI signaling on the client network device side.
.
UNI-N: The logical entity that terminates UNI signaling on the transport network side.
2.3.3.3 The OIF or Public UNI
The OIF UN I Connection Services
The primary service offered by the transport network over the UNI is the ability to create and delete connections on demand. A connection is a fixed-bandwidth circuit between ingress and egress APs (i.e., ports) in the transport network, with specified framing [61]. The connection can be either unidirectional or bidirectional. Under OIF UNI 1.0, this definition is restricted to being a TDM connection of payload bandwidth 50.112 Mbit/s (e.g., SONET STS-1 or SDH VC-3) and higher.
Multilayer Interactions and Network Models 83
The properties of the connection are defined by the attributes specified during connection establishment. Four activities are supported across the UNI, as listed below and illustrated with RSVP-TE [29,48]:
.
connection establishment (signaling) – Path/Resv exchange sequences;
.
connection deletion (signaling) – PathTear/ResvTear exchange sequences;
.
connection notification (signaling) – PathErr/ResvErr/Notify exchange sequences;
.
status exchange (signalin g) – discovery of connection status;
.
autodiscovery (signaling) – discovery of connectivity between a client, the network, and the services available from the network.
Actual traffic (usage of the established connections) takes place in the data plane, not over the service control interface.
For each activity there is a client and a server role.
The OIF UN I Signaling Sequences
UNI signaling refers to the message exchange between a UNI-C and a UNI-N entity to invoke transport network services. Under UNI 1.0 signaling, the following actions may be invoked:
1. Connection creation: This action allows a connection with the specified attributes to be
created between a pair of APs. Connection creation may be subject to network-de fined policies (e.g., user group connectivity restrictions) and security procedures.
2. Connection deletion: This action allows an existing connection to be deleted.
3. Connection status enquiry: This action allows the status of certain parameters of the
connection to be queried.
OIF UNI Supporting Procedures
UNI Neighbor Discovery (Optional)
The neighbor discovery procedure is fundamental for dynamically establishing the interface mapping between a client and a TNE. It aids in verifying local port connectivity between the TNE and the client devices. It also allows the UNI signaling control channel to be brought up and maintained.
Service Discovery (Optional)
Service discovery is the process by which a client device obtains information about the available connectivity from the transport network, and the transport network obtains informa­tion about the client UNI signaling (i.e., UNI-C) and port capabilities.
Signaling Control Channel Maintenance
UNI signaling requires a control channel between the client-side and the network-side signaling entities. Different control channel configurations are possible, as defined in the OIF UNI specification [57]. OIF UNI supports procedures for maintenance of the control channel under all these configurations.
There are two service invocation models, one called direct invocation and the other called indirect invocation. Under both models, the client-side and network-side UNI signaling agents are referred to as UNI-C and UNI-N respectively. In the direct invocation model,
84 The Advances in Control and Management for Transport Networks
the UNI-C functionality is present in the client itself. In the indirect invocation model, an entity called the proxy UNI-C performs UNI functions on behalf of one or more clients. The clients are not required to be collocated with the proxy UNI-C.
A control channel is required between the UNI-C and the UNI-N to transport signaling
messages. The OIF UNI specification supports an in-fiber signaling transport configuration, where the signaling messages are carried over a communication channel embedded in the data-carrying optical link between the client and the TNE. This type of signaling applies only to the direct service invocation. An out-of-fiber signaling transport configuration is also supported, where the signaling messages are carr ied over a dedicate d communication link between the UNI-C and the U NI-N, separate from the data-bearing optical links. This type of signaling applies to the direct service invocation model as well as the indirect service invocation model.
Discovery Functions (Optional)
The neighbor discovery procedures allow TNEs and directly attached client devices to determine the identities of each other and the identities of remote ports to which their local ports are connected. The IP control channel (IPCC) maintenance procedures allow TNEs and clients to continuously monitor and maintain the list of available IPCCs.
The protocol mechanisms are based on the LMP.
Service discovery in OIF UNI is optional, and it can be based on OIF-specific LMP extensions [57]. Service discovery is the procedure by which a UNI-C indicates the client device capabilities it represents to the network, and obtains information concerning transport network services from the UNI-N; that is, the signaling protocols used and UNI versions supported, client port-level service attributes, transparency service support, and network routing diversity support.
OIF UNI Extensions [58]
The primary service offered by the transport network over the UNI is the ability to trigger the creation, the deletion, and the modification of optical connections on demand. In the context of the NOBEL project, a connection is a fixed-bandwidth circuit between ingress and egress APs (i.e., ports) in the transport network, with specified framing. The connection can be either unidirectional or bidirectional. Under UNI 2.0, this connection can be a SONET service of bandwidth VT1.5 and higher, or an SDH service of bandwidth VC-11 and higher,an Ethernet service, or a G.709 service. The properties of the connection are defined by the attributes specified during connection establishment.
