COMPUTER NETWORKS
`
`ANDREW S. TANENEAUM
`
`CoIIisio
`
`5
`
`4 2
`
`4|001:0O04|
`
`01.
`
`1100110110
`
`0110110011
`
`Broadcastin
`
`SHALL I COMPARE THEE
`TO A SUMMEFI'S DAY?
`
`WELCOME
`
`TO THE
`
`INFORMATION
`
`SUPER
`
`HIGHWAY
`
`APPLE 1015 - Page 1
`
`
`
`
`

`
`
`
`Library of Congress Cataloging in Publication Data
`
`Tanenbaum, Andrew S. 1944-.
`
`Computer networks / Andrew S. Tanenbaum. —- 3rd ed.
`p.
`cm.
`
`Includes bibliographical references and index.
`ISBN 0—13—349945—6
`1.Computer networks.
`TK5105.5.T36 1996
`0O4.6——dc2O
`
`I. Title.
`
`96-4121
`CIP
`
`Editorial/production manager: Camille Trentacoste
`Interior design and composition: Andrew S. Tanenbaum
`Cover design director: Jerry Votta
`Cover designer: Don Martinetti, DM Graphics, Inc.
`Cover concept: Andrew S. Tanenbaum, from an idea by Marilyn Tremaine
`Interior graphics: Hadel Studio
`Manufacturing manager: Alexis R. Heydt
`Acquisitions editor: Mary Franz
`Editorial Assistant: Noreen Regina
`
`
`
`© 1996 by Prentice Hall PTR
`Prentice-Hall, Inc.
`A Simon & Schuster Company
`Upper Saddle River, New Jersey 0745 8
`
`The publisher offers discounts on this book when ordered in bulk quantities. For more information,
`contact:
`
`Corporate Sales Department, Prentice Hall PTR, One Lake Street, Upper Saddle River, NJ 07458.
`Phone: (800) 382-3419; Fax: (201) 236-7141. E-mail: corpsales@prenhal1.com
`
`All rights reserved. No part of this book may be reproduced, in any form or by any means, without
`permission in writing from the publisher.
`
`All product names mentioned herein are the trademarks of their respective owners.
`
`Printed in the United States of America
`10
`9
`8
`7
`6
`5
`4
`
`ISBN O-13-349945-6
`
`Prentice-Hall International (UK) Limited, London
`Prentice-Hall of Australia Pty. Limited, Sydney
`Prentice-Hall Canada Inc., Toronto
`Prentice-Hall Hispanoamericana, S.A., Mexico
`Prentice-Hall of India Private Limited, New Delhi
`Prentice—Ha11 of Japan, Inc., Tokyo
`Simon & Schuster Asia Pte. Ltd., Singapore
`Editora Prentice-Hall do Brasil, Ltda., Rio de Janeiro
`
`APPLE 1015 - Page 2
`
`APPLE 1015 - Page 2
`
`

`
`SEC. 2.4
`
`THE TELEPHONE SYSTEM
`
`131
`
`town, and the other one’s wife was the town telephone operator. He quickly saw
`that either he was going to have to invent automatic telephone switching equip-
`ment or he was going to go out of business. He chose the first option. For nearly
`100 years,
`the circuit switching equipment used worldwide was known as
`Strowger gear.
`(History does not record whether the now—unemployed switch-
`board operator got a job as an information operator, answering questions such as:
`What is the phone number of an undertaker?
`The model shown in Fig. 2—34(a) is highly simplified of course, because parts
`of the “copper” path between the two telephones may,
`in fact, be microwave
`links onto which thousands of calls are multiplexed. Nevertheless, the basic idea
`is valid: once a call has been set up, a dedicated path between both ends exists and
`will continue to exist until the call is finished.
`
`when call is made
`
`Physical copper
`connection set up
`
`\/
`
`Switching office
`
`COWPUYGV
`
`
`
`
`Packets queued up
`for subsequent
`transmission
`
`Computer
`
`Fig. 2-34. (a) Circuit switching. (b) Packet switching.
`
`An important property of circuit switching is the need to set up an end—to—end
`path before any data can be sent. The elapsed time between the end of dialing and
`the start of ringing can easily be 10 sec, more on long—distance or international
`calls. During this time interval, the telephone system is hunting for a copper path,
`as shown in Fig. 2—35(a). Note that before data transmission can even begin, the
`call
`request signal must propagate all
`the way to the destination, and be
`
`APPLE 1015 - Page 3
`
`APPLE 1015 - Page 3
`
`

