LAN SYSTEM

LAN SYSTEM

The medium access control technique and topology

are key characteristics used in the classification of LANs and in the

development of standards. The following systems are discussed in this lesson:'

Ethernet and Fast Ethernet (CSMA/CD)

Token RingIFDDI

100VG-AnyLAN

ATM LANs

Fibre Channel

Wireless LANs

Ethernet/Fast ethernet(CSMA/CD)

The most commonly used medium access control technique for busltree and star

topologies is carrier-sense multiple access with collision detection (CSMAICD).

The original baseband version of this technique was developed by Xerox as part of

the Ethernet LAN. The original broadband version was developed by MITRE as

part of its MITREnet LAN. All of this work formed the basis for the IEEE 802.3

standard.

In this lesson, we will focus on the IEEE 802.3 standard. As with other LAN

standards, there is both a medium access control layer and a physical layer, which

are considered in turn in what follows.

IEEE 802.3 Medium Access Control

It is easier to understand the operation of CSMAICD if we look first at some earlier

schemes from which CSMAICD evolved.

Precursors

CSMAICD and its precursors can be termed random access, or contention, techniques.

They are random access in the sense that there is no predictable or scheduled

time for any station to transmit; station transmissions are ordered randomly.

They exhibit contention in the sense that stations contend for time on the medium.

The earliest of these techniques, known as ALOHA, was developed for

packet radio networks. However, it is applicable to any shared transmission

medium. ALOHA, or pure ALOHA as it is sometimes called, is a true free-for-all.

Whenever a station has a frame to send, it does so. The station then listens for an

amount of time equal to the maximum possible round-trip propagation delay on the

network (twice the time it takes to send a frame between the two most widely separated

stations) plus a small fixed time increment. If the station hears an acknowl-

edgment during that time, fine; otherwise, it resends the frame. If the station fails to

receive an acknowledgment after repeated transmissions, it gives up. A receiving

station determines the correctness of an incoming frame by examining a framecheck-

sequence field, as in HDLC. If the frame is valid and if the destination

address in the frame header matches the receiver's address, the station immediately

sends an acknowledgment. The frame may be invalid due to noise on the channel

or because another station transmitted a frame at about the same time. In the latter

case, the two frames may interfere with each other at the receiver so that neither

gets through; this is known as a collision. If a received frame is determined to be

invalid, the receiving station simply ignores the frame.

ALOHA is as simple as can be, and pays a penalty for it. Because the number

of collisions rise rapidly with increased load, the maximum utilization of the channel

is only about 18% (see [STAL97]).

To improve efiiciency, a modification of ALOHA, known as slotted ALOHA,

was developed. In this scheme, time on the channel is organized into uniform slots

whose size equals the frame transmission time. Some central clock or other technique

is needed to synchronize all stations. Transmission is permitted to begin only

at a slot boundary. Thus, frames that do overlap will do so totally. This increases the

maximum utilization of the system to about 37%.

Both ALOHA and slotted ALOHA exhibit poor utilization. Both fail to take

advantage of one of the key properties of both packet radio and LANs, which is that

propagation delay between stations is usually very small compared to frame transmission

time. Consider the following observations. If the station-to-station propagation

time is large compared to the frame transmission time, then, after a station

launches a frame, it will be a long time before other stations know about it. During

that time, one of the other stations may transmit a frame; the two frames may interfere

with each other and neither gets through. Indeed, if the distances are great

enough, many stations may begin transmitting, one after the other, and none of

their frames get through unscathed. Suppose, however, that the propagation time is

small compared to frame transmission time. In that case, when a station launches a

frame, all the other stations know it almost immediately. So, if they had any sense,

they would not try transmitting until the first station was done. Collisions would be

rare because they would occur only when two stations began to transmit almost

simultaneously. Another way to look at it is that a short delay time provides the stations

with better feedback about the state of the network; this information can be

used to improve efficiency.

The foregoing observations led to the development of carrier-sense multiple

access (CSMA). With CSMA, a station wishing to transmit first listens to the

medium to determine if another transmission is in progress (carrier sense). If the

medium is in use, the station must wait. If the medium is idle, the station may transmit.

It may happen that two or more stations attempt to transmit at about the same

time. If this happens, there will be a collision; the data from both transmissions will

be garbled and not received successfully. To account for this, a station waits a reasonable

amount of time, after transmitting, for an acknowledgment, taking into

account the maximum round-trip propagation delay, and the fact that the acknowledging

station must also contend for the channel in order to respond. If there is no

acknowledgment, the station assumes that a collision has occurred and retransmits.

One can see how this strategy would be effective for networks in which the

average frame transmission time is much longer than the propagation time. Collisions

can occur only when more than one user begins transmitting within a short

time (the period of the propagation delay). If a station begins to transmit a frame,

and there are no collisions during the time it takes for the leading edge of the packet

to propagate to the farthest station, then there will be no collision for this frame

because all other stations are now aware of the transmission.

The maximum utilization achievable using CSMA can far exceed that of

ALOHA or slotted ALOHA. The maximum utilization depends on the length of

the frame and on the propagation time; the longer the frames or the shorter the

propagation time, the higher the utilization. This subject is explored in Appendix

13A.

With CSMA, an algorithm is needed to specify what a station should do if the

medium is found busy. The most common approach, and the one used in IEEE

802.3, is the 1-persistent technique. A station wishing to transmit listens to the

medium and obeys the following rules:

I. If the medium is idle, transmit; otherwise, go to step 2.

2. If the medium is busy, continue to listen until the channel is sensed idle; then

transmit immediately.

If two or more stations are waiting to transmit, a collision is guaranteed.

Things get sorted out only after the collision.

Description of CSMAICD

CSMA, although more efficient than ALOHA or slotted ALOHA, still has one

glaring inefficiency: When two frames collide, the medium remains unusable for the

duration of transmission of both damaged frames. For long frames, compared to

propagation time, the amount of wasted capacity can be considerable. This waste

can be reduced if a station continues to listen to the medium while transmitting.

This leads to the following rules for CSMAICD:

1. If the medium is idle, transmit; otherwise, go to step 2.

2. If the medium is busy, continue to listen until the channel is idle, then transmit

immediately.

3. If a collision is detected during transmission, transmit a brief jamming signal

to assure that all stations know that there has been a collision and then cease

transmission.

4. After transmitting the jamming signal, wait a random amount of time, then

attempt to transmit again. (Repeat from step 1.)

Figure 13.1 illustrates the technique for a baseband bus. At time to, station A

begins transmitting a packet addressed to D. At tl, both B and C are ready to transmit.

B senses a transmission and so defers. C, however, is still unaware of A's transmission

and begins its own transmission. When A's transmission reaches C, at t2, C

detects the collision and ceases transmission. The effect of the collision propagates

back to A, where it is detected some time later, t3, at which time A ceases transmission.

