An Introduction to

ATM Networks

by

Harry Perros

To

Helen, Nick, and Mikey Foreword

ATM networks was the subject of intense research and development from the late 1980s
to the late 1990s. Currently, ATM is a mature networking technology and it is taught
regularly in Universities and in short professional courses. This book was written with a
view to be used as a text book in a second course on computer networks at the graduate
level or senior undergraduate level. Also, it was written for networking engineers out in
the field who would like to learn more about ATM networks. A pre-requisite for this
book is basic knowledge of computer networking principles.
The book is organized into the following four parts:


architectures, and the following three different classes of ATM switch architectures are
presented: space-division switches, shared memory switches, and shared medium
switches. We describe various architectures that have been proposed within each of these
three classes. Also, to give the reader a feel of a real-life switch, the architecture of a
commercial switch is described. We conclude this Chapter by describing various
algorithms for scheduling the transmission of cells out of an output port of an ATM
switch. Finally, Chapter 7 deals with the interesting problem of congestion control in
ATM networks. We first present the various parameters used to characterize ATM traffic,
the various quality of service (QoS) parameters, and the standardized ATM classes. In the
rest of the Chapter, we focus on the two classes of congestion control schemes, namely,
the preventive and reactive congestion control. We introduce the preventive congestion
control scheme, and we present various call admission control algorithms, the GCRA
bandwidth enforcement algorithm, and cell discard policies. Finally, we present the
available bit rate (ABR) scheme, a reactive congestion control scheme standardized by
the ATM Forum.
Part Three “Deployment of ATM”, deals with the two different topics, namely,
how IP traffic is transported over ATM, and ADSL-based access networks. It consists of
Chapters 8 and 9. In Chapter 8 we describe various schemes used to transport IP traffic
over ATM. We first present ATM Forum’s LAN emulation (LE), a solution that enables
existing LAN applications to run over an ATM network. Then, we describe IETF’s
schemes classical IP and ARP over ATM and next hop routing protocol (NHRP)
designed for carrying IP packets over ATM. The remaining of the Chapter is dedicated to
the three techniques IP switching, tag switching, and multi-protocol label switching
(MPLS). IP switching inspired the development of tag switching, which at this moment is
Forward

x
being standardized by IETF under the name of multi-protocol label switching. Chapter 9
is dedicated to the asynchronous digital subscriber line (ADSL) technology which can be
used in residential access networks to provide basic telephone services and access to the

Harry Perros
Cary, February 13th, 2001
Contents PART ONE: INTRODUCTION AND BACKGROUND

1. Introduction 3
1.1 The Asynchronous Transfer Mode (ATM) 3
1.2 Standards committees 5
Problems 11

2. Basic Concepts From Computer Networks 13
2.1 Communication networking techniques 13
2.2 The Open System Interconnection (OSI) Reference Model 16
2.3 Data link layer 17
2.4 The high data link control (HDLC) protocol 22
2.5 Synchronous time division multiplexing (TDM) 24
2.6 The logical link control (LLC) layer 27
2.7 Network access protocol X.25 29
2.8 The internet protocol (IP) 32
2.8.1 The IP header 32
2.8.2 IP addresses 34
2.8.3 ARP, RARO and ICMP 37
2.8.4 IP version 6 (IPv6) 38
Problems 38


4.6 UTOPIA and WIRE 78
Problems 79

5. The ATM Adaptation Layer 81
5.1 Introduction 81
5.2 ATM Adaptation Layer 1 (AAL 1) 84
5.2.1 The AAL 1 SAR sublayer 84
5.2.2 The AAL 1 CS sublayer 85
5.3 ATM Adaptation Layer 2 (AAL 2) 88
5.4 ATM Adaptation Layer 3/4 (AAL 3/4) 92
5.5 ATM Adaptation Layer 5 (AAL 5) 95
Contents
xv
Problems 97

6. ATM Switch Architectures 99
6.1 Introduction 99
6.2 Space-division switch architectures 102
6.2.1 The cross-bar switch 102
6.2.2 Banyan networks 105
6.2.3 Clos networks 113
6.2.4 Switch architectures with N
2
disjoint paths 114
6.3 Shared memory ATM switch architectures 115
6.4 Shared medium ATM switch architectures 118
6.5 Non-blocking switches with output buffering 120
6.6 Multicasting in an ATM switch 121
6.7 Scheduling algorithms 123
6.8 The Lucent AC120 Switch 126


8. Transporting IP Traffic Over ATM 177
8.1 Introduction 177
8.2 LAN emulation 179
8.3 Classical IP and ARP over ATM 183
8.3.1 ATMARP 184
8.3.2 IP multicasting over ATM 187
8.4 Next hop routing protocol (NHRP) 191
8.5 IP switching 194
8.6 Tag switching 198
8.7 Multi-protocol label switching (MPLS) 206
Problems 208

9. ADSL-Based Access Networks 211
9.1 Introduction 211
9.2 The ADSL technology 215
9.2.1 The discrete multi-tone (DMT) technique 217
9.2.2 Bearer channels 219
9.2.3 The ADSL super frame 220
9.3 Schemes for accessing network service providers 221
9.3.1 The L2TP access aggregation scheme 222
Contents
xvii
9.3.2 The PPP terminated aggregation scheme 224
Problems 224

PART FOUR: SIGNALLING IN ATM NETWORKS

10. Signalling Over the UNI 229
10.1 Connection types 229

11.3 The PNNI signalling protocol 269
Problems 271

List of standards 273

Glossary of abbreviations 277

Index 283

PART ONE:
INTRODUCTION AND BACKGROUND
In Part One, we present several topics which are part of the background necessary for
understanding ATM networks. It consists of Chapters 1, 2 and 3. Some of the material
presented in these Chapters can be skipped by the knowledgeable reader.