`
`SEC. 2.4
`
`THE TELEPHONE SYSTEM
`
`131
`
`town, and the other one’s wife was the town telephone operator. He quickly saw
`that either he was going to have to invent automatic telephone switching equip-
`ment or he was going to go out of business. He chose the first option. For nearly
`100 years,
`the circuit switching equipment used worldwide was known as
`Strowger gear.
`(History does not record whether the now-unemployed switch-
`board operator got a job as an information operator, answering questions such as:
`What is the phone number of an undertaker?
`The model shown in Fig. 2—34(a) is highly simplified of course, because parts
`of the “copper” path between the two telephones may,
`in fact, be microwave
`links onto which thousands of calls are multiplexed. Nevertheless, the basic idea
`is valid: once a call has been set up, a dedicated path between both ends exists and
`will continue to exist until the call is finished.
`
`when call is made
`
`Physical copper
`connection set up
`
`Switching office
`/ .
`
`
`Packets queued up
`for subsequent
`transmission
`
`
`
`Computer
`
`(13)
`
`Fig. 2-34. (a) Circuit switching. (b) Packet switching.
`
`An important property of circuit switching is the need to set up an end—to~end
`path before any data can be sent. The elapsed time between the end of dialing and
`the start of ringing can easily be 10 sec, more on long—distance or international
`calls. During this time interval, the telephone system is hunting for a copper path,
`as shown in Fig. 2—35(a). Note that before data transmission can even begin, the
`call
`request signal must propagate all
`the way to the destination, and be
`
`APPLE 1015 - Page 4
`
`APPLE 1015 - Page 4
`
`

`
`132
`
`THE PHYSICAL LAYER
`
`CHAP. 2
`
`acknowledged. For many computer applications (e. g., point—of—sa1e credit verifi-
`cation), long setup times are undesirable.
`
`Call request signal
`l
`
`Propagation
`delay
`
`:1-
`r
`
`Call
`
`accept
`it signal
`
`<—-————Time
`
`(3)
`
`(b)
`
`(C)
`
`Fig. 2-35. Timing of events in (a) circuit switching,
`(c) packet switching.
`
`(b) message switching,
`
`As a consequence of the copper path between the calling parties, once the
`setup has been completed, the only delay for data is the propagation time for the
`electromagnetic signal, about 5 msec per l0OO km. Also as a consequence of the
`established path, there is no danger of congestion—that is, once the call has been
`put through, you never get busy signals, although you might get one before the
`connection has been established due to lack of switching or trunk capacity.
`An alternative switching strategy is message switching, shown in Fig.2-
`35(b). When this form of switching is used, no physical copper path is established
`in advance between sender and receiver. Instead, when the sender has a block of
`data to be sent, it is stored in the first switching office (i.e., router) and then for-
`warded later, one hop at a time. Each block is received in its entirety, inspected
`
`APPLE 1015 - Page 5
`
`APPLE 1015 - Page 5
`
`