With CSMAICD, the amount of wasted capacity is reduced to the time it takes

to detect a collision. Question: how long does that take? Let us consider first the

case of a baseband bus and consider two stations as far apart as possible. For example,

in Figure 13.1, suppose that station A begins a transmission and that just before

that transmission reaches D, D is ready to transmit. Because D is not yet aware of

A's transmission, it begins to transmit. A collision occurs almost immediately and is

recognized by D. However, the collision must propagate all the way back to A

before A is aware of the collision. By this line of reasoning, we conclude that the

amount of time that it takes to detect a collision is no greater than twice the end-toend

propagation delay. For a broadband bus, the delay is even longer. Figure 13.2

shows a dual-cable system. This time, the worst case occurs for two stations as close

together as possible and as far as possible from the headend. In this case, the maximum

time to detect a collision is four times the propagation delay from an end of

the cable to the headend.

An important rule followed in most CSMAICD systems, including the IEEE

standard, is that frames should be long enough to allow collision detection prior to

the end of transmission. If shorter frames are used, then collision detection does not

occur, and CSMAKD exhibits the same performance as the less efficient CSMA

protocol.

Although the implementation of CSMAICD is substantially the same for

baseband and broadband, there are differences. One is the means for performing

carrier sense; for baseband systems, this is done by detecting a voltage pulse train.

For broadband, the RF carrier is detected.

Collision detection also differs for the two systems. For baseband, a collision

should produce substantially higher voltage swings than those produced by a single

transmitter. Accordingly, the IEEE standard dictates that the transmitter will

detect a collision if the signal on the cable at the transmitter tap point exceeds the

maximum that could be produced by the transmitter alone. Because a transmitted

signal attenuates as it propagates, there is a potential problem: If two stations far

apart are transmitting, each station will receive a greatly attenuated signal from the

other. The signal strength could be so small that when it is added to the transmitted

signal at the transmitter tap point, the combined signal does not exceed the CD

threshold. For this reason, among others, the IEEE standard restricts the maximum

length of coaxial cable to 500 m for 10BASE5 and to 200 m for 10BASE2.

A much simpler collision detection scheme is possible with the twisted pair

star-topology approach (Figure 12.13). In this case, collision detection is based on

logic rather than on sensing voltage magnitudes. For any hub, if there is activity

(signal) on more than one input, a collision is assumed. A special signal called the

collision presence signal is generated. This signal is generated and sent out as long

as activity is sensed on any of the input lines. This signal is interpreted by every

node as an occurrence of a collision.

There are several possible approaches to collision detection in broadband systems.

The most common of these is to perform a bit-by-bit comparison between

transmitted and received data. When a station transmits on the inbound channel, it

begins to receive its own transmission on the outbound channel after a propagation

delay to the headend and back. Note the similarity to a satellite link. Another

approach, for split systems, is for the headend to perform detection based on garbled

data.

MAC Frame

Figure 13.3 depicts the frame format for the 802.3 protocol; it consists of the following

fields:

Preamble. A 7-octet pattern of alternating 0s and 1s used by the receiver to

establish bit synchronization.

Start frame delimiter. The sequence 10101011, which indicates the actual start

of the frame and which enables the receiver to locate the first bit of the rest

of the frame.

Destination address (DA). Specifies the station(s) for which the frame is

intended. It may be a unique physical address, a group address, or a global

address. The choice of a 16- or 48-bit address length is an implementation

decision, and must be the same for all stations on a particular LAN.

Source address (SA). Specifies the station that sent the frame.

Length. Length of the LLC data field.

LLC data. Data unit supplied by LLC.

Pad. Octets added to ensure that the frame is long enough for proper CD

operation.

Frame check sequence (FCS). A 32-bit cyclic redundancy check, based on all

fields except the preamble, the SFD, and the FCS.

IEEE 802.3 10-Mbgs Specifications (Ethernet)

The IEEE 802.3 committee has been the most active in defining alternative physical

configurations; this is both good and bad. On the good side, the standard has

been responsive to evolving technology. On the bad side, the customer, not to mention

the potential vendor, is faced with a bewildering array of options. However, the

committee has been at pains to ensure that the various options can be easily integrated

into a configuration that satisfies a variety of needs. Thus, the user that has

a complex set of requirements may find the flexibility and variety of the 802.3 standard

to be an asset.

To distinguish among the various implementations that are available, the

committee has developed a concise notation:

<data rate in Mbps><signaling method><maximum segment length in hundreds of

meters>

The defined alternatives are:

Note that 10BASE-T and 10-BASE-F do not quite follow the notation; "T'

stands for twisted pair, and "F" stands for optical fiber. Table 13.1 summarizes these

options. All of the alternatives listed in the table specify a data rate of 10 Mbps. In

addition to these alternatives, there are several versions that operate at 100 Mbps;

these are covered later in this lesson.

1OBASES Medium Specification

10BASE5 is the original 802.3 medium specification and is based on directly on Ethernet.

10BASE5 specifies the use of 50-ohm coaxial cable and uses Manchester digital

signaling.3 The maximum length of a cable segment is set at 500 meters. The

length of the network can be extended by the use of repeaters, which are transparent

to the MAC level; as they do no buffering, they do not isolate one segment from

another. So, for example, if two stations on different segments attempt to transmit

at the same time, their transmissions will collide. To avoid looping, only one path of

segments and repeaters is allowed between any two stations. The standard allows a

maximum of four repeaters in the path between any two stations, thereby extending

the effective length of the medium to 2.5 kilometers.

l0BASE2 Medium Specification

To provide a lower-cost system than 10BASE5 for personal computer LANs,

10BASE2 was added. As with 10BASE5, this specification uses 50-ohm coaxial

cable and Manchester signaling. The key difference is that 10BASE2 uses a thinner

cable, which supports fewer taps over a shorter distance than the 10BASE5 cable.

Because they have the same data rate, it is possible to combine 10BASE5 and

10BASE2 segments in the same network, by using a repeater that conforms to

10BASE5 on one side and 10BASE2 on the other side. The only restriction is that

a 10BASE2 segment should not be used to bridge two 10BASE5 segments, because

a "backbone" segment should be as resistant to noise as the segments it connects.

10BASE-T Medium Specification

By sacrificing some distance, it is possible to develop a 10-Mbps LAN using the

unshielded twisted pair medium. Such wire is often found prewired in office buildings

as excess telephone cable, and can be used for LANs. Such an approach is specified

in the 10BASE-T specification. The 10BASE-T specification defines a starshaped

topology. A simple system consists of a number of stations connected to a

central point, referred to as a multiport repeater, via two twisted pairs. The central

point accepts input on any one line and repeats it on all of the other lines.