Chapter 1: Introduction
In this Chapter we identify the various forces that gave rise to ATM networks, and
describe some of the well-known standards bodies.


ATM is a technology that provides a single platform for the transmission of voice, video,
and data at specified quality of service and at speeds varying from fractional T1, i.e.,
nX64 Kbps, to Gbps. Voice, data and video are currently transported by different
networks. Voice is transported by the public telephone network, and data by a variety of
packet-switched networks. Video is transported by networks based on coaxial cables,
satellites, and radio waves, and to a limited extent, by packet-switched networks.
In order to understand what caused the development of ATM, we have to go back
to the 80’s! During that decade, we witnessed the development of the workstation and the
evolution of the optical fiber. A dramatic reduction in the cost of processing power and
associated peripherals, such as main memory and disk drives, lead to the development of
powerful workstations capable of running large software. This was a significant
improvement over the older “dumb terminal”. These workstations were relatively cheap
to buy, easy to install and interconnect, and they enabled the development of distributed
systems. As distributed systems became more commonplace, so did the desire to move
files over the network at a higher rate. Also, there was a growing demand for other
applications, such as, video conferencing, multi-media, medical imaging, remote
An Introduction to ATM Networks

4
processing and remote printing of a newspaper. At the same time, optical fiber
technology evolved very rapidly, and by the end of the 80s there was a lot of optical fiber
installed. Optical fiber permitted high bandwidth and very low bit-error rate.
These technological developments coupled with the market needs for faster
interconnectivity, gave rise to various high-speed wide-area networks and services, such
as frame relay, Asynchronous Transfer Mode (ATM), and Switched Multimegabit Data
Services (SMDS).
ATM was standardized by ITU-T in 1987. It is based on packet-switching and it
is connection oriented. An ATM packet, known as a cell, is a small fixed-size packet with
a payload of 48 bytes and a 5-byte header. The reason for using small packets was

connections. Also, there are on going projects in telecommunication companies aiming at
replacing the existing trunks used in the telephone network with an ATM network.
On a smaller scale, ATM is used to provide circuit emulation, a service that
emulates a point-to-point T1/E1 circuit and a point-to-point fractional T1/E1 circuit over
an ATM network. ATM is the preferred solution for ADSL-based residential access
networks used to provide access to the Internet and basic telephone services over the
phone line. Also, it is used in passive optical networks (PON) deployed in residential
access networks.
We conclude this section by noting that arguments in favour and against existing
and emerging new networking technologies will most likely continue for a long time.
There is no argument, however, that these are indeed very exciting times as far as
communication systems are concerned!

1.2 Standards committees

Standards allow vendors to develop equipment to a common set of specifications.
Providers and end-users can also influence the standards so that the vendors’ equipment
conform to certain characteristics. As a result of the standardization process, one can
purchase equipment from different vendors without being bound to the offerings of a
single vendor.
There are two types of standards, namely de facto and de jure. De facto standards
are those which were first developed by a single vendor or a consortium, and then they
were accepted by the standards bodies. De jure standards are those generated through
consensus within national or international standards bodies. ATM, for instance, is the
result of the latter type of standardization.
An Introduction to ATM Networks

6
Several national and international standards bodies are involved with the
standardization process in telecommunication, such as the International

Introduction
7
Contributions to standards are generated by companies, and they are first submitted to
national technical coordination groups, resulting to national standards. These national
coordinating bodies may also pass on contributions to regional organizations or directly
to ITU-T, resulting in regional or world standards. ITU more recently started
recommending and referencing standards adopted by the other groups, instead of re-
writing them.
ITU-T is organized in 15 technical study groups. At present, more than 2500
recommendations (standards) or some 55,000 pages are in force. They are non-binding
standards agreed by consensus in the technical study groups. Although, non-binding, they
are generally complied with due to their high quality and also because they guarantee the
inter-connectivity of networks, and enable telecommunications services to be provided on
a worldwide scale.
ITU-T standards are published as recommendations, and they are organized into
series. Each series of recommendations is referred to by a letter of the alphabet. Some of
the well-known recommendations are the I, Q, and X. Recommendations I are related to
integrated services digital networks. For instance, I.321 describes the B-ISDN protocol
reference architecture, I.370 deals with congestion management in frame relay, and I.371
deals with congestion management in ATM networks. Recommendations Q are related to
switching and signalling. For instance, Q.2931 describes the signalling procedures used
to establish a point-to-point ATM switched virtual connection over the private UNI, and
Q.2971 describes the signalling procedures used to establish a point-to-multipoint ATM
switched virtual connection over the private UNI. Recommendations X are related to data
networks and open system communication. For instance, X.700 describes the
management framework for the OSI basic reference model, and X.25 deals with the
interface between a DTE and a DCE terminal operating in a packet mode and connected
to a public data networks by dedicated circuit.

The International Organization for Standardization (ISO)