GIGABIT-CAPABLE NETWORKSdownload.e-bookshelf.de/download/0000/5923/24/L-G... · 2013. 7. 23. ·...
Transcript of GIGABIT-CAPABLE NETWORKSdownload.e-bookshelf.de/download/0000/5923/24/L-G... · 2013. 7. 23. ·...
-
GIGABIT-CAPABLEPASSIVE OPTICAL
NETWORKS
-
GIGABIT-CAPABLEPASSIVE OPTICAL
NETWORKS
Dave HoodElmar Trojer
-
Copyright � 2012 by John Wiley & Sons. All rights reserved.
Published by John Wiley & Sons, Inc., Hoboken, New Jersey
Published simultaneously in Canada
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, scanning, or otherwise,
except as permitted under Section 107 or 108 of the 1976 United States Copyright Act, without either
the prior written permission of the Publisher, or authorization through payment of the appropriate
per-copy fee to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923,
978-750-8400, fax 978-750-4470, or on the web at www.copyright.com. Requests to the Publisher
for permission should be addressed to the Permissions Department, John Wiley & Sons, Inc.,
111 River Street, Hoboken, NJ 07030, 201-748-6011, fax 201-748-6008, or online at
http://www.wiley.com/go/permission.
Limit of Liability/Disclaimer of Warranty: While the publisher and author have used their best efforts
in preparing this book, they make no representations or warranties with respect to the accuracy or
completeness of the contents of this book and specifically disclaim any implied warranties of
merchantability or fitness for a particular purpose. No warranty may be created or extended by sales
representatives or written sales materials. The advice and strategies contained herein may not be
suitable for your situation. You should consult with a professional where appropriate. Neither the publisher
nor author shall be liable for any loss of profit or any other commercial damages, including
but not limited to special, incidental, consequential, or other damages.
For general information on our other products and services or for technical support, please contact
our Customer Care Department within the United States at 877-762-2974, outside the United States
at 317-572-3993 or fax 317-572-4002.
Wiley also publishes its books in a variety of electronic formats. Some content that appears in print
may not be available in electronic formats. For more information about Wiley products, visit our
web site at www.wiley.com.
Library of Congress Cataloging-in-Publication Data:
Hood, Dave, 1945-
Gigabit-capable passive optical networks / Dave Hood, Elmar Trojer.
p. cm.
Includes bibliographical references and index.
ISBN 978-0-470-93687-0 (cloth)
1. Passive optical networks. 2. Gigabit communications. I. Trojer, Elmar. II. Title.
TK5103.592.P38H66 2011
621.380275–dc232011028223
Printed in the United States of America
10 9 8 7 6 5 4 3 2 1
http://www.copyright.comhttp://www.wiley.com/go/permissionhttp://www.wiley.com
-
To Kent McCammon
-
CONTENTS
Acknowledgments xi
1 Introduction 1
1.1 Target Audience / 3
1.2 Evolution of G-PON Technology and Standards / 3
2 System Requirements 9
2.1 G-PON Operation / 10
2.2 ONU Types / 13
2.3 Network Considerations / 16
2.4 OLT Variations and Reach Extenders / 26
2.5 ONU Powering / 33
2.6 Technology Requirements / 42
2.7 Management Requirements / 44
2.8 Maintenance / 46
3 Optical Layer 49
3.1 Introduction / 50
3.2 Optical Fiber / 54
3.3 Connectors and Splices / 61
3.4 WDM Devices and Optical Filters / 64
3.5 Passive Optical Splitters / 67
3.6 Power Budget / 71
3.7 Coexistence / 77
3.8 Optical Transmitters / 82
3.9 Optical Receivers / 91
3.10 G-PON Transceiver Modules / 106
vii
-
3.11 Optical Amplifiers / 113
3.12 Reach Extension / 118
4 Transmission Convergence Layer 127
4.1 Framing / 129
4.2 ONU Activation / 166
4.3 ONU Transmission Timing and Equalization Delay / 177
4.4 ONU Registration / 187
4.5 ONU Energy Conservation / 190
4.6 Security / 200
5 Management 219
5.1 The Toolkit / 219
5.2 Equipment Management / 240
5.3 Reach Extender Management / 249
5.4 PON Maintenance / 252
5.5 Obsolete Fragments of Information Model / 254
6 Services 255
6.1 Basic Ethernet Management / 256
6.2 Multicast / 275
6.3 Quality of Service / 285
6.4 IP Services / 303
6.5 POTS / 306
6.6 Pseudowires / 317
6.7 Digital Subscriber Line UNIs / 333
6.8 RF Video / 337
7 Other Technologies 339
7.1 Ethernet PON, EPON / 340
7.2 Wireless Broadband / 352
7.3 Copper / 353
7.4 Ethernet, Point to Point / 354
7.5 WDM PON / 357
7.6 Access Migration / 360
viii CONTENTS
-
Appendix I – FEC and HEC in G-PON 363
I.1 Redundancy and Error Correction / 363
I.2 Forward Error Correction / 364
I.3 Hybrid Error Correction / 374
Appendix II – PLOAM Messages 375
II.1. PLOAM Messages in G.987 XG-PON / 375
II.2. PLOAM Messages in G.984 G-PON / 388
References 403
International Telecommunications Union, Telecommunication
Standardization Sector / 403
Broadband Forum / 405
Institute of Electrical and Electronics Engineers / 405
Internet Engineering Task Force / 406
Other / 407
Acronyms 409
Index 423
CONTENTS ix
-
ACKNOWLEDGMENTS
We wish to thank the many who encouraged our effort, in particular, Dave Allan,
Dave Ayer, Chen Ling, Dave Cleary, Jack Cotton, Lou De Fonzo, Jacky Hood, Einar
In de Betou, Denis Khotimsky, Lynn Lu, KentMcCammon, DonMcCullough, Derek
Nesset, Peter €Ohlen, Dave Piehler, Albert Rafel, Bj€orn Skubic, and Mara Williamsfor their help in reviewing various parts of the manuscript and the premanuscript
studies. Tom Anschutz, Paul Feldman, Richard Goodson, David Sinicrope, and
Zheng Ruobin helped to clarify issues that came up in the course of the work.