`
`SEC. 2.4
`
`THE TELEPHONE SYSTEM '
`
`133
`
`for errors, and then retransmitted. A network using this technique is called a
`store-and-forward network, as mentioned in Chap. 1.
`The first electromechanical telecommunication systems used message switch-
`ing, namely for telegrams. The message was punched on paper tape off—line at the
`sending office, and then read in and transmitted over a communication line to the
`next office along the way, where it was punched out on paper tape. An operator
`there tore the tape off and read it in on one of the many tape readers, one per out-
`going trunk. Such a switching office was called a torn tape office.
`With message switching, there is no limit on block size, which means that
`routers (in a modern system) must have disks to buffer long blocks.
`It also means
`that a single block may tie up a router—router line for minutes, rendering message
`Switching useless for interactive traffic. To get around these problems, packet
`switching was invented. Packet—switching networks place a tight upper limit on
`block size, allowing packets to be buffered in router main memory instead of on
`disk. By making sure that no user can monopolize any transmission line very long
`(milliseconds), packet—switching networks are well suited to handling interactive
`traffic. A further advantage of packet switching over message switching is shown
`in Fig. 2—35(b) and (c): the first packet of a multipacket message can be forwarded
`before the second one has fully arrived, reducing delay and improving throughput.
`For these reasons, computer networks are usually packet switched, occasionally
`circuit switched, but never message switched.
`Circuit switching and packet switching differ in many respects. The key
`difference is that circuit switching statically reserves the required bandwidth in
`advance, whereas packet switching acquires and releases it as it is needed. With
`circuit switching, any unused bandwidth on an allocated circuit is just wasted.
`With packet switching it may be utilized by other packets from unrelated sources
`going to unrelated destinations, because circuits are never dedicated. However,
`just because no circuits are dedicated, a sudden surge of input
`traffic may
`overwhelm a router, exceeding its storage capacity and causing it to lose packets.
`In contrast, with circuit switching, when packet switching is used,
`it
`is
`straightforward for the routers to provide speed and code conversion. Also, they
`can provide error correction to some extent.
`In some packet—switched networks,
`however, packets may be delivered in the wrong order tothe destination. Reor-
`dering of packets can never happen with circuit switching.
`Another difference is that circuit switching is completely transparent. The
`sender and receiver can use any bit rate, format, or framing method they want to.
`The carrier does not know or care. With packet switching, the carrier determines
`the basic parameters. A rough analogy is a road versus a railroad.
`In the former,
`the user determines the size, ‘speed, and nature of the vehicle; in the latter, the car-
`rier does. It is this transparency that allows voice, data, and fax to coexist within
`the phone system.
`A final difference between circuit and packet switching is the charging algo-
`rithm. Packet carriers usually base their charge on both the number of bytes (or
`
`APPLE 1015 - Page 6
`
`APPLE 1015 - Page 6
`
`

`
`134
`
`THE PHYSICAL LAYER
`
`CHAP. 2
`
`packets) carried and the connect time. Furthermore, transmission distance usually
`does not matter, except perhaps internationally. With circuit switching,
`the
`charge is based on the distance and time only, not the traffic. The differences are
`summarized in Fig. 2~36.
`
`Item
`
`Circuit—switched
`
`Packet-switched
`
`“‘r—"
`
`FDedlcated “copper” path
`Bandwidth available
`Potentially wasted bandwidth
`
`Store-and-forward transmission
`
`Each packet follows the same route
`
`Yes
`Fixed
`Yes
`
`No
`
`Yes
`
`No
`Dynamic
`No
`
`Yes
`
`No
`
`-1
`
`Call setup
`When can congestion occur
`Charging
`
`K‘
`
`Required
`At setup time
`Per minute
`
`Not needed
`On every packet
`Per packet
`
`1
`
`Fig. 2-36. A comparison of circuit—switched and packet-switched networks.
`
`Both circuit switching and packet switching are so important, we will come
`back to them shortly and describe the various technologies used in detail.
`
`The Switch Hierarchy
`
`It is worth saying a few words about how the routing between switches is
`done within the current circuit—switched telephone system. We will describe the
`AT&T system here, but other companies and countries use the same general prin-
`ciples. The telephone system has five classes of switching offices, as illustrated
`in Fig. 2-37. There are 10 regional switching offices, and these are fully intercon-
`nected by 45 high—bandwidth fiber optic trunks. Below the regional offices are 67
`sectional offices, 230 primary offices, 1300 toll offices, and 19,000 end offices.
`The lower four levels were originally connected as a tree.
`Calls are generally connected at the lowest possible level. Thus if a sub-
`scriber connected to end office 1 calls another subscriber connected to end office
`
`1, the call will be completed in that office. However, a call from a customer
`attached to end office 1 in Fig. 2-37 to a customer attached to end office 2 will
`have to go toll office 1. However, a call from end office 1 to end office 4 will
`have to go up to primary office 1, and so on. With a pure tree, there is only one
`minimal route, and that would normally be taken.
`During years of operation, the telephone companies noticed that some routes
`were busier than others. For example, there were many calls from New York to
`Los Angeles. Rather than go all the way up the hierarchy, they simply installed
`direct trunks for the busy routes. A few of these are shown in Fig. 2-37 as
`
`APPLE 1015 - Page 7
`
`APPLE 1015 - Page 7
`
`