The following features are added in OIF UNI 2.0 [58]:
.
separation of call and connection controllers as recommende d in [39]
.
dual homing for diverse network infrastructure provider routing
.
nondisruptive connection modification through rerouting traffic-engineered tunnel LSPs; that is, implementing make-before break [29]
.
1:N signaled protection (N 1) through segment LSP protection [8] or end-to-end LSP protection [7]
.
Sub-TDM signal rate connections (SONET/STS-1, SDH/VC-12, SDH/VC-3, etc.)
.
transport of carrier-grade Ethernet services [22,26]
.
transport of wavelength connection services as recommended with G.709 interfaces
.
enhanced security.
Multilayer Interactions and Network Models 85
UNI Abstract Messages
This section describes the different signaling abstract messages. They are termed “abstract” since the actual realization depends on the signaling protocol used. OIF UNI describes LDP and RSVP-TE signaling messages corresponding to the abstract messages that can be exchanged between a pair of CE nodes implementing the UNI-C function and the PE node implementing the UNI-N functions. Abstract messages comprise connection create request (Path), connec­tion create response (Resv), connection create confirmation (ResvConf), downstream connec­tion delete request (PathTear), upstream connection delete request (ResvTear), connection status enquiry, connection status response, and notification (Notify).
The attributes are classified into identification-related, signaling-related, routing-related, policy-related, and miscellaneous. The encoding of these attributes would depend on the signaling protocol used.
2.3.3.4 The IETF GMPLS UNI
The following section describes the UN I from the IETF point of view, termed a private UNI [2]. In Ref. [62] the IETF describes the signaling messages exchanges that can be configured between IP network clients of optical network servers. In the NOBEL project, the network model considered consists of IP routers
5
attached to a set of optical core networks and connected to their peers over dynamically signaled optical channels. In this environment, an optical sub-network may consist entirely of transparent PXCs or OXCs with optical– electrical–optical (OEO) conversions. The core network itself is composed with optical nodes incapable of switching at the granularity of individual IP packets.
With reference to [62], three logical control interfaces are differentiated by the type and the possible control of information exchanges: the client–optical internetwork interface (UNI), the internal node-to- node interface within an optical network domain (I-NNI), and the E-NNI between two network domains. The UNI typically represents a connection service boundary between the client packet LSRs and the OXC network [15]. The distinction between the I-NNI and the E-NNI is that the former is an interface within a given network under a single administration (e.g., one single carrier network company), while the latter indicates an interface at the administrative boundary between two carrier networks. The I-NNI is typically configured between two sets of network equipments within the same routing area or autono­mous system. The I-NNI and E-NNI may thus differ in the policies that restrict routing information flow between nodes. Ideally, the E-NNI and I-NNI will both be standardized and vendor-independent interfaces. However, standardization efforts have so far concentrated on the E-NNI [59,60]. The degree to which the I-NNI will become the subject for standardization is yet to be defined within a roadmap.
The client and server parts of the UNI are essentially two different roles: the client role (UNI-C) requests a service connection from a server; the server role (UNI-N) can trigger the establishment of new optical connections to fulfill the QoS parameters of the connection request, and assures that all relevant admission control conditions are satisfied. The signaling messages across the UNI are dependent on the set of connection services defined across it and the manner in which the connection services may be accessed.
5
The routers that have direct physical connectivity with the optical network are referred to as “edge routers” with
respect to the optical network.
86 The Advances in Control and Management for Transport Networks
The service available at this interface can be restricted, depending on the public/private configuration of the UNI. The UNI can be categorized as public or private, depending upon context and service models. Routing information (i.e., topology and link state information) can be exchanged across a private UNI. On the other hand, such information is not exchanged across a public UNI, or such information may be exchanged with a very explicit routing engine configuration.
Connection Service Signaling Models
Two service-models are currently defined at the IETF, namely the unified service model (vertically) and the domain services model (horizontally). Under the unified model, the IP and optical networks are treated together as a single integrated network from a routing domain point of view. In principle, there is no distinction between the UNI, NNI, and any other control interfaces.
The optical domain services model does not deal with the type and nature of routing protocols within and across optical networks. An end-system (i.e., UNI-C and UNI-N) discovery procedure may be used over the UNI to verify local port connectivity between the optical and client devices, and allows each device to bootstrap the UNI control channel. This model supports the establishment of a wavelength connection between routers at the edge of the optical network. The resulting overlay model for IP over optical networks is discussed later. Under the domain service model, the permitted services through the UNI are as follows:
.
lightpath creation
.
lightpath deletion
.
lightpath modification
.
lightpath status enquiry
.
service discovery, restricted between UNI-C and UNI-N.
Routing Approaches
Introduction
The following routing approaches are closely related to the definition of the UNI and the interconnection models considered (overlay, augmented, peer).
Under the peer model, the IP control plane acts as a peer of the OTN control plane (single instance).
Under the overlay model, the client layer routing, topology distribution, and signaling protocols are independent of the routing, topology distribution, and signaling protocols within the optical domain. The two distinct control planes interact through a user network interface defining a separated client–server relationship. As a consequence, this model is the most opaque, offers less flexibility, and requires specific rules for multilayer routing.