We would especially like to recognize Dewi Williams, an ideal match for our
target reader profile, who was willing to provide intelligent and thoughtful comment
and discussion well beyond the call of duty.
Needless to say, the inevitable remaining errors and inconsistencies are entirely
our own responsibility.
Finally, special thanks to Jacky and Antonia for their support and patience through
the process.
PUBLISHER’S BRIEF REVIEW
Although thoroughly grounded in the G-PON standards, this book is far more than
just a rehash of the standards. Two experts in G-PON technology explain G-PON in a
way that is approachable without being superficial. As well as thorough coverage of
all aspects of G-PON and its 10Gb/s evolution into XG-PON, this book describes
the alternatives and the reasons for the choices that were made, the history and
the tradeoffs.
xi
-
1
INTRODUCTION
Fiber optic access networks have been a dream for at least 30 years. As speeds
increase, as the disparate networks of the past converge on Ethernet and IP (Internet
protocol), as the technology and business case improve, that dream is becoming
reality.
The access network is that part of the telecommunications network that connects
directly to subscriber endpoints. This book details one of the technologies for fiber in
the loop (FITL), namely gigabit-capable passive optical network (G-PON) technol-
ogy, along with its 10-Gb sibling XG-PON. Figure 1.1 shows how G-PON and XG-
PON fit into the telecommunications network hierarchy. This book is about the
G-PON family.
For quick reference, Table 1.1 summarizes the common PON technologies,
including both the G-PON and EPON families. The G-PON family is standardized
by the International Telecommunications Union—Telecommunication Standardiza-
tion Sector (ITU-T), while EPON comes from the Institute of Electrical and
Electronic Engineers (IEEE). Chapter 7 includes a comparison of G-PON and
EPON.
Because they share many properties, this book uses the term G-PON generically
to refer to either ITU-T G.984 or G.987 systems unless otherwise stated. Where a
distinction needs to be made, we make it explicit: G.984 G-PON or G.987 XG-PON.
Figure 1.2 illustrates the fundamental components of a PON. The head end is
called the optical line terminal (OLT). It usually resides in a central office and usually
Gigabit-capable Passive Optical Networks, First Edition. Dave Hood and Elmar Trojer.� 2012 John Wiley & Sons, Inc. Published 2012 by John Wiley & Sons, Inc.
1
-
serves more than one PON.* The PON contains a trunk fiber feeding an optical power
splitter, or often a tree of splitters. From the splitter, a separate drop fiber goes to each
subscriber, where it terminates on an optical network unit (ONU). ONUs of various
kinds offer a full panoply of telecommunications services to the subscriber.
Telecommunications networks
Access networks
Wireline access networks
Optical access networks
Passive optical networks: PONs
G-PON and XG-PON
1G-EPON and 10G-
EPON
Point-to-point fiberand WDM
PON
Twisted pair copper networks
Co-axial cable
networks
Wireless networks
Core and aggregation networks
Figure 1.1 G-PON taxonomy.
* The term OLT is therefore ambiguous: it may refer only to the terminating optoelectronics and MAC
functionality of a single PON or it may mean the entire access node, terminating a number of PONs and
forwarding traffic to and from an aggregation network.
TABLE 1.1 PON Family Values
Technology Standard Speed
G-PON ITU-T G.984 2.5Gb/s downstream, 1.25Gb/s upstream
XG-PON ITU-T G.987 10Gb/s downstream, 2.5Gb/s upstream
EPON IEEE 802.3 1Gb/s symmetric
10G-EPON IEEE 802.3 10Gb/s downstream, 1 or 10Gb/s upstream
OLT
ONU
Optical power splitter
Opticaldrop
Optical trunkline OTL
Access network
Optical distribution network (ODN)PON
ONU
Downstream Upstream
Bidirectional single mode optical fiber
To aggregation and core network
Subscriber premises network
Figure 1.2 G-PON terminology.
