Nyamdavaa: December 2010

Not long ago, desktop computers were rare, and the Internet was the

province of a handful of intellectuals in universities, government agencies, and

large companies. Most people had no idea what a modem was or why they would

ever need one. Now, desktop computers are selling for well under $1000,

hard drive capacities run into the multi-gigabytes, the applications are almost

unlimited, and the Internet has attracted millions to the data communications

world.

Telecommunications and the computer are partners in a marriage that has

changed the way people store, access, and use information. In the heyday of the

mainframe, databases were centralized and a specialized program was required to

access them. That program was an application that ran on the mainframe, which

transmitted information over a structured network to the terminal that requested

it. The terminal could modify the record only to the extent that the program

permitted. The mainframe is quite efficient at what it does, but its applications

are limited. No computer of the mainframe era has applications such as word

processing, spreadsheet, or e-mail that approach the effectiveness of those on

desktop computers.

The merger of mainframe and telecommunications makes possible many

applications including automatic teller machines, airline reservation systems, and

credit card verification networks. These use text-based clients, and work fine on

dumb terminals. More effective, however, are graphics-based applications that

enable the user to merge images and multi-color graphics with text. These require

desktop processing power, graphic displays, and storage capacity that outstrip the

capabilities of computers and networks of a few years ago.

This chapter is the second of several that deal with data communications.

Chapter 3 was an overview intended to put data communications in perspective.

That chapter omitted details and used terminology that is difficult to grasp

without a deeper understanding of how the major components of a data network

function down at the bit level.

As we discussed in Chapter 3, data protocols function in layers. The lowest

level is the physical layer, in which bits move across a transmission medium,

which is a pair of wires, a coaxial cable, a radio channel, or an optical fiber pair.

Before computers came on the scene, data communications used only the physical

network. Two teletypewriters (or two desktop computers with their serial ports

connected) can exchange data over two pairs of wires, but with limitations.

Disturbances from a variety of sources can mutilate bits and render the message

useless or worse if the error passes undetected.

This chapter expands on Chapter 3 with explanations of how data devices

ensure error-free transport across telecommunications media that are subject to a

variety of disturbances. This chapter is hardware-oriented, with enough discussion

of protocols to prepare you for a discussion of pulse-code modulation in Chapter 5

and a more detailed protocol discussion in Chapter 6.

DATA COMMUNICATIONS FUNDAMENTALS

Data communications is an endless quest for perfection. Billions of dollars move

among financial institutions throughout the world, software traverses the

Internet, and goods and services move through electronic document interchange.

Billions of transactions are carried daily without as much as one misplaced bit.

Binary digits, from which the word bit is derived, are either right or wrong and a

single bit in error can convert a file to rubbish. Some data applications can tolerate

an occasional error, but most require absolute integrity regardless of whether the

transmission medium is the finest fiber-optic cable or a deteriorated rural wire

line running through a swamp.

The devices that originate and receive data are called data terminal equipment

(DTE). These can range from computers to simple receive-only terminals or

printers. DTE connects to the telecommunications network through data

communications equipment (DCE), which converts the DTE’s output to a signal

suitable for the transmission medium. DCE ranges from line drivers to complex

modems and multiplexers.

The basic information element that a computer processes is the bit, which

is represented by the two digits 0 and 1. Processors manipulate data in groups of

eight bits known as bytes or octets. To make binary digits easier for humans to

manipulate, octets are often split into groups of four bits and represented as the

hexadecimal digits 0 to 9 and A to F. Inside the computer data travels over parallel

paths. Parallel transmission is suitable for short distances to peripherals such

as printers, but for communications over a range of more than a few feet, the eight

parallel bits are converted to serial as Figure 4-1 shows. This serial bit stream is

coupled to telecommunications circuits through some type of DCE.

Coding

The number of characters that binary numbers can encode depends on the

number of bits in the code. Early teletypewriters used a five-bit code called

Baudot that had a capacity of 25 or 32 characters. Today’s telecommunications

device for the deaf (TDD) uses this code. Thirty-two characters are not enough to

send a full range of upper and lower case plus special characters. In the Baudot

code, shift characters trigger the machines between upper and lower cases.

The receiving device continues in its current mode until it receives a shift

character. If it misses a shift, the device continues to print and the transmission

is garbled.