Stations attach to the multiport repeater via a point-to-point link. Ordinarily,

the link consists of two unshielded twisted pairs. Because of the high data rate and

the poor transmission qualities of unshielded twisted pair, the length of a link is limited

to 100 meters. As an alternative, an optical fiber link may be used. In this case,

the maximum length is 500 m.

10BROAD36 Medium Specification

The 10BROAD36 specification is the only 802.3 specification for broadband. The

medium employed is the standard 75-ohm CATV coaxial cable. Either a dual-cable

or split-cable configuration is allowed. The maximum length of an individual segment,

emanating from the headend, is 1800 meters; this results in a maximum endto-

end span of 3600 meters.

The signaling on the cable is differential phase-shift keying (DPSK). In ordinary

PSK, a binary zero is represented by a carrier with a particular phase, and a

binary one is represent by a carrier with the opposite phase (180-degree difference).

DPSK makes use of differential encoding, in which a change of phase occurs when

a zero occurs, and there is no change of phase when a one occurs. The advantage of

differential encoding is that it is easier for the receiver to detect a change in phase

than to determine the phase itself.

The characteristics of the modulation process are specified so that the resulting

10 Mbps signal fits into a 14 MHz bandwidth.

10BASE-F Medium Specification

The 10BASE-F specification enables users to take advantage of the distance and

transmission characteristics available with the use of optical fiber. The standard

actually contains three specifications:

a 10-BASE-FP (passive). A passive-star topology for interconnecting stations

and repeaters with up to 1 km per segment.

a 10-BASE-FL (link). Defines a point-to-point link that can be used to connect

stations or repeaters at up to 2 km.

10-BASE-FB (backbone). Defines a point-to-point link that can be used to

connect repeaters at up to 2 km.

All three of these specifications make use of a pair of optical fibers for each

transmission link, one for transmission in each direction. In all cases, the signaling

scheme involves the use of Manchester encoding. Each Manchester signal element

is then converted to an optical signal element, with the presence of light corresponding

to high and the absence of light corresponding to low. Thus, a 10-Mbps

Manchester bit stream actually requires 20 Mbps on the fiber.

The 10-BASE-FP defines a passive star system that can support up to 33 stations

attached to a central passive star, of the type described in Lesson 3. 10-

BASE-FL and 10-BASE-FB define point-to-point connections that can be used to

extend the length of a network; the key difference between the two is that 10-

BASE-FB makes use of synchronous retransmission. With synchronous signaling,

an optical signal coming into a repeater is retimed with a local clock and retransmitted.

With conventional asynchronous signaling, used with 10-BASE-FL, no such

retiming takes place, so that any timing distortions are propagated through a series

of repeaters. As a result, 10BASE-FB can be used to cascade up to 15 repeaters in

sequence to achieve greater length.

IEEE 802.3 100-Mbps Specifications (Fast Ethernet)

Fast Ethernet refers to a set of specifications developed by the IEEE 802.3 committee

to provide a low-cost, Ethernet-compatible LAN operating at 100 Mbps. The

blanket designation for these standards is 100BASE-T. The committee defined a

number of alternatives to be used with different transmission media.

Figure 13.4 shows the terminology used in labeling the specifications and indicates

the media used. All of the 100BASE-T options use the IEEE 802.3 MAC protocol

and frame format. 100BASE-X refers to a set of options that use the physical

medium specifications originally defined for Fiber Distributed Data Interface

(FDDI; covered in the next lesson). All of the 100BASE-X schemes use two physical

links between nodes: one for transmission and one for reception. 100BASE-TX

makes use of shielded twisted pair (STP) or high-quality (Category 5) unshielded

twisted pair (UTP). 100BASE-FX uses optical fiber.

In many buildings, each of the 100BASE-X options requires the installation of

new cable. For such cases, 100BASE-T4 defines a lower-cost alternative that can

use Category-3, voice grade UTP in addition to the higher-quality Category 5 UTP.~

To achieve the 100-Mbps data rate over lower-quality cable, 100BASE-T4 dictates

the use of four twisted pair lines between nodes, with the data transmission making

use of three pairs in one direction at a time.

100 BASE-X

For all of the 100BASE-T options, the topology is similar to that of 10BASET,

namely a star-wire topology.

Table 13.2 summarizes key characteristics of the 100BASE-T options.

For all of the transmission media specified under 100BASE-X, a unidirectional

data rate of 100 Mbps is achieved by transmitting over a single link (single twisted

pair, single optical fiber). For all of these media, an efficient and effective signal

encoding scheme is required. The one chosen was originally defined for FDDI, and

can be referred to as 4Bl5B-NRZI. See Appendix 13A for a description.

The 100BASE-X designation includes two physical-medium specifications,

one for twisted pair, known as 100BASE-TX, and one for optical fiber, known as

100-BASE-FX.

100BASE-TX makes use of two pairs of twisted pair cable, one pair used for

transmission and one for reception. Both STP and Category 5 UTP are allowed.

The MTL-3 signaling scheme is used (described in Appendix 13A).

100BASE-FX makes use of two optical fiber cables, one for transmission and

one for reception. With 100BASE-FX, a means is needed to convert the 4Bl5BNRZI

code groups stream into optical signals. The technique used is known as

intensity modulation. A binary 1 is represented by a burst or pulse of light; a binary

0 is represented by either the absence of a light pulse or by a light pulse at very low

intensity.

100BASE-T4

100BASE-T4 is designed to produce a 100-Mbps data rate over lower-quality Category

3 cable, thus taking advantage of the large installed base of Category 3 cable

in office buildings. The specification also indicates that the use of Category 5 cable

is optional. 100BASE-T4 does not transmit a continuous signal between packets,

which makes it useful in battery-powered applications.

For 100BASE-T4 using voice-grade Category 3 cable, it is not reasonable to

expect to achieve 100 Mbps on a single twisted pair. Instead, 100BASE-T4 specifies

that the data stream to be transmitted is split up into three separate data streams,

each with an effective data rate of 33Mbps. Four twisted pairs are used. Data are

transmitted using three pairs and received using three pairs. Thus, two of the pairs

must be configured for bidirectional transmission.

As with 100BASE-X, a simple NRZ encoding scheme is not used for

100BASE-T4; this would require a signaling rate of 33 Mbps on each twisted pair

and does not provide synchronization. Instead, a ternary signaling scheme known as

8B6T is used .

TOKEN RING/FDDI

Token ring is the most commonly used MAC protocol for ring-topology LANs. In

this lesson, we examine two standard LANs that use token ring: IEEE 802.5 and

FDDI.