2 INTRODUCTION
-
A single-fiber connection is used for both directions, through specification of
separate wavelengths for each direction. As we shall see, wavelength separation
also allows for coexistence of other technologies on the same optical distribution
network (ODN).
After a brief overview and history in this chapter, Chapter 2 outlines the require-
ments and constraints of a G-PON access network. Chapter 3 explains the optical layer
of the network. Moving up the stack, Chapter 4 covers the transmission convergence
layer, the home of most of the features that uniquely distinguish G-PON from other
access technologies. Chapter 5 introduces the management model in the context of
equipment and software management, while Chapter 6 shows how the management
model is used to construct telecommunications services. Finally, Chapter 7 describes
current and future alternatives, competitors, and partners of G-PON.
Two appendices, a list of references, a guide to acronyms and abbreviations, and
an index appear at the end of the book.
1.1 TARGET AUDIENCE
This book is written for the experienced telecommunications or data communica-
tions professional whose knowledge base does not yet extend into the domain of PON
or, in particular, G-PON and XG-PON. We also address this book to the advanced
student, who cannot be expected to have a grounding in the ancient and forgotten lore
of telecoms. Hoping to strike a balance against excessive redundancy, we neverthe-
less include a certain amount of background material, for example, on DS1 and E1
TDM services, and we always try to indicate where to find additional information.
We argue that a simple restatement of the standards adds no value. Accordingly,
we structure this book with a view toward explaining and comparing the standards,
rather than simply paraphrasing them. This is most evident in the frequent side-by-
side comparisons of G-PON and XG-PON. This book also addresses many important
aspects of real-world access networks that lie beyond the scope of the standards.
Disclaimers The complete and authoritative specifications are in the standards
themselves. While we make every attempt to be accurate, we have necessarily elided
any number of secondary details, especially in the peripheral standards. We trust that
the reader who ventures into the formal standards will find few surprises. Although
both authors are employed by Ericsson, we should also state that this book is not
sponsored by Ericsson, and the views expressed do not necessarily represent Ericsson
positions.
1.2 EVOLUTION OF G-PON TECHNOLOGY AND STANDARDS
PON technology began in the 1980s with the idea of a fiber ring dropping service to
each subscriber. Ring topology was abandoned early for reasons that will be apparent
EVOLUTION OF G-PON TECHNOLOGY AND STANDARDS 3
-
upon reflection, and subsequent PONs have been based on optical trees. In the early
days, several companies* developed products around the integrated services digital
network (ISDN) standards. The systems delivered plain old telephone service
(POTS), but offered few advantages over copper-fed POTS. Although there were
some deployments, the technology (cost) and the market need (revenue) were too far
apart to justify a realistic business case.
By the 1990s, optical communications were starting to mature in the long-haul
network, speeds were increasing, and the industry was starting down the cost curve
on the technology side. At the same time, a market for Internet access was
developing, and subscribers were increasingly frustrated with modems running at
9.6 or even at the once-impressive speed of 56 kb/s. The PON industry tried again.
The first generation of what might be termed modern PON was based on
asynchronous transfer mode (ATM), originally designated A-PON. Significant
commercial deployments of ATM PON occurred under the moniker B-PON (broad-
band PON). B-PON was standardized and rolled out in the last years of the twentieth
century.
Like G-PON, B-PON is defined by the ITU-T, in the G.983 series of recom-
mendations. Several data rates are standardized. Early deployments delivered data
services only, at aggregate bit rates of 155Mb/s, both upstream and down. This is one
of the bit rates used by the synchronous digital hierarchy (SDH), an optical transport
technology developed during the 1980s and 1990s and widely deployed today,
albeit having evolved over the years to the higher rates of 2.5 and 10Gb/s and beyond.
B-PON specified the same bit rates as SDH, with the intention to reuse component
technology and also parts of the SDH standards for specifications such as jitter.
In the early years of the millennium, B-PON matured in several ways:
. Service definitions expanded beyond best-efforts data, most notably to include
POTS and voice over Internet protocol (VoIP).
. The original downstream wavelength spectrum was redefined into two bands, a
basic band for use by the B-PON protocols and an enhancement band intended
for radio frequency (RF) content such as broadcast video. Spectrum was also
identified for use by independent dense wavelength division multiplex
(DWDM) access, coexisting on the same fiber plant.
. To improve the utilization of upstream capacity, dynamic bandwidth allocation
(DBA) was defined and standardized.
. Although the common upstream rate remained at 155Mb/s, the technology,
cost, and market requirements had evolved to the point that 622Mb/s—another
SDH speed—became the expected downstream rate.