To overcome this 32-character limitation, teletypewriters evolved to a sevenlevel

code known as the American Standard Code for Information Interchange

(ASCII). This code, which Table 4-1 shows, provides 27 or 128 combinations.

Although the ASCII code uses seven bits for characters, eight bits are transmitted.

The eighth bit is used for error detection as described later.

Several other codes are used for data communications. Many IBM devices

use Extended Binary Coded Decimal Interchange Code (EBCDIC), which

Table 4-2 shows. EBCDIC is an eight-bit code, allowing the full 256 characters to

be encoded. Neither EBCDIC nor ASCII can begin to represent all of the characters

and technical symbols in the world’s languages. Unicode is a 16-bit code that

provides a platform for encoding any symbol into binary. Not only are all current

and archaic languages encoded, but the standard also includes musical symbols

and geometric shapes. Code compatibility between machines is essential. Because

EBCDIC, ASCII, and Unicode are widely used, some applications will require

code conversion. Most Web browsers and many application programs support

Unicode and can convert between the codes.

Data Communication Speeds

Data communications speeds are measured in bits per second (bps) or some

multiple: kilobits (Kbps), megabits (Mbps), gigabits (Gbps), and terabits (Tbps).

It is easy to confuse this with storage and file sizes, which are always quoted in

bytes. As a matter of convention, bits are shown in lower case and bytes in upper.

For example, 1 Mbps is megabits and 1 MBps is megabytes.

Early data applications were limited by the speed at which an operator could

type or toggle a telegraph key. When punched paper tape teletypewriters were

developed, the operator could type off-line, and then send at the full speed of the

device. Teletypewriters using the Baudot code around the time of World War II

ran at 60 words per minute, which was roughly 30 bps. Later ASCII machines

upped the speed to 100 wpm, still a fraction of the data-carrying capacity of a

voice circuit. Analog multiplexers at the time subdivided voice channels so they

could carry multiple telegraph or teletypewriter signals.

The telephone channel bandwidth of 300 to 3300 Hz imposes an upper limit

on data transmission speeds. Many people use bit rate and baud rate interchangeably

to express the data-carrying capacity of a circuit, but they are not

technically synonymous. Bit rate is the number of bits per second the channel can

carry. Baud rate is the number of cycles or symbol changes per second the channel

can support. The bandwidth of a voice channel is limited to 2400 baud, but higher

bit rates are transmitted by encoding more than one bit per baud. A 19,200-bps

modem, for example, encodes eight bits per baud. The latest version of high-speed

modems uses somewhat higher baud rates than 2400 to achieve downstream

speeds as high as 56 Kbps. This speed requires one end of the connection to have

digital connectivity to the central office.

Modulation Methods

A data signal leaves the serial interface of the DTE as a series of baseband voltage

pulses as Figure 4-2 shows. Baseband means that the varying voltage level from

the DTE is impressed without modulation directly on the transmission medium.

Baseband pulses can be transmitted over limited distances across copper wire

from a computer’s serial interface. RS-232, or more accurately EIA-232, is the

standard that most computers support. The standard limits the cable length to

50 ft (15 m), although many users operate it successfully over longer distances.

The length limitations of EIA-232 result from its use of unbalanced transmit

and receive leads, each of which is a single wire, sharing a common return path. A

balanced transmission medium such as EIA-423 can transmit over much greater distances

because the transmit and receive paths are separate. EIA-423 has a length limit

of 4000 ft (1200 m). EIA-232 can operate over longer distances by using a limited distance

modem or a line driver that matches a short serial interface to a balanced cable

pair. Abalanced transmission medium is inherently less susceptible to noise.

For transmission over voice-grade channels, a modem modulates data

pulses into a combination of analog tones and amplitude and phase changes that

fit within the channel pass band. The digital signal modulates the frequency, the

amplitude, or the phase of an audio signal as Figure 4-2 shows. High-speed modems

use all three methods. Amplitude modulation by itself is the least-used method

because it is susceptible to noise-generated errors. It is frequently used, however,

Data Modulation Methods

in conjunction with frequency and phase changes. Frequency modulation is an

inexpensive method used with low-speed modems. To reach speeds of more than

300 bps, phase shift modems are employed.