`
`SEC. 2.4
`
`THE TELEPHONE SYSTEM
`
`135
`
`G,/—
`
`_=-'—-'‘‘ > —-"-S‘ __‘''-~‘:;
`
`\ 4’
`‘u.
`/*
`§~'.£4z..*...,y/-‘ ;s'§,'«{
`4'” ¢I.'D¢‘‘= *-
`$23}
`
`10 Regional
`offices
`
`t d)
`
`“i"'y
`
`in erconnec e
`
`I
`
`’II
`
`\‘
`
`s\
`
`,
`
`'1
`
`67 Sectional
`offices
`
`00 0 OOCO@@@
`
`\
`
`\
`,
`: X‘
`X ‘.
`I
`\\
`‘\
`‘,
`I
`\\
`,
`\
`\ \
`\\\\
`'
`\
`\\
`CCCCC CCCCCCCCC COCO ‘9'°°°E"“'
`offices
`
`offices
`
`1300 Toll
`offices
`
`1
`
`2
`
`3
`
`4
`
`5
`
`200 Miliionteiephones_
`
`19,000
`
`Fig. 2-37. The AT&T telephone hierarchy. The dashed lines are direct trunks.
`
`dashed lines. As a consequence, many calls can now be routed along many paths.
`The actual route chosen is generally the most direct one, but if the necessary
`trunks along it are full, an alternative is chosen. This complex routing is now pos-
`sible because a switching machine, like the AT&T 5 ESS, is in fact just a general
`purpose computer with a large amount of very specialized I/O equipment.
`
`Crossbar Switches
`
`Let us now turn from how calls are routed among switches to how individual
`switches actually work inside. Several kinds of switches are (or were) common
`within the telephone system. The simplest kind is the crossbar switch (also
`called a crosspoint switch), shown in Fig. 2-38.
`In a switch with n input lines
`and n output
`lines (i.e.,
`11
`full duplex lines),
`the crossbar switch has n2
`
`APPLE 1015 - Page 8
`
`APPLE 1015 - Page 8
`
`

`
`136
`
`THE PHYSICAL LAYER
`
`CHAP. 2
`
`intersections, called crosspoints, where an input and an outputline may be con—
`nected by a semiconductor switch, as shown in Fig. 2—38(a).
`In Fig. 2—38(b) we
`see an example in which line 0 is connected to line 4, line 1 is connected to line 7,
`and line 2 is connected to line 6. Lines 3 and 5 are not connected. All the bits
`
`that arrive at the switch from line 4, for example, are immediately sent out of the
`switch on line 0. Thus the crossbar switch implements circuit switching by mak-
`ing a direct electrical connection, just like the jumper cables in the first—generation
`switches, only automatically and within microseconds.
`
`Potential connection
`
`Actual connection
`
`inputs
`
`(b) A crossbar switch
`Fig. 2-38. (a) A crossbar switch with no connections.
`with three connections set up: 0 with 4, l with 7, and 2 with 6.
`
`The problem with a crossbar switch is that the number of crossbars grows as
`the square of the number of lines into the switch.
`If we assume that all lines are
`full duplex and that there are no self—connections, only the crosspoints above the
`diagonal are needed. Still, n(rz —- 1)/2 crosspoints are needed. For n = 1000, we
`need 499,500 crosspoints. While building a VLSI chip with this number of
`transistor switches is possible, having 1000 pins on the chip is not. Thus a single
`crossbar switch is only useful for relatively small end offices.
`
`Space Division Switches
`
`By splitting the crossbar switch into small chunks and interconnecting them, it
`is possible to build multistage switches with many fewer_crosspoints. These are
`called space division switches. Two configurations are illustrated in Fig. 2—39.
`To keep our example simple, we will consider only three—stage switches, but
`
`APPLE 1015 - Page 9
`
`APPLE 1015 - Page 9
`
`