*One of the original companies—DSC-Optilink—can be tied to today’s Alcatel-Lucent; another—
Raynet—can be linked to Ericsson.
4 INTRODUCTION
-
Reasonable B-PON deployment volume was achieved at these levels. At the time of
writing, in 2011, B-PON was still being installed to fill out empty slots in existing
chassis.
Although the B-PON standards are ITU-T recommendations, a group called
FSAN (Full Service Access Network) guided the requirements and recommenda-
tions and continues to guide its successors to this day. FSAN is an informal
organization of telecommunications operators founded in 1995. Unlike formal
standards development organizations (SDOs) such as ITU-T and IEEE, and non-
SDOs such as Broadband Forum (BBF), FSAN has no membership fees and no staff.
Equipment and component vendors are members by invitation only. While FSAN
emphasizes that it is not an SDO, the same companies and the same people carry their
discussions from FSAN into ITU-T for the formal standardization work, often on
successive days of the same meeting. In the early days, the text of the recommenda-
tions was actually developed under the FSAN umbrella, then passed to ITU-T for
formal review and consent.
Around the turn of the millennium, digital video began to come out of the lab. The
bandwidth limitations of B-PON became a concern—622Mb/s distributed across 32
subscribers is only (!) 20Mb/s average rate per subscriber, much of which would be
consumed by a single contemporary high-definition digital video stream—while the
cost of technology had continued to improve. Standardization discussions began on a
new generation of PON, this one known as gigabit-capable PON, or G-PON. The first
G-PON standards, the ITU-T G.984 series, were published in 2003. As with B-PON,
the G-PON standards recognize several data rates, but the only rate of practical
interest runs at 2.488Gb/s downstream, with 1.244Gb/s in the upstream direction,
capacity shared among the ONUs. These are also SDH bit rates.
The initial versions of the G.984 series recognized the ATM of B-PON, but ATM
was subsequently deprecated as a fading legacy technology. Another capability that
was initially standardized and later deprecated was provision for G-PON to directly
carry TDM (time division multiplex) traffic. This might have been useful for services
such as DS1 (digital signal level 1) or SDH, but detailed mappings were never
defined, and it was overtaken by the development of standards for pseudowires, about
which we shall learn in Chapter 6.
The only form of payload transport that remains in G.984 today is the G-PON
encapsulation method (GEM), usually encapsulating Ethernet frames.* It is perfectly
accurate to think of G-PON as an Ethernet transport network, notwithstanding
marketing claims from the EPON competition that G-PON is not real Ethernet.
Another important evolutionary step from B-PON to G-PON is accommodation
of the operators’ requirement for incremental upgrade of already deployed installa-
tions. The B-PONwavelength plan did not provide for the coexistence of B-PON and
G-PON on the same optical network. The eventual need to upgrade access network
technology thus presented a dilemma:
*Other mappings are also defined in G.984 G-PON; some were recognized to be of no market interest
and were not carried forward into G.987 XG-PON. The only additional mapping in G.987 XG-PON is
multiprotocal label switching (MPLS) over GEM. Time will tell whether it is useful.
EVOLUTION OF G-PON TECHNOLOGY AND STANDARDS 5
-
. It was usually not economical to install a new optical distribution network,
particularly the distribution and drop segments, in parallel with an existing one.
. It was not feasible to replace all ONUs on a PON at the same instant. Imagine an
army of 32 service technicians calling on 32 subscribers at precisely 10 AM
next Tuesday morning—or at any other time for that matter!
. It was unacceptable to shut down telecommunications service to a group of
subscribers for several hours or days to allow for a realistic number of service
technicians to schedule realistic service appointments with subscribers.
. And maybe only 1 of those 32 subscribers was willing to pay for upgraded
service anyway.
The upshot of this consideration was a requirement for G-PON networks to reserve
wavelengths to allow incremental upgrade to the next generation of PON technology,
whatever that might be, and to include the necessary wavelength blocking filters in
ONUs. Chapters 2 and 3 discuss this in further detail.
G-PON began to be deployed in substantial volume in 2008 and 2009.
Once a standard is implemented and deployed, it is natural to want to confirm that
everyone has the same interpretation and that the various implementations will
interwork. This led to a series of interoperability test events, beginning with the basic
ability of an OLT to discover and activate an ONU on the PON. The first G-PON
plugfest occurred in January 2006, and there have been two to four events per year
since then. Today’s G-PON equipment is largely interoperable, although the final
proof remains to be seen: there have not yet been widespread live deployments of
multivendor access networks.
As testing moved further up the stack, it became apparent that the flexibility of the
ONU management and control interface (OMCI) was not an unmixed blessing.
Different vendors supported given features in different ways. If the OLT tried to
provision the feature in oneway and the ONU supported only some different way, the
pair would not interoperate.
Interoperability was a primary motivation for standardization. FSAN therefore
created the OMCI implementation study group (OISG) with the charter to develop
best practices, recommendations for the preferred ways to implement various
features. OISG was and is a vendors-only association, theoretically free to discuss
implementation considerations under mutual nondisclosure agreements.