Quadrature Amplitude Modulation (QAM)

Modems use increasingly complex modulation methods for encoding multiple

bits per baud to reach speeds approaching the theoretical limit of a voice grade

circuit. Since an analog channel is nominally limited to 2400 baud, to send

9600 bps, e.g., four bits per baud must be encoded. This yields 24 or 16 combinations

that each symbol can represent. High-speed modems use a method known

as QAM. In QAM, two carrier tones combine in quadrature to produce the modem’s

output signal. The receiving end demodulates the quadrature signal to recover the

transmitted signal. In 16 QAM, each symbol carries one of 16 signal combinations.

As Figure 4-3 shows, any combination of four bits can be encoded into a particular

pair of X–Y plot points, each of which represents a phase and amplitude combination

that corresponds to a 4-bit sequence. The four bits can be expressed as

hexadecimal in 24 or 16 combinations. This two-dimensional diagram is called a

signal constellation.

The receiving modem demodulates the signal to determine what pair of X–Y

coordinates was transmitted, and the four-bit signal combination passes from the

modem to the DTE. If line noise or phase jitter affect the signal, the received point

will be displaced from its ideal location, so the modem must make a best guess

Signal Constellation in a 16-Bit (24) QAM Signal

as to which plot point was transmitted. If the signal is displaced far enough, the

receiver guesses wrong, and the resulting signal is in error.

Even higher rates can be modulated, with each additional bit doubling the

number of signal points. A64-QAM signal encodes 26 bits per symbol, a 128-QAM

signal results in 27 combinations, and a 256-QAM signal results in 28 combinations,

bringing the signal points closer together and increasing the susceptibility

of the modem to impairments. DSL modems use QAM to modulate a broadband

data signal above the voice circuit on a telephone cable pair.

TCM is an encoding method that makes each symbol dependent on adjacent

symbols. In a 14,400-bps modem, for example, data is presented to a TCM modulator

in six-bit groups. Two of the six bits are separated from the signal, and a code

bit is added. The resulting signal is two groups: one three-bit and one four-bit.

These combine into 27 bits, which are mapped into a signal point and selected

from a 128-point signal constellation. Since only six of the seven bits are required

to transmit the original signal, not all 128 points are needed to transmit the signal,

and only certain patterns of signal points are defined as valid. If an error causes an

invalid pattern at the receiver, the decoder selects the most likely valid sequence

and forwards it to the DTE. If it guesses wrong, the DTE’s error-correction mechanism

arranges for retransmission. TCM reduces the signal’s susceptibility to line

impairments and reduces the number of retransmissions.

Full- and Half-Duplex Mode

Full-duplex data systems transmit in both directions simultaneously. Half-duplex

systems transmit in only one direction at a time; the channel reverses for

transmission in the other direction. Half-duplex is used only on dedicated line

facilities where the application is inherently half-duplex. A good example is an

automatic teller machine where the user interacts with a central computer to make

deposits and withdrawals with each end of the transmission sending a short

message that identifies the user and actuates the transaction.

LECs provide dedicated analog circuits as either two-wire or four-wire, but

almost all data private lines use four-wire facilities. LECs also offer dedicated

digital circuits. A 56-Kbps digital channel (64 Kbps in Europe) provides a fourwire

digital channel. This service uses a bipolar modulation method, which is

discussed in Chapter 5.

The LECs’ and IXCs’ inter-office facilities are inherently four-wire. To

provide end-to-end four-wire facilities, the LEC assigns two cable pairs in the

local loop. Two-wire facilities can support full-duplex operation by using modems

that separate the two directions of transmission. Split channel modems provide

the equivalent of four-wire operation by dividing the voice channel into two

segments, one for transmit and one for receive. Dial-up modems support 2400-bps

full-duplex communication over two-wire circuits using the ITU V.22 bis modulation

method, 9600 bps using V.32, 14,400 bps with V.32 bis, 33,600 bps with V.34,

and 56 Kbps with V.92 modulation.

Synchronizing Methods

All data communications channels require synchronization to keep the sending

and receiving ends in step. The signal on a baseband data communications

channel is a series of rapid voltage changes, and synchronization enables the

receiving terminal to determine which pulse is the first bit in a character.

The simplest synchronizing method is asynchronous, sometimes called

stop–start synchronization. Asynchronous signals, illustrated in Figure 4-4, are in

the one or mark state when no characters are being transmitted. Acharacter begins

with a start bit at the zero or space level followed by eight data bits and a stop bit

at the one level. The terms mark and space originated in telegraphy and extend

to teletypewriters. A teletypewriter needs line current to hold it closed when it is

not receiving characters. Some asynchronous terminals also use a current loop

over ranges greater than the EIA-232 serial standard supports. Current loops have

largely disappeared from public networks because they generate noise and the

ILECs cannot guarantee that circuits will be assigned to metallic cable.