`
`SEC. 2.4
`
`THE TELEPHONE SYSTEM
`
`137
`
`N=16,n=4,k=2
`
`N=16,n=4,k=3
`
`N
`-5- Crossbars
`
`N
`—n- Crossbars
`
`N
`F Crossbars
`
`N
`-n— Crossbars
`
`
`
`Fig. 2-39. Two space division switches with different parameters.
`
`switches with more stages are also possible. In these examples, we have a total of
`N inputs and N outputs. Instead of building a single N X N crossbar, we build the
`switch out of smaller rectangular Crossbars.
`In the first stage, each crossbar has n
`inputs, so we need N/n of them to handle all N incoming lines.
`The second stage has k Crossbars, each with N/n inputs and N/n outputs. The
`third stage is a repeat of the first stage, but reversed left to right. Each intermedi-
`ate crossbar is connected to each input crossbar and each output crossbar. Conse-
`quently, it is possible to connect every input to every output using either the first
`intermediate crossbar in Fig. 2—39(a) or using the second one.
`In fact, there are
`two disjoint paths from each input to each output, depending which intermediate
`crossbar is chosen. In Fig. 2—39(b) there are three paths for each input/output pair.
`With k intermediate stages (k is a design parameter), there are k disjoint paths.
`Let us now compute the number of crosspoints needed for a three—stage
`switch.
`In the first stage, there are N/n Crossbars, each with nk crosspoints, for a
`total of Nk.
`In the second stage,
`there are k Crossbars, each with (N/n)2
`crosspoints. The third stage is the same as the first. Adding up the three stages,
`we get
`
`Number of crosspoints = 2kN + k(N/n)2
`
`For N = 1000, n = 50 and k = 10, we need only 24,000 crosspoints instead of the
`499,500 required by a 1000 X 1000 single-stage crossbar.
`
`APPLE 1015 - Page 10
`
`
`
`.ss:_<wzrz«:w-.<~w.~».:,..
`
`APPLE 1015 - Page 10
`
`

`
`138
`
`THE PHYSICAL LAYER
`
`CHAP. 2
`
`Unfortunately, as usual, there is no free lunch. The switch can block. Con-
`sider Fig. 2—39(a) again. Stage 2 has eight inputs, so a maximum of eight calls
`can be connected at once. When call nine comes by, it will have to get a busy sig-
`nal, even though the destination is available. The switch of Fig. 2—39(b) is better,
`handling a maximum of 12 calls instead of 8, but it uses more crosspoints. Some-
`times when making a phone call you may have gotten a busy signal before you
`finished dialing. This was probably caused by blocking part way through the net-
`work.
`
`It should be obvious that the larger k is, the more expensive the switch and the
`lower the blocking probability.
`In 1953, Clos showed that when k = 2n — 1, the
`switch will never block (Clos, 1953). Other researchers have analyzed calling
`patterns in great detail to construct switches that theoretically can block but do so
`only rarely in practice.
`
`Time Division Switches
`
`A completely different kind of switch is the time division switch, shown in
`Fig. 240. With time division switching, the n input lines are scanned in sequence
`to build up an input frame with n slots. Each slot has k bits. For Tl switches, the
`slots are 8 bits, with 8000 frames processed per second.
`
`Time slot
`
`interchanger
`
`.
`
`n output
`lines
`
`Output frame
`
`n word mapping table
`
`
`
`
`Time slots
`RAM buffer of
`n k-bit words
`
`C
`
`Fig. 2-40. A time division switch.
`
`The heart of the time division switch is the time slot interchanger, which
`accepts input frames and produces output frames in which the time slots have
`been reordered.
`In Fig. 2-40, input slot 4 is output first, then slot 7, and so on.
`Finally, the output frame is demultiplexed, with output slot 0 (input slot 4) going
`
`APPLE 1015 - Page 11
`
`APPLE 1015 - Page 11
`
`