In 2009, OISG released an implementers’ guide of OMCI best practices, originally
published as a supplement to the OMCI specification ITU-TG.984.4. As G.984.4 was
migrated into G.988, the implementers’ guide material was incorporated into G.988,
where it resides today.
OMCI best practices continue to evolve as minor questions arise, but the issues
that spawned OISG have largely been resolved. OISG’s charter also evolved and it
became an early preview forum for OMCI maintenance, an opportunity for sanity
checks and consensus building before new OMCI proposals were formally submitted
to the ITU-T process.
6 INTRODUCTION
-
Although OISG has not been formally disbanded, it is now dormant, both because
of its success at resolving interoperability issues and because of the shift of
responsibility from FSAN to BBF. Test plans are published as BBF technical reports,
specifically TR-247 and TR-255. Responsibility for plugfests also shifted from
FSAN to.
Broadband Forum entered the G-PON scene only recently, but in a major way.
Previously known as DSL Forum, BBF changed its name and expanded its scope to
include, among other things, the entire access network and everything attached to it.
In 2006, BBF published TR-101, which defined requirements for migration of the
access network from ATM to Ethernet, but still with a DSL mind-set. The ink had
scarcely dried on TR-101 when BBF began a project to define its applicability to the
special aspects of G-PON, resulting in TR-156 (2008).
Although there are some rough edges at the organizational boundaries, the scope
addressed by BBF is theoretically disjoint from the scope of the ITU-T recommenda-
tions. ITU-T specifies an interface between OLT and ONU and the fundamentals
of its operation. ITU-T includes tools for maintenance and tools from which
applications can be constructed and extends the tool set as necessary when new
applications arise.
In contrast, BBF views the overall network architecture as its scope. BBF takes the
ITU-T tools as a given, identifies preferred configuration options, and writes
network- and service-level requirements to serve these options. Vendors and
operators are, of course, free to develop other applications on the same base, but
if BBF does its job well, its model architectures prove to be satisfactory for most real-
world needs and are suitable as reference models even for applications that lie
beyond the strict bounds of the BBF architecture.
If BBF does its job well? In fact, TR-156 has largely been accepted by operators
worldwide as a satisfactory model for simple ONUs that deliver Ethernet service
to end users. Additional BBF technical reports (TR-142, TR-167) define how the
G-PON toolkit can be used to control only the ONU’s PON interface, with the
remainder of the ONU managed through other means. Chapter 2 expands this topic.
Once the G-PON standards began to mature and vendors busied themselves
bringing product to market, FSAN turned its attention to the question of the next
logical step after G-PON. Starting in about 2007, as the outline became clearer, the
operators launched a white paper project to define the requirements for the next
generation. FSAN completed the next-gen white paper in mid-2009. Parallel work
had begun in late 2008 to develop the details of the necessary recommendations.
FSAN structured its view of the future into two domains: Next-gen 1 (NG-1) was
the set of PON architectures that was required to coexist on the same fiber
distribution network with G-PON, while next-gen 2 designated the realm of
possibilities freed from that constraint, an invitation to take a long view of technology
to see what might make sense at some unspecified time in the future.
As it turned out, NG-1 PON bifurcated further, into versions known as XG-PON1
and XG-PON2, or often just XG-1 and XG-2, where X is the Roman numeral 10,
designating the nominal 10 Gb/s downstream rate. Both versions run downstream
data at 9.953Gb/s, another SDH bit rate. The upstream capacity of XG-PON1 is
EVOLUTION OF G-PON TECHNOLOGY AND STANDARDS 7
-
2.488Gb/s, while XG-PON2 runs at 9.953Gb/s upstream. The reason for the
distinction was the substantial difference in technological challenge, coupled with
the perception that market need was insufficient to justify the significantly higher
cost of high-speed upstream links.
XG-PON1 is standardized in the ITU-T G.987 series of recommendations. At the
time of writing, XG-PON2 is being held in abeyance, pending the convergence of
market demand and technological feasibility. Because IEEE has already standard-
ized a symmetric 10G form of EPON (10G-EPON, Chapter 7), it is possible that
XG-PON2 will not be pursued further. If that comes to pass, operators who need
symmetric 10G will deploy 10G-EPON, and the G-PON community will work
toward next-next-generation access, probably WDM PON.
8 INTRODUCTION
-
2
SYSTEM REQUIREMENTS
In this chapter:
. Overview of power-splitting PONs
. Optical network considerations
T Split ratio
T Maximum reach, differential reach
T PON protection
. Reach extenders
. Coexistence with future generations
. Types of ONU
. ONU powering
. Introduction to
T Dynamic bandwidth assignment; details in Chapter 6
T Quality of service; details in Chapter 6
T Security; details in Chapter 4
T Management; details in Chapter 5
Before diving into the details of how a G-PON works, we need to understand
something about the business case. We return repeatedly to business case questions
over the course of this book because, ultimately, everything we do must add value to
someone for something.