IEEE 802.5 Medium Access Control

MAC Protocol

The token ring technique is based on the use of a small frame, called a token, that

circulates when all stations are idle. A station wishing to transmit must wait until it

detects a token passing by. It then seizes the token by changing one bit in the token,

which transforms it from a token into a start-of-frame sequence for a data frame.

The station then appends and transmits the remainder of the fields needed to construct

a data frame.

When a station seizes a token and begins to transmit a data frame, there is no

token on the ring, so other stations wishing to transmit must wait. The frame on the

ring will make a round trip and be absorbed by the transmitting station. The transmitting

station will insert a new token on the ring when both of the following conditions

have been met:

The station has completed transmission of its frame.

The leading edge of the transmitted frame has returned (after a complete circulation

of the ring) to the station.

If the bit length of the ring is less than the frame length, the first condition

implies the second; if not, a station could release a free token after it has finished

transmitting but before it begins to receive its own transmission. The second condition

is not strictly necessary, and is relaxed under certain circumstances. The advantage

of imposing the second condition is that it ensures that only one data frame at

a time may be on the ring and that only one station at a time may be transmitting,

thereby simplifying error-recovery procedures.

Once the new token has been inserted on the ring, the next station downstream

with data to send will be able to seize the token and transmit. Figure 13.5

illustrates the technique. In the example, A sends a packet to C, which receives it

and then sends its own packets to A and D.

Nota that under lightly loaded conditions, there is some inefficiency with

token ring because a station must wait for the token to come around before transmitting.

However, under heavy loads, which is when it matters, the ring functions in

a round-robin fashion, which is both efficient and fair. To see this, consider the configuration

in Figure 13.5. After station A transmits, it releases a token. The first station

with an opportunity to transmit is D. If D transmits, it then releases a token and

C has the next opportunity, and so on.

The principal advantage of token ring is the flexible control over access that it

provides. In the simple scheme just described, the access if fair. As we shall see,

schemes can be used to regulate access to provide for priority and for guaranteed

bandwidth services.

The principal disadvantage of token ring is the requirement for token rnaintenance.

Loss of the token prevents further utilization of the ring. Duplication of the

token can also disrupt ring operation. One station must be selected as a monitor to

ensure that exactly one token is on the ring and to ensure that a free token is reinserted,

if necessary.

MAC Frame

Figure 13.6 depicts the frame format for the 802.5 protocol. It consists of the following

fields:

Starting delimiter (SD). Indicates start of frame. The SD consists of signaling

patterns that are distinguishable from data. It is coded as follows: JKOJKOOO,

where J and K are nondata symbols. The actual form of a nondata symbol

depends on the signal encoding on the medium.

Access control (AC). Has the format PPPTMRRR, where PPP and RRR are

3-bit priority and reservation variables, and M is the monitor bit; their use is

explained below. T indicates whether this is a token or data frame. In the case

of a token frame, the only remaining field is ED.

Frame control (FC). Indicates whether this is an LLC data frame. If not, bits 7

in this field control operation of the token ring MAC protocol.

Destination address (DA). As with 802.3.

Source address (SA). As with 802.3.

Data unit. Contains LLC data unit.

Frame check sequence (FCS). As with 802.3.

End delimiter (ED). Contains the error-detection bit (E), which is set if

any repeater detects an error, and the intermediate bit (I), which is used to

indicate that this is a frame other than the final one of a multiple-frame

transmission.

FCS

Frame status (FS). Contains the address recognized (A) and frame-copied

(C) bits, whose use is explained below. Because the A and C bits are outside

the scope of the FCS, they are duplicated to provide a redundancy check to

detect erroneous settings.

We can now restate the token ring algorithm for the case when a single priority

is used. In this case, the priority and reservation bits are set to 0. A station wishing

to transmit waits until a token goes by, as indicated by a token bit of 0 in the AC

field. The station seizes the token by setting the token bit to 1. The SD and AC

fields of the received token now function as the first two fields of the outgoing

frame. The station transmits one or more frames, continuing until either its supply

of frames is exhausted or a token-holding timer expires. When the AC field of the

last transmitted frame returns, the station sets the token bit to 0 and appends an ED

field, resulting in the insertion of a new token on the ring.

Stations in the receive mode listen to the ring. Each station can check passing

frames for errors and can set the E bit to 1 if an error is detected. If a station detects

its own MAC address, it sets the A bit to 1; it may also copy the frame, setting the

C bit to 1. This allows the originating station to differentiate three results of a frame

transmission:

e Destination station nonexistent or not active (A = 0, C = 0)

Destination station exists but frame not copied ( A = 1, C = 0)

Frame received (A = 1, C = 1)

Token Ring Priority

The 802.5 standard includes a specification for an optional priority mechanism.

Eight levels of priority are supported by providing two 3-bit fields in each data

frame and token: a priority field and a reservation field. To explain the algorithm,

let us define the following variables:

Pf = priority of frame to be transmitted by station

P, = service priority: priority of current token

Pr = value of P, as contained in the last token received by this station

R, = reservation value in current token

Rr = highest reservation value in the frames received by this station during

the last token rotation

The scheme works as follows:

1. A station wishing to transmit must wait for a token with P, 5 Pf.

2. While waiting, a station may reserve a future token at its priority level (Pf).

If a data frame goes by, and if the reservation field is less than its priority

(R, < Pf), then the station may set the reservation field of the frame to its

priority (R, t Pf). If a token frame goes by, and if (R, < Pf AND Pf < P.,),

then the station sets the reservation field of the frame to its priority (R, c Pf).

This setting has the effect of preempting any lower-priority reservation.

3. When a station seizes a token, it sets the token bit to 1 to start a data frame,

sets the reservation field of the data frame to 0, and leaves the priority field

unchanged (the same as that of the incoming token frame).

4. Following transmission of one or more data frames, a station issues a new

token with the priority and reservation fields set as indicated in Table 13.3.

The effect of the above steps is to sort the competing claims and to allow the

waiting transmission of highest priority to seize the token as soon as possible. A

moment's reflection reveals that, as stated, the algorithm has a ratchet effect on priority,

driving it to the highest used level and keeping it there. To avoid this, a station

that raises the priority (issues a token that has a higher priority than the token

that it received) has the responsibility of later lowering the priority to its previous

level. Therefore, a station that raises priority must remember both the old and the

new priorities and must downgrade the priority of the token at the appropriate

time. In essence, each station is responsible for assuring that no token circulates

indefinitely because its priority is too high. By remembering the priority of earlier

transmissions, a station can detect this condition and downgrade the priority to a

previous, lower priority or reservation.

To implement the downgrading mechanism, two stacks are maintained by

each station, one for reservations and one for priorities:

S, = stack used to store new values of token priority

S, = stack used to store old values of token priority

The reason that stacks rather than scalar variables are required is that the priority

can be raised a number of times by one or more stations. The successive raises

must be unwound in the reverse order.