`
`
`
`SEC. 2.4
`
`THE TELEPHONE SYSTEM
`
`139
`
`to line 0, and so on. In essence, the switch has moved a byte from input line 4 to
`output line 0, another byte from input line 7 to output line 1, and so on. Viewed
`from the outside, the whole arrangement is a circuit switch, even though there are
`no physical connections.
`The time slot interchanger works as follows: When an input frame is ready to
`be processed, each slot (i.e., each byte in the input frame) is written into a RAM
`buffer inside the interchanger. The slots are written in order, so buffer word 1‘
`contains slot i.
`
`After all the slots of the input frame have been stored in the buffer, the output
`frame is constructed by reading out the words again, but in a different order. A
`counter goes from 0 to n —— 1. At step j, the contents of word j of a mapping table
`is read out and used to address the RAM table. Thus if word 0 of the mapping
`table contains a 4, word 4 of the RAM buffer will be read out first, and the first
`slot of the output frame will be slot 4 of the input frame. Thus the contents of the
`mapping table determine which permutation of the input frame will be generated
`as the output frame, and thus which input line is connected to which output line.
`Time division switches use tables that are linear in the number of lines, rather
`than quadratic, but they have another limitation.
`It is necessary to store it slots in
`the buffer RAM and then read them out again within one frame period of 125
`usec. If each of these memory accesses takes T microsec, the time needed to pro-
`cess a frame is 2nT microsec, so we have 2nT = 125 or n = 125/2T. For a
`memory with lO0—nsec cycle time, we can support at most 625 lines. We can also
`turn this relation around and use it to determine the required memory cycle to sup—
`port a given number of lines. As with a crossbar switch, it is possible to devise
`multistage switches that split the work up into several parts and then combine the
`results in order to handle larger numbers of lines.
`\.
`
`2.5. NARROWBAND ISDN
`
`For more than a century, the primary international telecommunication infras-
`tructure has been the public circuit—switched telephone system. This system was
`designed for analog voice transmission and is inadequate for modern communica-
`tion needs. Anticipating considerable user demand for an end-to~end digital ser-
`vice (i.e., not like Fig. 2-17 which is part digital and part analog), the world’s tele-
`phone companies and PTTs got together in 1984 under the auspices of CCITT and
`agreed to build a new, fully digital, circuit—switched telephone system by the early
`part of the 21st Century. This new system, called ISDN (Integrated Services
`Digital Network), has as its primary goal the integration of voice and nonvoice
`services.
`It is already available in many locations and its use is growing slowly.
`In the following sections we will describe what it does and how it works. For
`further information, see (Dagdeviren et al., 1994; and Kessler, 1993).
`
`APPLE 1015 - Page 12
`
`APPLE 1015 - Page 12
`
`

`
`140
`
`THE PHYSICAL LAYER
`
`CHAP. 2
`
`2.5.1. ISDN Services
`
`The key ISDN service will continue to be voice, although many enhanced
`features will be added. For example, many corporate managers have an intercom
`button on their telephone that rings their secretaries instantly (no call setup time).
`One ISDN feature is telephones with multiple buttons for instant call setup to
`arbitrary telephones anywhere in the world. Another feature is telephones that
`display the caller’s telephone number, name, and address on a display while ring-
`ing. A more sophisticated version of this feature allows the telephone to be con-
`nected to a computer, so that the caller’s database record is displayed on the
`screen as the call comes in. For example, a stockbroker could arrange that when
`she answers the telephone, the caller’s portfolio is already on the screen along
`with the current prices of all the caller’s stocks. Other advanced voice services
`include call forwarding and conference calls worldwide.
`Advanced nonvoice services are“remote electricity meter reading, and on—line
`medical, burglar, and smoke alarms that automatically call the hospital, police, or
`fire department, respectively, and give their address to speed up response.
`
`2.5.2. ISDN System Architecture
`
`It is now time to look at the ISDN architecture in detail, particularly the
`customer’s equipment and the interface between the customer and the telephone
`company or PTT. The key idea behind ISDN is that of the digital bit pipe, a con-
`ceptual pipe between the customer and the carrier through which bits flow.
`Whether the bits originated from a digital telephone, a digital terminal, a digital
`facsimile machine, or some other device is irrelevant. All that matters is that bits
`
`can flow through the pipe in both directions.
`The digital bit pipe can, and normally does, support multiple independent
`channels by time division multiplexing of the bit stream. The exact format of the
`bit stream and its multiplexing is a carefully defined part of the interface specifi-
`cation for the digital bit pipe. Two principal standards for the bit pipe have been
`developed, a low bandwidth standard for home use and a higher bandwidth stan-
`dard for business use that supports multiple channels that are identical to the home
`use channel. Furthermore, businesses may have multiple bit pipes if they need
`additional capacity beyond what the standard business pipe can provide.
`In Fig. 2—4l(a) we see the normal configuration for a home or small business.
`The carrier places a network terminating device, NT1, on the customer’s premises
`and connects it to the ISDN exchange in the carrier’s office, several kilometers
`away, using the twisted pair that was previously used to connect to the telephone.
`The NTl box has a connector on it into which a passive bus cable can be inserted.
`Up to eight ISDN telephones, terminals, alarms, and other devices can be con-
`nected to the cable, similar to the way devices are connected to a LAN. From the
`customer’s point of view, the network boundary is the connector on NTl.
`
`APPLE 1015 - Page 13
`
`APPLE 1015 - Page 13
`
`