Gigabit-capable Passive Optical Networks, First Edition. Dave Hood and Elmar Trojer.� 2012 John Wiley & Sons, Inc. Published 2012 by John Wiley & Sons, Inc.
9
-
As with most companies, telecommunications operators are driven forward by
market opportunity, cost reduction, and competitive pressure, and they are held back
by existing investment and existing practices. New technology is comparatively easy
to justify in a greenfield development—we have to do something, so let us go for the
latest and greatest!—but most of the potential market is already served in one way or
another, even if it’s no more than ADSL (asymmetric digital subscriber line) from
a central office. The difficulty arises in making a business case for the deployment
of a new technology that may be of immediate interest to only a small number of
existing subscribers, in what is called, for contrast, a brownfield.
Civil works—right of way acquisition, permits, trenching for underground cable,
poles for aerial cable—are a very large part of the up-front cost of a change in
technology, for example, from copper to fiber. Estimates range from 65 to as much as
80% of the total cost. No matter how economical the equipment itself may be, this
cost must be paid. Once the business case has been made to install new fiber in the
outside plant infrastructure, it makes sense to place large fiber-count cables, or at
least a lot of empty ducts, through which fiber can easily be blown at a later date. The
cost of additional ducted fibers is comparatively small, even vanishingly small, in
fiber trunks. With the optical infrastructure in place and spare fibers available, it
becomes much easier to take subsequent evolutionary steps.
It is easier to develop a business case if all telecommunications services can be
provided by a single network. This is the idea behind the oft-heard term convergence,
a concerted effort to eliminate parallel networks, each of which serves only a subset
of the service mix. Software-defined features, Ethernet, and IP are major steps along
the road to convergence. The contribution of standards and of the network equipment
is to ensure that the investment, once made, can be used for a complete range of
services for decades to come. In keeping with the full-service focus of its FSAN
parent, G-PON is designed to deliver any telecommunications service that may be
needed.
2.1 G-PON OPERATION
2.1.1 Physical Layer
To recapitulate the brief overview in Chapter 1, a PON in general, and a G-PON in
particular, is built on a single-fiber optical network whose topology is a tree, as shown
in Figure 2.1. The OLT is at the root, and some number of ONUs connect at the
leaves. Downstream optical power from the OLT is split at the branching points of
the tree. Each split allocates an equal fraction of the power to each branch. The
achievable reach of a PON is a tradeoff of fiber loss against the division of power at
the splitter. Chapter 3 describes splitters and fiber loss in detail.
A power splitter is symmetric: Its loss upstream is the same as down, 3 dB for each
power of 2 in the split ratio.
The OLT transmits a continuous downstream signal that conveys timing, control,
management, and payload to the ONUs. The OLT is master of the PON. Based on
10 SYSTEM REQUIREMENTS
-
service-level commitments and traffic offered by the ONUs, the OLT continuously
develops an upstream capacity allocation plan for the near future—typically 1 or
2ms—and transmits this so-called bandwidth map to the ONUs. The ONU is
permitted to transmit only when explicitly given permission by a grant contained
in a bandwidth map. During its allocated time, the ONU sends a burst of data
upstream, data that includes control, management, and payload.
For the bursts to arrive at the OLT at precisely the proper interleaved times, each
ONU must offset its notion of a zero reference transmission time by a value
determined by its round-trip delay,* the time it takes for the signal from the OLT
to reach the ONU, plus ONU processing delay and the time it takes for the signal
from the ONU to reach the OLT. The OLT measures the round-trip delay of each
ONU during activation and programs the ONU with the compensating equalization
delay value.
Another low-level requirement on a PON is the discovery of new ONUs, be they
either newly installed devices or existing devices that have been offline for reasons
such as fiber failure or absence of power, whether intentional or not. The OLT
periodically broadcasts a discovery grant, which authorizes any ONU that is not yet
activated on the PON to transmit its identity. Since the round-trip time of a new ONU
is unknown, the OLT opens a quiet window, a discovery window, also called a
ranging window, a time interval during which only unactivated ONUs are permitted
to transmit.
It is possible that more than one unactivated ONU could attempt to activate at
the same time; if their transmissions overlapped in time, neither would succeed.
Worse, they could deadlock, repeatedly colliding on every discovery grant forever.
The ranging protocol therefore specifies that the ONU introduce a random delay in its
response to the OLT’s invitation. Even though the transmissions from two ONUs
may collide during a given discovery cycle, they will sooner or later appear as distinct
activation requests in some subsequent interval.
The size of the discovery window depends on the expected fiber distance between
the farthest possible ONU and the nearest possible ONU. This is called maximum
Downstream
SplitterOLTONU
Bidirectional single mode optical fiber
Optical power splitter
Aggregation and core network
Trunk fiber,Optical trunk link OTL
Optical drop
Home network
Access network
PONUpstream
Figure 2.1 Tree structure of a PON.