Asynchronous signals are transmitted in a character mode, i.e., each character

is individually synchronized and unrelated to any other character in the

transmission. One drawback of asynchronous communication is the extra two

overhead bits per character that carry no information. Asynchronous communication

also lacks the ability to correct errors. Asynchronous has a major advantage

of being a simple and universal standard. Nearly every desktop computer has a

serial port that can be attached to a modem for communications wherever a telephone

can be found. Modems overcome much of the asynchronous deficiency by

implementing an error detection and correction dialogue.

Protocols intended for LANs and WANs use synchronous or block mode

protocol. Figure 4-5 shows a High-Level Datalink Control (HDLC) synchronous

frame. The PDU structure is different for other datalink protocols such as

Ethernet, but the principles are the same. An information block is sandwiched

Asynchronous Transmission

between header and trailer records. The header contains addressing and control

information, and the trailer handles error detection and correction as explained in

the next section. A starting flag contains a unique bit pattern that prepares the

receiving device to receive the frame. The header and trailer lengths are set in the

protocol and the control octet contains, among other things, the length of the data

block. The network administrator adjusts the length of the block to fit the needs of

the application and the characteristics of the network.

Error Detection and Correction

Errors occur in all data communications circuits. Where the transmission is text

that people will interpret, a few errors can be tolerated because the meaning can

be derived from context. Teletypewriters have no error-correction capability, but

they use other techniques to flag errors and users can interpret the message from

context. Data applications are not tolerant of errors, but voice and video can

accept a high rate of errors with imperceptible effect. This section discusses

causes, detection, and correction of data communication errors.

Causes of Data Errors

The type of transmission medium and the modulation method have the greatest

effect on the error rate. Any analog transmission medium is subject to external noise,

which affects the amplitude of the signal. Atmospheric conditions, such as lightning

that cause static bursts, and noise induced from external sources such as power lines

all cause errors. Digital circuits carried on fiber optics are immune to these influences,

but digital radio is susceptible to these as well as signal fades. Fiber-optic systems

exhibit an infinitesimal error rate until something fails in the electronics and the

system switches to a standby channel. Technicians probably cause the bulk of data

errors. Any circuit is subject to errors during maintenance activities and external

damage or interruption by vandalism. Even LANs within a single building are subject

to occasional interruptions due to equipment failure or human error. The best

error mitigation program is a design that reduces the susceptibility of the service to

errors. Nevertheless, errors are inevitable and corrective measures are essential.

The simplest way of detecting errors is parity checking, or vertical redundancy

checking (VRC), a technique used on asynchronous circuits, particularly on

teletypewriters. In the ASCII code set, the eighth bit is reserved for parity. Parity

is set as odd or even, referring to the number of 1 bits in the character. As

Figure 4-6 shows, DTE adds an extra bit, if necessary, to cause each character to

match the parity established for the network.

Most asynchronous terminals can be set to send and receive odd, even, or no

parity. When a parity error occurs, some terminals can register an alarm, but for the

most part parity is useless in computers. Parity has two drawbacks: there is no way

to tell what the original character should have been and, worse, if an even number

of error occurs, parity checking will not even detect that there was an error. In today’s

data networks, parity is irrelevant, although it is part of modem setup strings.

Echo Checking

The receiving computer may echo the received characters back to the sending end.

This technique, called echo checking, is suitable for detecting some errors in keyboarded

text. The typist sees an unexpected echoed character, backspaces, and

retypes it. An error in an echoed character is as likely to have occurred on the return

trip as in the original transmission, so the receiver may have the correct character

but the transmitter believes it was received in error. Although echo checking is ineffective

in machine-to-machine communications, some computers still use it. You

may inadvertently set up your modem to display characters locally and they are also

echoed from the distant end. In this case double characters appear on the screen.