Finally,under the augmented model, there are separate routing instances in the IP and optical domains, but certain types of information from one routing instance can be passed through to the other routing instance.
Integrated Routing (GMPLS)
This routing approach supports the peer model with the control from a single administrative domain. Under the integrated routing, the IP and optical networks are assumed to run the same
Multilayer Interactions and Network Models 87
instance of an IGP routing protocol (e.g., OSPF-TE) with suitable TE extensions for the “optical networks” and for the “IP networks.” These TE extensions must capture optical link parameters and any routing constraints that are specific to optical networks. The virtual topology and link state information stored in the TEDB and maintained by all nodes’ (OXCs and routers) routing engines may be identical , but not necessarily. This approach permits a router to compute an end-to-end path to another router considering the link sta tes of the optical network.
The selection of the resources in all layers can be optimized as a whole, in a coordinated manner (i.e.,, taking all layers into account). For example, the number of wavelength LSPs carrying packet LSPs can be minimized.
It can be routed wavelength LSPs that provide a virtual topology to the IP network client without reserving their bandwidth in the absence of traffic at the IP network client, since this bandwidth could be used for other traffic.
Domain-Specific Routing
The domain-specific routing approach supports the augmented interconnection model. Under this approach, the routing processes within the optical and IP domains are separated, with a standard border gateway routing protocol running between domains. IP inter-domain routing based on the border gateway protocol (BGP) is usually the reference model.
Overlay Routing
The overlay routing approach supports the overlay interconnection model. Under this approach, an overlay mechanism that allows edge routers to register and query for external addresses is implemented. This is conceptually similar to the address resolution mechanism used for IP over ATM. Under this approach, the optical network could implement a registry that allows edge routers to register IP addresses and VPN identifiers. An edge router may be allowed to query for external addresses belonging to the same set of VPNs that it belongs to. A successful query would return the address of the egress optical port through which the external destination can be reached.
IETF GMPLS UNI Functionality [50]
Routing information exchange may be enabled at the UNI level, according to the different routing approaches above. This would constitute a significant evolution: even if the routing instances are kept separate and independent, it would still be possible to dynamically exchange reachability and other types of routing information.
6
Addressing
The IETF proposes two addressing schemes. The following policies are relevant:
.
In an overlay or augmented model, an end client (edge node) is identified by either a single IP address representing its node-ID, or by one or more numbered TE links that connect the client and the core node.
.
In the peer model, a common addressing scheme is used for the optical and client networks.
6
Another more sophisticated step would be to introduce dynamic routing at the E-NNI level. This means that any neighboring networks (independent of internal switching capability) would be capable of exchanging routing information with peers across the E-NNI.
88 The Advances in Control and Management for Transport Networks
Table 2.3 IETF models and approaches
Signaling control Interconnection model Routing control UNI functionality
Uniform end-to-end Peer Integrated (I-NNI) Signaling þ common link
state database (TEDB)
Overlay (edge-to-edge) Augmented Separated (E-NNI) Signaling þ inter-layer routing
Overlay No routing Signaling
Signaling through the UNI
The IETF proposes to use standard GMPLS signaling for the UNI, which can be configured for a client–server model in the case of a per domain signaling model and for end-to-end integrated signaling in the case of a unified service model. A comparison of the different UNI signaling features is shown in Tables 2.3 and 2.4.
Overlay Service Model
The signaling for UNI considers a client–server relationship between the client and the optical network. Usually the switching capability of the client network is lower than the transport network. The source/destination client addresses are routable, and the identifier of the session is edge-to-edge significant. In principle, this implies several signaling sessions used throughout the UNI, I-NNI, and E-NNI that are involved in the connection.
The starting point for the IETF overlay model (IETF GMPLS UNI) is the use of the GMPLS RSVP-TE protocol specified in Ref. [10]. Based on that protocol, the draft GMPLS specifies mechanisms for UNI signaling that are fully compliant with the signaling specified in Refs. [10,48]. There is a single end-to-end RSVP session for the user connection. The first and last hops constitute the UNI, and the RSVP session carries the LSP parameters end to end.
Furthermore, the extensions described in GMPLS address the OIF UNI shortcomings and provide capabilities that are required in support of multilayer recovery.
Unified Service Model
In this model, the IP and optical networks are treated together as a single integrated network from a control plane point of view. In principle, there is no distinction between the
Table 2.4 Comparison between the OIF UNI and the IETF UNI
UNI and Service Model OIF UNI IETF UNI (overlay) IETF UNI (unified)
Signaling Direct and indirect Direct Direct Symmetry/scope Asymmetrical/local Asymmetrical/
edge-to-edge Routing protocol None None/optional Link state preferred Routing information None UNI-N may reveal
reachability
based on policy Address space Must be distinct Can be common
in part Discovery Optional Optional Through routing Security No trust Limited trust High trust
Symmetrical/end-to-end
Reachability
(augmented) and TE attributes
Common
Loading...