*True, the correction could instead be applied locally by the OLTas an offset in the bandwidth map, but it is
not.
G-PON OPERATION 11
-
differential reach, standardized with 10-, 20-, and 40-km options in G.984 G-PON.
In G.987 XG-PON, the maximum differential reach options are 20 and 40 km.
Chapter 4 goes into detail on all of these aspects of G-PON operation.
2.1.2 Layer 2
In terms of the OSI seven-layer communications model,* the access network largely
exists at layer 2. Perhaps the single most important concept underlying an Ethernet-
based access network—which G-PON is—is that of the virtual local area network
(VLAN), specified in IEEE 802.1Q. A very substantial part of the hardware,
software, and management of a G-PON is dedicated to classifying traffic into
VLANs, then forwarding the traffic according to VLAN to the right place with the
right quality of service (QoS). Although an ONU is modeled as an IEEE 802 MAC
bridge, MAC addresses are usually less important at the ONU than are VLAN tags.
The access network operates at layer 2, but it judiciously includes some layer 3
functions as well, particularly for multicast management. For practical purposes,
multicast means IPTV (Internet protocol television) service; it is expected to
represent a large fraction of the traffic and to yield a large part of the revenue
derived from a G-PON. The PON architecture is ideally suited for multicast
applications because a single copy of a multicast signal on the fiber can be
intercepted by as many ONUs as need it. Each ONU extracts only the multicast
groups (video channels) that are requested by its subscribers.
To determine which groups are requested at any given time, the ONU includes at
least an IGMP/MLD{ snoop function, about which we shall learn more in Chapter 6.
Snooping involves monitoring transmissions from the subscriber’s set-top box
(STB), based on which the ONU compares the requested channels with a local
access control list (ACL). If the requested content is authorized and is already
available on the PON, the ONU delivers it immediately without further ado. Also
acting as an IGMP/MLD snoop, or more likely as a proxy, the OLT likewise
determines whether a given multicast group is already available, or whether it needs
to be requested from yet a higher authority. As seen by a multicast router further up
the hierarchy, a proxy aggregates a number of physical STBs into a single virtual
STB, thereby avoiding unnecessary messages to the router and improving network
scalability.
It is an open question what statistical capacity gain should be expected from
multicast, now and in the future. Even if 80% of subscribers are watching the same
10 channels, a long statistical tail would require substantial capacity to carry
content of interest only to the remaining few. Will a PON with 50 subscribers,
each with 2 or more television sets and a recording device, need 50 multicast groups?
Thirty? Twenty?
* See ITU-T X.200.{ IGMP: Internet group management protocol (IPv4), MLD: multicast listener discovery (IPv6). We
usually spell out acronyms the first time they appear; acronyms are also listed in a separate section at the
end of the book.
12 SYSTEM REQUIREMENTS
-
There is a general expectation that video will move toward unicast, but no one is
prepared to say how soon. At the end of the day, it may not matter. The considerations
described above suggest that we should expect a busy hour load of two or three
multicast groups per subscriber. At bit rates on the order of 5Mb/s per multicast
group, G-PON has enough downstream capacity for that level of loading. If it were to
materialize, mass market demand for ultrahigh bandwidth unicast video, up to
65Mb/s per channel, could motivate further access network upgrade.
2.2 ONU TYPES
A variety of product configurations seeks to fit the range of operators’ needs
completely and optimally. Here we outline a few of the possibilities.
2.2.1 Single-Family ONU
At least in some markets, the single-family unit (SFU) is the most common form of
ONU. Predictably, there are many variations on the SFU theme. The SFU may be
located indoors or out. Power is always supplied by the subscriber, but the SFU may
or may not include battery backup. The SFU may be regarded as a part of the
telecommunications network, owned and managed by the operator, or it may be
considered to be customer premises equipment (CPE), owned by the subscriber.
The simplest SFU, such as the one illustrated in Figure 2.2 with its cover off,
delivers one Ethernet drop; it is essentially a G-PON-to-Ethernet conversion device.
This one is intended to be mounted on a wall, at the demarcation point between the
drop fiber and the subscriber’s home network. The single Ethernet feed would then
Figure 2.2 Single-family ONU with power brick.
ONU TYPES 13
-
be connected to a residential gateway (RG) at some location convenient to the
subscriber’s device layout.
Many SFUs, such as the one in Figure 2.3, add value by including several bridged
Ethernet drops, suitable for direct connection to several subscriber devices, for
example, two or three PCs and a set-top box. Some may also include built-in
terminations for one or two POTS lines. Other applications for home use might
include low-rate telemetry, for example, to read utility meters or to monitor intrusion
detectors. M2M (machine to machine) communications are expected to mushroom
over the next few years, and the SFU will surely play a part in backhauling
information to centralized servers.