Cyclical Redundancy Checking (CRC)

Synchronous data networks use CRC to detect and correct errors. The sending

DCE processes the bits in each frame against a complex polynomial that always

results in a remainder. The remainder is entered in an error check block following

the data block. The receiving DCE recalculates the CRC field against the header

and data block and compares it to the received CRC. If the two match, the frame

is acknowledged; otherwise the protocol returns a message that instructs the

sender to retransmit. The sender must, therefore, buffer PDUs until it receives

an acknowledgement. The probability of an undetected error with CRC is so slight

that it can be considered error-free. Synchronous datalink protocols acknowledge

which frames have been received correctly through a process that Chapter 6

describes. If a PDU is not acknowledged before the protocol times out, the sending

end retransmits it. This can result in duplicate PDUs, so the protocol must detect

these and kill the duplicate.

One bit in error in a frame is fully as detrimental as a long string of errors.

Most carriers quote the bit-error rate (BER) or error-free seconds (EFS) in their

SLAs. Errors often come in groups and since one bit-error destroys the frame,

a high BER may be a somewhat deceptive quality measurement. A better

measurement of datalink quality is the block error rate (BLER), which is

calculated by dividing the number of errored blocks or frames received over a

period by the total number of blocks transmitted. A device such as a front-end

processor or a protocol analyzer can compute BLER.

Forward Error Correction (FEC)

When the BLER of a circuit is excessive, throughput may be reduced to an

unacceptable level. The longer the data block, the worse the problem because of

the amount of data that must be retransmitted. FEC can help bring the error rate

down to a manageable level. In FEC systems, an encoder on the transmitting end

processes the incoming signal and generates redundant code bits. The transmitted

signal contains both the original information bits plus the redundant bits. At the

receiving end, the redundant bits are regenerated from the information bits and

compared with the redundant bits in the received signal. When a discrepancy

occurs, the FEC circuitry on the receiving end uses the redundant bits to generate

the most likely bit combination and passes it to the DTE. Although FEC is fallible,

it reduces the BLER and the number of retransmissions.

Throughput

One critical measure of a data communication circuit is its throughput, defined as

the number of information bits correctly transferred per unit of time. Although it

would be theoretically possible for the throughput of a data channel to approach

its maximum bit rate, in practice this can never be realized because of overhead

bits and the retransmission of errored PDUs. The following are the primary factors

that limit the throughput of a data channel:

_ Modem speed. The faster the modem the less the time taken to transmit

a block of data.

_ Half- or full-duplex mode. With other factors equal on a private line circuit,

full-duplex circuits have greater throughput because the modems do not

have to reverse between transmitting and receiving.

_ Error rate. The higher the error rate, the more the retransmissions and

the lower the throughput.

66 PART 1 Introduction

_ Protocol. Different protocols have different overhead bits and errorhandling

methods. Also, some protocols cannot be transmitted over

a particular network and must be encapsulated in another protocol,

which increases overhead.

_ Size of data block. If the error rate is high, short data blocks are more

efficient because the retransmission time is high. If the error rate is low,

long data blocks are more efficient. The shorter the data block, the

more significant the header and trailer as a percentage of PDU length.

When the data block is too long, each error necessitates retransmitting

considerable data. Optimum block length is a balance between time

consumed in overheads and in error retransmission.

_ Propagation speed. This factor is the time required for data to traverse

the circuit. It depends on the length of the circuit and the type of

transmission medium. Satellite circuits have the greatest delay.

The network administrator optimizes the throughput of a data channel by

reaching a balance between the above variables.

DATA COMMUNICATIONS EQUIPMENT

An effective data communications network is a compromise involving many

variables. The nature of data transmission varies so greatly with the application

that designs are often empirically determined. The network designer arrives at

the most economical balance of performance and cost, evaluating equipment

alternatives as discussed in this section.

Terminals

The dumb terminal of the past is giving way to the PC and a variety of hand-held

and wireless devices that communicate with special hosts. Dumb terminals still

have their uses. For example, PBXs, routers, multiplexers, and other such devices

are equipped with EIA-232 ports so they can be configured from a terminal. Since

a serial port is a standard feature of most desktop computers, it is simple for a

computer to emulate an asynchronous terminal. Telecommunications software

ranges in features from simple dumb terminal emulation to full-featured intelligent

terminal applications. In the latter category, a desktop computer can upload

and download files from and to its own disk, select and search for files on the host,

and even interact with the host without a human attendant.