`
`SEC. 2.5
`
`NARROWBAND ISDN
`
`141
`
`
`Customer's office
`<————— Carrier's office —-——>
`I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I
`I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I
`
`‘I
`
`ISDN Terminal
`
`lI
`
`ISDN
`T I
`Exchange
`
`;
`;
`ISDN
`ISDN
`ISDN
`:
`:
`Alarm
`Terminal
`Telephone
`._ ________________________________ ___J
`I______________________________________ _.._J
`
`
`
`-I—-0 internal
`“9IW°"<
`
`Customer's equipment
`
`Carrier's equipment
`
`(a)
`
`
`
`I
`
`Customer's office
`
`> <—
`
`Carrier's office
`
`
`
`T '
`I
`
`I
`|
`
`.
`
`" ““““““““““““““““““““““““ “I
`.******************************* "‘
`I
`I
`TE1 T
`I
`I
`I
`I
`TefITSII?I;\I H S
`I
`5 j E
`\’ \
`E
`I T: - U
`T°.
`,
`I
`I
`I
`carriers I
`N-I-2 jj 1- Exchan e mo internal
`S
`:
`-E...-1 / ISDN
`'
`9
`network:
`II
`I
`..I
`
`ISDN
`
`____________________________________ __
`Digital
`Carrier's equipment
`bit pipe
`
`A
`
`1
`
`ISDN
`
`Telephone
`
`,
`3
`I
`I j
`=15)“
`53% 1| ;
`'
`ISDN
`Terminal
`LAN Gateway
`I
`'
`
`I
`
`I E
`
`0 0 0 0
`
`EI
`
`__________________________________ _-J
`
`Customer's equipment
`
`(b)
`
`Fig. 2-41. (21) Example ISDN system for home use. (b) Example ISDN system
`with a PBX for use in large businesses.
`
`APPLE 1015 - Page 14
`
`APPLE 1015 - Page 14
`
`

`
`142
`
`THE PHYSICAL LAYER
`
`CHAP. 2
`
`the model of Fig. 2-4l(a) is inadequate because it is
`For large businesses,
`common to have more telephone conversations going on simultaneously than the
`bus can handle. Therefore, the model of Fig. 2-4l(b) is used.
`In this model we
`find a device, NT2, called a PBX (Private Branch eXchange), connected to NTl
`and providing the real interface for telephones, terminals and other equipment.
`An ISDN PBX is not very different conceptually from an ISDN switch, although
`it is usually smaller and cannot handle as many conversations at the same time.
`CCITT defined four reference points, called R, S, T, and U, between the
`various devices. These are marked in Fig. 2-41. The U reference point is the con-
`nection between the ISDN exchange in the carrier’s office and NTl. At present it
`is a two-wire copper twisted pair, but at some time in the future it may be
`replaced by fiber optics. The T reference point is what the connector on NTl pro-
`vides to the customer. The S reference point is the interface between the ISDN
`PBX and the ISDN terminals. The R reference point is the connection between
`the terminal adapter and non-lSDN‘<=~terminals. Many different kinds of interfaces
`will be used at R.
`
`2.5.3. The ISDN Interface
`
`The ISDN bit ‘pipe supports multiple channels interleaved by time division
`multiplexing. Several channel types have been standardized:
`
`A — 4-kHz analog telephone channel
`B — 64—kbps digital PCM channel for voice or data
`C — 8- or-16 kbps digital channel
`D - 16-kbps digital channel for out-of-band signaling
`E — 64—kbps digital channel for internal ISDN signaling
`H — 384-, 1536-, or l920-kbps digital channel
`
`It was not CClTT’s intention to allow an arbitrary combination of channels on the
`digital bit pipe. Three combinations have been standardized so far:
`
`1. Basic rate: 2B + lD
`
`2. Primary rate: 23B + lD (U.S. and Japan) or 30B + lD (Europe)
`
`3. Hybrid: 1A + lC
`
`The basic rate and primary rate channels are illustrated in Fig. 2-42.
`The basic rate should be viewed as a replacement for POTS (Plain Old Tele-
`phone Service) for home or small business use. Each of the 64—kbps B channels
`can handle a single PCM voice channel with 8-bit samples made 8000 timesea
`second (note that 64 kbps means 64,000 here, not 65,536). Signaling is on a
`separate 16-kbps D channel, so the full 64 kbps are available to the user (as in the
`CCITT 2.048-Mbps system and unlike the U.S. and Japanese Tl
`system).
`
`APPLE 1015 - Page 15
`
`APPLE 1015 - Page 15
`
`