To summarize, a station having a higher priority frame to transmit than the

current frame can reserve the next token for its priority level as the frame passes by.

When the next token is issued, it will be at the reserved priority level. Stations of

lower priority cannot seize the token, so it passes to the reserving station or an intermediate

station with data to send of equal or higher priority level than the reserved

priority level. The station that upgraded the priority level is responsible for downgrading

it to its former level when all higher-priority stations are finished. When

that station sees a token at the higher priority after it has transmitted, it can assume

that there is no more higher-priority traffic waiting, and it downgrades the token

before passing it on.

Figure 13.7 is an example. The following events occur:

1. A is transmitting a data frame to B at priority 0. When the frame has completed

a circuit of the ring and returns to A, A will issue a token frame. However,

as the data frame passes D, D makes a reservation at priority 3 by setting

the reservation field to 3.

2. A issues a token with the priority field set to 3.

3. If neither B nor C has data of priority 3 or greater to send, they cannot seize

the token. The token circulates to D, which seizes the token and issues a data

frame.

4. After D's data frame returns to D, D issues a new token at the same priority

as the token that it received: priority 3.

5. A sees a token at the priority level that it used to last issue a token; it therefore

seizes the token even if it has no data to send.

6. A issues a token at the previous priority level: priority 0.

Note that, after A has issued a priority 3 token, any station with data of priority

3 or greater may seize the token. Suppose that at this point station C now has

priority 4 data to send. C will seize the token, transmit its data frame, and reissue a

priority 3 token, which is then seized by D. By the time that a priority 3 token

arrives at A, all intervening stations with data of priority 3 or greater to send will

have had the opportunity. It is now appropriate, therefore, for A to downgrade the

token.

Early Token Release.

When a station issues a frame, if the bit length of the ring is less than that of the

frame, the leading edge of the transmitted frame will return to the transmitting sta

tion before it has completed transmission; in this case, the station may issue a token

as soon as it has finished frame transmission. If the frame is shorter than the bit

length of the ring, then after a station has completed transmission of a frame, it must

wait until the leading edge of the frame returns before issuing a token. In this latter

case, some of the potential capacity of the ring is unused.

To allow for more efficient ring utilization, an early token release (ETR)

option has been added to the 802.5 standard. ETR allows a transmitting station to

release a token as soon as it completes frame transmission, whether or not the

frame header has returned to the station. The priority used for a token released

prior to receipt of the previous frame header is the priority of the most recently

received frame.

One effect of ETR is that access delay for priority traffic may increase when

the ring is heavily loaded with short frames. Because a station must issue a token

before it can read the reservation bits of the frame it just transmitted, the station

will not respond to reservations. Thus, the priority mechanism is at least partially

disabled.

Stations that implement ETR are compatible and interoperable with those

that do not complete such implementation.

IEEE 802.5 Physical Layer Specification

The 802.5 standard (Table 13.4) specifies the use of shielded twisted pair with data

rates of 4 and 16 Mbps using Differential Manchester encoding. An earlier specification

of a 1-Mbps system has been dropped from the most recent edition of the

standard.

A recent addition to the standard is the use of unshielded twisted pair at

4 Mbps.

FDDI Medium Access Control

FDDI is a token ring scheme, similar to the IEEE 802.5 specification, that is designed

for both LAN and MAN applications. There are several differences that are designed

to accommodate the higher data rate (100 Mbps) of FDDI.

MAC Frame

Figure 13.8 depicts the frame format for the FDDI protocol. The standard defines

the contents of this format in terms of symbols, with each data symbol corresponding

to 4 data bits. Symbols are used because, at the physical layer, data are encoded

in 4-bit chunks. However, MAC entities, in fact, must deal with individual bits, so

the discussion that follows sometimes refers to 4-bit symbols and sometime to bits.

A frame other than a token frame consists of the following fields:

Preamble. Synchronizes the frame with each station's clock. The originator of

the frame uses a field of 16 idle symbols (64 bits); subsequent repeating stations

may change the length of the field, as consistent with clocking requirements.

The idle symbol is a nondata fill pattern. The actual form of a nondata

symbol depends on the signal encoding on the medium.

Starting delimiter (SD). Indicates start of frame. It is coded as JK, where J and

K are nondata symbols.

Frame control (FC). Has the bit format CLFFZZZZ, where C indicates

whether this is a synchronous or asynchronous frame (explained below); L

indicates the use of 16- or 48-bit addresses; FF indicates whether this is an

LLC, MAC control, or reserved frame. For a control frame, the remaining 4

bits indicate the type of control frame.

Destination address (DA). Specifies the station(s) for which the frame is

intended. It may be a unique physical address, a multicast-group address, or a

broadcast address. The ring may contain a mixture of 16- and 48-bit address

lengths.

Source address (SA). Specifies the station that sent the frame.

0 Information. Contains an LLC data unit or information related to a control

operation.

Frame check sequence (FCS). A 32-bit cyclic redundancy check, based on the

FC, DA, SA, and information fields.

Ending delimiter (ED). Contains a nondata symbol (T) and marks the end of

the frame, except for the FS field.

0 Frame Status (FS). Contains the error detected (E), address recognized (A),

and frame copied (F) indicators. Each indicator is represented by a symbol,

which is R for "reset" or "false" and S for "set" or "true."

A token frame consists of the following fields:

Preamble. As above.

Starting delimiter. As above.

Frame control (FC). Has the bit format 10000000 or 11000000 to indicate that

this is a token.

Ending delimiter (ED). Contains a pair of nondata symbols (T) that terminate

the token frame.

A comparison with the 802.5 frame (Figure 13.6) shows that the two are quite

similar. The FDDI frame includes a preamble to aid in clocking, which is more

demanding at the higher data rate. Both 16- and 48-bit addresses are allowed in the

same network with FDDI; this is more flexible than the scheme used on all the 802

standards. Finally, there are some differences in the control bits. For example,

FDDI does not include priority and reservation bits; capacity allocation is handled

in a different way, as described below.

MAC Protocol

The basic (without capacity allocation) FDDI MAC protocol is fundamentally the

same as IEEE 802.5. There are two key differences:

1. In FDDI, a station waiting for a token seizes the token by aborting (failing to

repeat) the token transmission as soon as the token frame is recognized. After

the captured token is completely received, the station begins transmitting one

or more data frames. The 802.5 technique of flipping a bit to convert a token

to the start of a data frame was considered impractical because of the high

data rate of FDDI.

2. In FDDI, a station that has been transmitting data frames releases a new

token as soon as it completes data frame transmission, even if it has not begun

to receive its own transmission. This is the same technique as the early token

release option of 802.5. Again, because of the high data rate, it would be too

inefficient to require the station to wait for its frame to return, as in normal

802.5 operation.