Modems

Since the early 1980s, modems have undergone a dramatic evolution. To discuss

modems, it is useful to classify them as dial-up and private line. In the dial-up

category, modems have almost become a commodity. They are manufactured to

international standards, and nearly every computer contains one. The interface

between DTE and the modem is standardized in most countries, with the

predominant interfaces being the EIA-232, EIA-449, and ITU V.35. EIA and ITU

standards specify the functions of the interface circuits but do not specify the

physical characteristics of the interface connector. Connectors have been adopted

by convention, e.g., the DB-25 connector has become a de facto standard for the

EIA-232 interface. Not all 25 pins of the DB-25 are necessary in most applications,

so many products use the nine-pin DB-9 connector.

Dial-Up Modems

Like other telecommunications products, modems have steadily become faster,

cheaper, and smarter, with V.92 modems being the modern norm. Modem setup

was once somewhat tricky, but most devices are now self-configuring except for

special terminal emulation functions. Dial-up modems either plug into a desktop

computer expansion slot or are self-contained devices that plug into the

computer’s serial port. Modems support an error correction protocol and

implement V.44 data compression. Most also support fax.

The switched telephone network carries a considerable share of asynchronous

data communication. Therefore, many modem features are designed to

emulate a telephone set. The most sophisticated modems, in combination with a

software package in an intelligent terminal, are capable of fully unattended

operation. Modems designed for unattended, and many designed for attended,

operation include these features:

_ dial tone recognition

_ automatic tone and dial pulse dialing

_ monitoring call progress tones such as busy and reorder

_ automatic answer

_ call termination

Dial-up modems operate in a full-duplex mode. When two modems connect,

they go through an elaborate exchange of signals to determine the features the

other modem supports. Such features as error correction and compression are

examined. High-speed modems test the line to determine the highest speed with

which they can communicate and fall back to that speed.

The V.92 Standard

An ordinary telephone circuit is designed to support voice communications and

has inherent characteristics that limit its bandwidth. For years, engineers believed

that 33.6 Kbps was the maximum speed that a voice-grade circuit could carry.

With the popularity of the Internet, companies began seeking ways to increase

modem speeds. Engineers reasoned that in virtually every telephone connection,

68 PART 1 Introduction

most of the circuit is digital and that only the local loop from the central office to

the user’s premise is analog.

As we will discuss in Chapter 5, every time a circuit undergoes an analog-todigital

conversion, a bit of the quality is lost. If the connection could be digital all

the way except for the loop on the modem user’s end of the circuit, only one analog

conversion would take place, and the majority of the connection would be digital.

Several companies began experimenting with an approach to increase modem

speed. Compatible modem protocols would be used at each end of the connection,

but the portion of the circuit from the central office to the ISP would be digital.

Several proprietary protocols came on the market before ITU-T approved the

V.90 standard in 1998. The V.92 standard followed and improved on V.90 with

innovations such as V.44 compression and a quick-connect procedure. Although

56 Kbps is possible, it is not always achieved. Poor phone-line quality limits

speed and one end of the connection must be digital. Note that V.92 modems are

asymmetric. They download at speeds up to 56 Kbps, but are limited to 33.6 Kbps

in the upstream direction.

Private Line Modems

Analog private lines are rapidly becoming outdated, and private line modems are

replaced by their digital equivalents, so little additional development work is

being conducted. Different manufacturers use proprietary formats to encode the

signal, compress data, and more important, communicate network management

information. Private line modems can be classed as synchronous or asynchronous,

half- or full-duplex, and two- or four-wire, with the latter being the most common.

Circuit throughput can be improved by using data compression. With data compression

and adaptive equalization, it is possible to operate at 19.2 Kbps or higher

over voice-grade lines.

Special Purpose Modems

The market offers many modems that fulfill specialized requirements. This section

discusses some of the equipment that is available:

_ Alarm reporting modem. This class of modem has connections for

accepting and relaying alarms from external devices. It may also

monitor the ASCII bit stream of a channel looking for particular bit

patterns. When alarms occur, the modem dials a predefined number.

_ DSL and cable modems. These devices, discussed in more detail

in Chapter 8, are used for access, primarily to the Internet.

_ Dial-backup modems. A dial-backup modem contains circuitry to restore

a failed leased line over a dial-up line. The restoral may be automatically

initiated on failure of the dedicated line. The modem may simulate

a four-wire private line over a single dial-up line, or two dial-up lines

may be required.

_ Fiber-optic modems. Where noise and interference are a problem,

fiber-optic modems can provide high bandwidth at a moderate cost.

Operating over one fiber pair, these modems couple directly to the

fiber-optic cable.