`
`SEC. 2.5
`
`NARROWBAND ISDN
`
`143
`
`Q
`.1 Basic rate
`...:.
`
`(8)
`
`: D (16 kbps)
`_ B1 to B2
`
`Fig. 2-42. (a) Basic rate digital pipe. (b) Primary rate digital pipe.
`
`Because ISDN is so focused on 64—kbps channels, we refer to it as N-ISDN (Nar-
`rowband ISDN),
`to contrast it with broadband ISDN (ATM) to be discussed
`later.
`
`The primary rate interface is intended for use at the T reference point for
`businesses with a PBX.
`It has 23 B channels and l D channel (at 64 kbps) in the
`United States and Japan and 30 B channels and l D channel (at 64 kbps) in
`Europe. The 23B + lD choice was made to allow an ISDN frame fit nicely on
`AT&T’s T1 system. The 30B + ID choice was made to allow an ISDN frame fit
`nicely in CCITT’s 2.048 Mbps system. The 32nd time slot in the CCITT system
`is used for framing and general network maintenance. Note that the amount of D
`channel per B channel in the primary rate is much less than in the basic rate, as it
`is not expected that there will be much telemetry or low bandwidth packet data
`there.
`
`2.5.4. Perspective on N-ISDN
`
`N-ISDN was a massive attempt to replac,e_ the analog telephone system with a
`digital one suitable for both voice and nonvoice traffic. Achieving worldwide
`agreement on the interface standard for the basic rate was supposed to lead to a
`large user demand for ISDN equipment,
`thus leading to mass production,
`economies of scale, and inexpensive VLSI ISDN chips. Unfortunately,
`the
`standardization process took years and the technology in this area moved very
`rapidly, so that once the standard was finally agreed upon, it was obsolete.
`For home use, the largest demand for new services will undoubtedly be for
`video on demand. Unfortunately,
`the ISDN basic rate lacks the necessary
`bandwidth by two orders of magnitude. For business use, the situation is even
`bleaker. Currently available LANS offer at least 10 Mbps and are now being
`replaced by lOO—Mbps LANS. Offering 64—kbps service to businesses in the 1980s
`was a serious proposition. In the 1990s, it is a joke.
`Oddly enough, ISDN may yet be saved, but by a totally unexpected applica—
`tion: Internet access. Various companies now sell ISDN adaptors that combine
`the 2B + D channels into a single 144-kbps digital channel. Many Internet service
`providers also support these adaptors. The result is that people can access the
`
`APPLE 1015 - Page 16
`
`APPLE 1015 - Page 16
`
`

`
`416
`
`THE NETWORK LAYER
`
`CHAP. 5
`
`routers on the way. Normally, this option would only provide a few routers, to
`force a particular path. For example, to force a packet from London to Sydney to
`go west instead of east,
`this option might specify routers in New York, Los
`Angeles, and Honolulu. This option is most useful when political or economic
`considerations dictate passing through or avoiding certain countries.
`The Record route option tells the routers along the path to append their IP
`address to the option field. This allows system managers to track down bugs in
`the routing algorithms (“Why are packets from Houston to Dallas all Visiting
`Tokyo first?”) When the ARPANET was first set up, no packet ever passed
`through more than nine routers, so 40 bytes of option was ample. As mentioned
`above, now it is too small.
`Finally, the Timestamp option is like the Record route option, except that in
`addition to recording its 32-bit IP address, each router also records a 32-bit time-
`stamp. This option, too, is mostly for debugging routing algorithms.
`“V
`
`5.5.2. IP Addresses
`
`Every host and router on the Internet has an IP address, which encodes its net—
`work number and host number. The combination is unique: no two machines
`have the same IP address. All IP addresses are 32 bits long and are used in the
`Source address and Destination address fields of IP packets. The formats used
`for IP address