Figure 13.9 gives an example of ring operation. After station A has seized the

token, it transmits frame F1, and immediately transmits a new token. F1 is

addressed to station C, which copies it as it circulates past. The frame eventually

returns to A, which absorbs it. Meanwhile, B seizes the token issued by A and transmits

F2 followed by a token. This action could be repeated any number of times, so

that, at any one time, there may be multiple frames circulating the ring. Each station

is responsible for absorbing its own frames based on the source address field.

A further word should be said about the frame status (FS) field. Each station

can check passing bits for errors and can set the E indicator if an error is detected.

If a station detects its own address, it sets the A indicator; it may also copy the

frame, setting the C indicator; this allows the originating station, when it absorbs a

frame that it previously transmitted, to differentiate among three conditions:

0 Station nonexistent/nonactive

0 Station active but frame not copied

0 Frame copied

When a frame is absorbed, the status indicators (E, A, C) in the FS field may be

examined to determine the result of the transmission. However, if an error or failure

to receive condition is discovered, the MAC protocol entitj does not attempt to

retransmit the frame, but reports the condition to LLC. It is the responsibility of

LLC or some higher-layer protocol to take corrective action.

Capacity Allocation

The priority scheme used in 802.5 will not work in FDDI, as a station will often issue

a token before its own transmitted frame returns. Hence, the use of a reservation

field is not effective. Furthermore, the FDDI standard is intended to provide for

greater control over the capacity of the network than 802.5 to meet the requirements

for a high-speed LAN. Specifically, the FDDI capacity-allocation scheme

seeks to accommodate a mixture of stream and bursty traffic,

To accommodate this requirement, FDDI defines two types of traffic: synchronous

and asynchronous. Each station is allocated a portion of the total capacity

(the portion may be zero); the frames that it transmits during this time are

referred to as synchronous frames. Any capacity that is not allocated or that is allocated

but not used is available for the transmission of additional frames, referred to

as asynchronous frames.

The scheme works as follows. A target token-rotation time (TTRT) is defined;

each station stores the same value for TTRT. Some or all stations may be provided

a synchronous allocation (SAi), which may,vary among stations. The allocations

must be such that

The assignment of values for SAi is by means of a station management protocol

involving the exchange of station management frames. The protocol assures that

the above equation is satisfied. Initially, each station has a zero allocation, and it

must request a change in the allocation. Support for synchronous allocation is

optional; a station that does not support synchronous allocation may only transmit

asynchronous traffic.

All stations have the same value of TTRT and a separately assigned value of

SAi. In addition, several variables that are required for the operation of the capacityallocation

algorithm are maintained at each station:

* Token-rotation timer (TRT)

* Token-holding timer (THT)

* Late counter (LC)

Each station is initialized with TRT set equal to TTRT and LC set to zero.'

When the timer is enabled, TRT begins to count down. If a token is received before

TRT expires, TRT is reset to TTRT. If TRT counts down to 0 before a token is

received, then LC is incremented to 1 and TRT is reset to TTRT and again begins

to count down. IF TRT expires a second time before receiving a token, LC is incremented

to 2, the token is considered lost, and a Claim process (described below) is

initiated. Thus, LC records the number of times, if any, that TRT has expired since

the token was last received at that station. The token is considered to have arrived

early if TRT has not expired since the station received the token-that is, if

LC = 0.

When a station receives the token, its actions will depend on whether the

token is early or late. If the token is early, the station saves the remaining time from

TRT in THT, resets TRT, and enables TRT:

THT ß TRT

TRT ß TTRT

enable TRT

The station can then transmit according to the following rules:

1. It may transmit synchronous frames for a time SAi.

2. After transmitting synchronous frames, or if there were no synchronous

frames to transmit, THT is enabled. The station may begin transmission of

asynchronous frames as long as THT > 0.

If a station receives a token and the token is late, then LC is set to zero and

TRT continues to run. The station can then transmit synchronous frames for a time

SAi. The station may not transmit any asynchronous frames.

This scheme is designed to assure that the time between successive sightings

of a token is on the order of TTRT or less. Of this time, a given amount is always

available for synchronous traffic, and any excess capacity is available for asynchronous

traffic. Because of random fluctuations in traffic, the actual token-circulation

time may exceed TTRT, as demonstrated below.

Figure 13.10 provides a simplified example of a 4-station ring. The following

assumptions are made:

1. Traffic consists of fixed-length frames.

2. TTRT = 100 frame times.

3. SAi = 20 frame times for each station.

4. Each station is always prepared to send its full synchronous allocation as many

asynchronous frames as possible.

5. The total overhead during one complete token circulation is 4 frame times

(one frame time per station).

One row of the table corresponds to one circulation of the token. For each

station, the token arrival time is shown, followed by the value of TRT at the time

of arrival, followed by the number of synchronous and asynchronous frames transmitted

while the station holds the token.

The example begins after a period during which no data frames have been

sent, so that the token has been circulating as rapidly as possible (4 frame times).

Thus, when Station 1 receives the token at time 4, it measures a circulation time of

4 (its TRT = 96). It is therefore able to send not only its 20 synchronous frames but

also 96 asynchronous frames; recall that THT is not enabled until after the station

has sent its synchronous frames. Station 2 experiences a circulation time of 120

(20 frames + 96 frames + 4 overhead frames), but is nevertheless entitled to transmit

its 20 synchronous frames. Note that if each station continues to transmit its

maximum allowable synchronous frames, then the circulation time surges to 180 (at

time 184), but soon stabilizes at approximately 100. With a total synchronous utilization

of 80 and an overhead of 4 frame times, there is an average capacity of 16

frame times available for asynchronous transmission. Note that if all stations always

have a full backlog of asynchronous traffic, the opportunity to transmit asynchro

nous frames is distributed among them.

FDDI Physical Layer Specification

The FDDI standard specifies a ring topology operating at 100 Mbps. Two media are

included (Table 13.5). The optical fiber medium uses 4Bl5B-NRZI encoding. Two

twisted pair media are specified: 100-ohm Category 5 unshielded twisted pair6 and

150-ohm shielded twisted pair. For both twisted pair media, MLT-3 encoding is

used. See Appendix 13A for a discussion of these encoding schemes.

100VG-ANYLAN

Like 100BASE-T, ~ O O V G - A ~ ~ iLs iAntNen~de d to be a 100-Mbps extension to the

10-Mbps Ethernet and to support IEEE 802.3 frame types. It also provides compatibility

with IEEE 802.5 token ring frames. 100VG-AnyLAN uses a new MAC

scheme known as demand priority to determine the order in which nodes share the

network. Because this specification does not use CSMAICD, it has been standardized

under a new working group, lEEE 802.12, rather than allowed to remain in the

802.3 working group.