The SFUmay also include a full residential gateway, with firewall, NAT (network
address translation) router, DHCP (dynamic host configuration protocol) server,
802.11 wireless access, USB ports, storage or print server, and more. This form of
SFU is typically managed jointly by the subscriber, by the ONU management and
control interface (OMCI, G.988) model of G-PON, and by an access control server
(ACS), the latter as defined in various Broadband Forum technical reports and
frequently short-handed as TR-69.*
2.2.2 Multi-Dwelling Unit ONU
The multiple dwelling unit (MDU) is an ONU that serves a number of residential
subscribers. It may be deployed in an apartment building, a condominium complex,
or at the curbside. The MDU is always considered to be part of the telecommunica-
tions network; that is, its power, management, and maintenance are the responsibility
of the operator. Depending on their target markets, MDUs typically serve from 8 to
24 subscribers. Very similar to the MDU, a G-PON-fed digital subscriber line access
multiplexer (DSLAM) may serve as many as 48 or even 96 subscribers.
Subscriber drops from an MDU may be Ethernet, but the IEEE 802.3 physical
layer is not specified to tolerate the stress of a full outdoor environment, specifically
lightning transients. Even if the MDU is housed indoors in the same building as the
Figure 2.3 SFU with enhanced functionality.
*BBF designates it TR-069. It is always pronounced without the zero, and we like to write it in the same
way we say it.
14 SYSTEM REQUIREMENTS
-
subscriber residences, it may be uneconomical to rewire the building with the cat-5
cable needed for Ethernet.
The alternative subscriber drop technology is DSL. When drops are short, the
preferred form of DSL is ITU-T G.993.2 VDSL2; such an MDUmay or may not also
offer POTS. Existing telephone-grade twisted pair runs from the MDU to the
subscriber premises, where there is a DSL modem and a splitter for POTS, if POTS
is included in the service. With the short drops implied by fiber to the curb, it is
feasible to deliver several tens of megabits per second—even 100Mb/s and more—
effectively overcoming the speed limitations of copper wiring. The rate-reach
maximum can be extended through bonding of services across two or more pairs,
while G.993.5 vectoring potentially increases attainable speed through crosstalk
cancellation.
2.2.3 Small-Business-Unit ONU
As well as the ubiquitous Ethernet service, a small business unit (SBU) is likely to
offer several POTS lines to a small-office customer. It may also support a few TDM
(time-division multiplex) services such as DS1 or E1 via pseudowire emulation
(Chapter 6 explains this). The SBU of Figure 2.4 has eight POTS lines, four Ethernet
drops, and four 2.048-Mb/s E1 TDM services.
The cellular backhaul unit (CBU) is a variation of the SBU—perhaps a new
category in its own right. In the cellular backhaul application, the ONU carries traffic
between the core network and a radio base station. Legacy mobile backhaul requires
interfaces such as DS1 or E1. As the cell network migrates from third to fourth
generation, Ethernet backhaul is displacing DS1 and E1. As well as the tightly
controlled frequency stability required of all TDM services, some wireless protocols
require a precise time of day reference, a function described in Chapter 4.
Another variation of the SBU is the multitenant unit (MTU), intended to be shared
by several small businesses. The target market is the small islands of commercial
activity common along major streets. The important distinction of the MTU from the
SBU is its need to isolate services one from another, both in terms of traffic—no
bridging between Ethernet ports—and in terms of service-level agreements (SLAs).
Figure 2.4 An SBU.
ONU TYPES 15
-
2.3 NETWORK CONSIDERATIONS
While some operators favor the CPE model, in which the ONU is indoors, located on
the subscriber’s desktop or perhaps mounted on an indoor wall, other operators wish
to deploy ONUs outdoors. To a considerable extent, this reflects a difference in the
operator’s perspective: ONU as part of the telecommunications network or ONU as
subscriber-owned device. Figure 2.5 illustrates such an outdoor ONU, which differs
from the device of Figure 2.2 in that it provides two POTS lines, as well as an
Ethernet drop, and is accessible to operator personnel without the need to enter the
subscriber’s home.
ONUs such as MDUs may go into equipment rooms or telecommunications
closets in buildings. ONUs may also be designed for curbside pedestals (Fig. 2.6) or
other outdoor housings, in which case they need to be fully hardened for outside plant
conditions. ONU components must generally be rated for the full industrial temper-
ature range, and ONU enclosures may be required to tolerate extremes of tempera-
ture and water exposure, including immersion (Fig. 2.7) and salt fog. Other
considerations for outdoor ONUs include lightning protection for all metallic wiring,
and insect and fungus resistance. All ONUs must satisfy regulatory requirements for
electromagnetic interference (EMI) generation and operator requirements for EMI
tolerance.
2.3.1 Power
ONU powering is indisputably a network consideration, but it warrants a separate
discussion in its own right. We defer this topic to Section 2.5.
Figure 2.5 Outdoor ONU, outer access cover open.
16 SYSTEM REQUIREMENTS