Topology

The topology for a 100VG-AnyLAN network is hierarchical star. The simplest configuration

consists of a single central hub and a number of attached devices. More

complex arrangements are possible, in which there is a single root hub, with one or

more subordinate level-2 hubs; a level-2 hub can have additional subordinate hubs

at level 3, and so on to an arbitrary depth.

Medium Access Control

The MAC algorithm for 802.12 is a round-robin scheme with two priority levels. We

first describe the algorithm for a single-hub network and then discuss the general

case.

Single-Hub Network

When a station wishes to transmit a frame, it first issues a request to the central hub

and then awaits permission from the hub to transmit. A station must designate each

request as normal-priority or high-priority.

The central hub continually scans all of its ports for a request in round-robin

fashion. Thus, an n-port hub looks for a request first on port 1, then on port 2, and

so on up to port n. The scanning process then begins again at port 1. The hub maintains

two pointers: a high-priority pointer and a normal-priority pointer. During one

complete cycle, the hub grants each high-priority request in the order in which the

requests are encountered. If at any time there are no pending high-priority requests,

the hub will grant any normal-priority requests that it encounters.

Figure 13.11 gives an example. The sequence of events is as follows:

1. The hub sets both pointers to port 1 and begins scanning. The first request

encountered is a low-priority request from port 2. The hub grants this request

and updates the low-priority pointer to port 3.

2. Port 2 transmits a low-priority frame. The hub receives this frame and retransmits

it. During this period, two high-priority requests are generated.

3. Once the frame from port 2 is transmitted, the hub begins granting high-priority

requests in round-robin order, beginning with port 1 and followed by

port 5. The high-priority pointer is set to port 6.

4.After the high-priority frame from port 5 completes, there are no outstanding

high-priority requests and the hub turns to the normal-priority requests.

Four requests have arrived since the last low-priority frame was transmitted:

from ports 2,7, 3, and 6. Because the normal-priority pointer is set to port 3,

these requests will be granted in the order 3, 6, 7, and 2 if no other requests

intervene.

5.The frames from ports 3, 6, and 7 are transmitted in turn. During the transmission

of frame 7, a high-priority request arrives from port 1 and a normal

priority request arrives from port 8. The hub sets the normal-priority pointer

to port 8.

6.Because high-priority requests take precedence, port 1 is granted access next.

7.After the frame from port 1 is transmitted, the hub has two outstanding normalpriority

requests. The request from port 2 has been waiting the longest; however,

port 8 is next in round-robin order to be satisfied and so its request is

granted, followed by that of port 2.

Hierarchical Network

In a hierarchical network, all of the end-system ports on all hubs are treated as a single

set of ports for purposes of the round-robin algorithm. The hubs are configured

to cooperate in scanning the ports in the proper order. Put another way, the set of

hubs is treated logically as a single hub.

Figure 13.12 indicates port ordering in a hierarchical network. The order is

generated by traversing a tree representation of the network, in which the branches

under each node in the tree are arranged in increasing order from left to right. With

this convention, the port order is generated by traversing the tree in what is referred

to as preorder traversal, which is defined recursively as follows:

1. Visit the root.

2. Traverse the subtrees from left to right.

This method of traversal is also known as a depth-first search of the tree.

Let us now consider the mechanics of medium access and frame transmission

in a hierarchical network. There are a number of contingencies to consider. First,

consider the behavior of the root hub. This hub performs the high-priority and

normal-priority round-robin algorithms for all directly attached devices. Thus, if

there are one or more pending high-priority requests, the hub grants these requests

in round-robin fashion. If there are no pending high-priority requests, the hub

grants any normal-priority requests in round-robin fashion. When a request is

granted by the root hub to a directly-attached end system, that system may immediately

transmit a frame. When a request is granted by the root hub to a directlyattached

level-2 hub, then control passes to the level-2 hub, which then proceeds to

execute its own round-robin algorithms.

Any end system that is ready to transmit sends a request signal to the hub to

which it attaches. If the end system is attached directly to the root hub, then the

request is conveyed directly to the root hub. If the end system is attached to a lowerlevel

hub, then the request is transmitted directly to that hub. If that hub does not

currently have control of the round-robin algorithm, then it passes the request up to

the next higher-level hub. Eventually, all requests that are not granted at a lower

level are passed up to the root hub.

The scheme described so far does enforce a round-robin discipline among all

attached stations, but two refinements are needed. First, a preemption mechanism

is needed. This is best explained by an example. Consider the following sequence of

events:

1. Suppose that the root hub (R) in Figure 13.12 is in control and that there are

no high-priority requests pending anywhere in the network. However, stations

5-1, 5-2, and 5-3 have all issued normal-priority requests, causing hub B to

issue a normal-priority request to R.

2. R will eventually grant this request, passing control to B.

3. B then proceeds to honor its outstanding requests one at a time.

4. While B is honoring its first normal-priority request, station 1-6 issues a highpriority

request.

5. In response to the request from 1-6, R issues a preempt signal to B; this tells

B to relinquish control after the completion of the current transmission.

6. R grants the request of 1-6 and then continues its round-robin algorithm.

The second refinement is a mechanism to prevent a nonroot hub from retaining

control indefinitely. To see the problem, suppose that B in Figure 13.12 has a

high-priority request pending from 5-1. After receiving control from R, B grants the

request to 5-1. Meanwhile, other stations subordinate to B issue high-priority re

quests. B could continue in round-robin fashion to honor all of these requests. If

additional requests arrive from other subordinates of B during these other transmissions,

then B would be able to continue granting requests indefinitely, even

though there are other high-priority requests pending elsewhere in the network. To

prevent this kind of lockup, a subordinate hub may only retain control for a single

round-robin cycle through all of its ports.

The IEEE 802.12 MAC algorithm is quite effective. When multiple stations

offer high loads, the protocol behaves much like a token ring protocol, with network

access rotating among all high-priority requesters, followed by low-priority requesters

when there are no outstanding high-priority requests. At low load, the protocol

behaves in a similar fashion to CSMAICD under low load: A single requester

gains medium access almost immediately.

100VG-AnyLANPhysical Layer Specification

The current version of IEEE 801.12 calls for the use of 4-pair unshielded twisted

pair (UTP) using Category 3,4, or 5 cable. Future versions will also support 2-pair

Category-5 UTP, shielded twisted pair, and fiber optic cabling. In all cases, the data

rate is 100 Mbps.

Signal Encoding

A key objective of the 100VG-AnyLAN effort is to be able to achieve 100 Mbps

over short distances using ordinary voice-grade (Category 3) cabling. The advantage

of this is that in many existing buildings, there is an abundance of voice-grade

cabling and very little else. Thus, if this cabling can be used, installation costs are

minimized.

With present technology, a data rate of 100 Mbps over one or two Category 3

pairs is impractical. To meet the objective, 100VG-AnyLAN specifies a novel

encoding scheme that involves using four pair to transmit data in a half-duplex

mode. Thus, to achieve a data rate of 100 Mbps, a data rate of only 25 Mbps is

needed on each channel. An encoding scheme known as 5B6B is used. (See Appendix

13A for a description.)

Data from the MAC layer can be viewed as a stream of bits. The bits from this

stream are taken five at a time to form a stream of quintets that are then passed

down to the four transmission channels in round-robin fashion. Next, each quintet

passes through a simple scrambling algorithm to increase the number of transitions

between 0 and 1 and to improve the signal spectrum. At this point, it might be possible

to simply transmit the data using NRZ. However, even with the scrambling,

the further step of 5B6B encoding is used to ensure synchronization and also to

maintain dc balance.

Because the MAC frame is being divided among four channels, the beginning

and ending of a MAC frame must be delimited on each of the channels, which is the

purpose of the delimiter generators. Finally, NRZ transmission is used on each

channel.

ATM LAN

A document on customer premises networks jointly prepared by Apple, Bellcore,

Sun, and Xerox [ABSX92] identifies three generations of LANs:

First Generation. Typified by the CSMNCD and Token Ring LANs. The first

generation provided terminal-to-host connectivity and supported clientlserver

architectures at moderate data rates.

Second Generation. Typified by FDDI. The second generation responds to

the need for backbone LANs and for support of high-performance workstations.

Third Generation. Typified by ATM LANs. The third generation is designed

to provide the aggregate throughputs and real-time transport guarantees that

are needed for multimedia applications.

Typical requirements for a third generation LAN include the following:

I. Support multiple, guaranteed classes of service. A live video application, for

example, may require a guaranteed 2-Mbps connection for acceptable performance,

while a file transfer program can utilize a background class of service.

2. Provide scalable throughput that is capable of growing in both per-host capacity

(to enable applications that require large volumes of data in and out of a

single host) and in aggregate capacity (to enable installations to grow from a

few to several hundred high-performance hosts).

3. Facilitate the interworking between LAN and WAN technology.

ATM is ideally suited to these requirements. Using virtual paths and virtual

channels, multiple classes of service are easily accommodated, either in a preconfigured

fashion (permanent connections) or on demand (switched connections).

ATM is easily scalable by adding more ATM switching nodes and using higher (or

lower) data rates for attached devices. Finally, with the increasing acceptance of

cell-based transport for wide-area networking, the use of ATM for a premises network

enables seamless integration of LANs and WANs.

The term ATM LAN has been used by vendors and researchers to apply to a

variety of configurations. At the very least, an ATM LAN implies the use of ATM

as a data transport protocol somewhere within the local premises. Among the possible

types of ATM LANs:

Gateway to ATM WAN. An ATM switch acts as a router and traffic concentrator

for linking a premises network complex to an ATM WAN.

Backbone ATM switch. Either a single ATM switch or a local network of

ATM switches interconnect other LANs.

Workgroup ATM. High-performance multimedia workstations and other end

systems connect directly to an ATM switch.

These are all "pure" configurations. In practice, a mixture of two or all three

of these types of networks is used to create an ATM LAN.

Figure 13.13 shows an example of a backbone ATM LAN that includes links

to the outside world. In this example, the local ATM network consists of four

switches interconnected with high-speed, point-to-point links running at the standardized

ATM rates of 155 and 622 Mbps. On the premises, there are three other

LANs, each of which has a direct connection to one of the ATM switches. The data

rate from an ATM switch to an attached LAN conforms to the native data rate of

that LAN. For example, the connection to the FDDI network is at 100 Mbps. Thus,

the switch must include some buffering and speed conversion capability to map the

data rate from the attached LAN to an ATM data rate. The ATM switch must also

perform some sort of protocol conversion from the MAC protocol used on the

attached LAN to the ATM cell stream used on the ATM network. A simple

approach is for each ATM switch that attaches to a LAN to function as a bridge or

router.'

An ATM LAN configuration such as that shown in Figure 13.13 provides a

relatively painless method for inserting a high-speed backbone into a local environment.

As the on-site demand rises, it is a simple matter to increase the capacity of

the backbone by adding more switches, increasing the throughput of each switch,

and increasing the data rate of the trunks between switches. With this strategy, the

load on individual LANs within the premises can be increased, and. the number of

LANs can grow.

However, this simple backbone ATM LAN does not address all of the needs

for local communications. In particular, in the simple backbone configuration, the

end systems (workstations, servers, etc.) remain attached to shared-media LANs

with the limitations on data rate imposed by the shared medium.

A more advanced, and more powerful approach, is to use ATM technology in

a hub. Figure 13.14 suggests the capabilities that can be provided with this

approach. Each ATM hub includes a number of ports that operate at different data

rates and that use different protocols. Typically, such a hub consists of a number of

rack-mounted modules, with each module containing ports of a given data rate and

protocol.

The key difference between the ATM hub shown in Figure 13.14 and the

ATM nodes depicted in Figure 13.13 is the way in which individual end systems are

handled. Notice that in the ATM hub, each end system has a dedicated point-topoint

link to the hub. Each end system includes the communications hardware and

software to interface to a particular type of LAN, but in each case, the LAN contains

only two devices: the end system and the hub! For example, each device

attached to a 10-Mbps Ethernet port operates using the CSMAICD protocol at

10 Mbps. However, because each end system has its own dedicated line, the effect

is that each system has its own dedicated 10-Mbps Ethernet. Therefore, each end

system can operate at close to the maximum 10-Mbps data rate.

The use of a configuration such as that of either Figure 13.13 or 13.14 has the

advantage that existing LAN installations and LAN hardware-so-called legacy

LANs-can continue to be used while ATM technology is introduced. The disadvantage

is that the use of such a mixed-protocol environment requires the implementation

of some sort of protocol conversion capability, . A simpler approach, but one that requires that end systems be

equipped with ATM capability, is to implement a "pure" ATM LAN.

One issue that was not addressed in our discussion so far has to do with the

interoperability of end systems on a variety of interconnected LANs. End systems

attached directly to one of the legacy LANs implement the MAC layer appropriate

to that type of LAN. End systems attached directly to an ATM network implement

the ATM and AAL protocols. As a result, there are three areas of compatibility to

consider:

1. Interaction between an end system on an ATM network and an end system on

a legacy LAN.

2. Interaction between an end system on a legacy LAN and an end system on

another legacy LAN of the same type (e.g., two IEEE 802.3 networks).

3. Interaction between an end system on a legacy LAN and an end system on

another legacy LAN of a different type (e.g., an IEEE 802.3 network and an

IEEE 802.5 network).