_ Limited distance modems. Many LECs offer limited distance circuits, which

are essentially a bare nonloaded cable pair between two points within

the same wire center. LDMs are inexpensive modems operating at speeds

of up to 19.2 Kbps. Where LDM capability is available, the modems are

significantly less expensive than long-haul 19.2-Kbps modems.

Data Service Units/Channel Service Unit (DSU/CSU)

ADSU/CSU connects DTE to a digital circuit. It provides signal conditioning and

testing points for digital circuits. For example, the bit stream from a data device

is generally a unipolar signal, which must be converted to a bipolar signal for

transmission on a digital circuit. The CSU/DSU does the conversion, and provides

a loop-back point for the carrier to make out-of-service tests on the circuit.

Operating at 56 and 64 Kbps, DSUs are full-duplex devices. They are available for

both point-to-point and multidrop lines.

Multiplexers and Concentrators

A data multiplexer subdivides a voice-grade line so it can support multiple

sessions, usually from dumb terminals. Multiplexers come in two varieties.

A standard data multiplexer carves the line into multiple channels. For example,

the 2400-baud capacity of a voice-grade circuit could be divided into 16 channels

of 150 baud each, with more capacity than a typist can use. The nature of many

data applications is such that the terminals are idle a great deal of the time. With

a straight multiplexer the idle time slots are wasted. A statistical multiplexer is able

to make use of this time by assigning time slots as necessary to meet the demand.

A typical statmux might provide 32 time slots on a single voice-grade line.

A concentrator is similar to a multiplexer, except that it is a single-ended

device. At the terminal end, devices connect to the concentrator exactly as they

would connect to a multiplexer, and the concentrator connects to the facility. At

the host end, the facility connects into the host or front-end processor. A concentrator

matches the characteristics of the host processor.

The primary application for multiplexers is in data networks that use asynchronous

terminals. Since many of these devices cannot be addressed and have no

error correction capability, they are of limited use by themselves in remote locations.

The multiplexer provides end-to-end error checking and correction and circuit sharing

to support multiple terminals. Although multiplexers are still available, they are

being displaced by local area networks linked over digital circuits.

LAN Equipment

Much of the hardware we have discussed in this section is of limited applicability

in today’s network because it is obsolete, replaced by LANs and equipment for

interconnecting them. We will go into these devices in considerably more detail in

subsequent chapters, but we discuss them here briefly to complete the equipment

picture and to prepare for the protocol discussion that follows in Chapter 6.

Hubs

The earliest LAN segments used coaxial cable as a transmission medium. Coax

is bulky and unwieldy and is now obsolete for LAN use. Modern networks

use unshielded twisted-pair (UTP) wire that is carefully designed and constructed

to support data communications up to 1 Gbps over distances of 100 m. See

Chapter 9 for additional information on the wiring infrastructure.

At first, stations were connected together with a central multi-port hub.

Many hubs still exist, but they are being phased out because they have a major

drawback. All of the stations connected to the segment share the bandwidth and

contend with one another for access. When stations attempt to transmit simultaneously,

they collide, their transmissions are mutilated, and they must retransmit

the frame. When the traffic reaches the point of excessive collisions, throughput

drops off and the LAN must be broken into smaller segments. This is accomplished

by means of a bridge.

LAN Bridges

ALAN bridge is a two-port device that interconnects two segments. Its method of

operation is simple. By listening to traffic on the network, it learns which MAC

addresses belong to which segment and builds a table. If the sending and receiving

addresses are on the same segment the bridge ignores the frame. If the

addressee is on the other segment, it lets the frame across the bridge. If the bridge

does not have the addressee in its table, it broadcasts a query on both ports.

Bridging has further limitations that we discuss in later chapters, among which is

its two-port limit. The solution is to eliminate both bridges and hubs by using

a switch.

Ethernet Switches

An Ethernet switch is, in effect, a multi-port bridge. Each station is assigned to a

port, so potential collisions are eliminated. The switch learns the station’s MAC

address on each port and connects the sending and receiving ports long enough

to pass a frame. Some LANs share switch ports with hubs, but the cost of switches

is so low that sharing ports is usually more trouble than it is worth. Switches

are effective devices that nearly every LAN uses, but they have a drawback

of their own: they cannot handle more than one route between the originating

and terminating ports. For this, we need a router and a different addressing scheme.

Switches operate on the MAC address, but packet flows require an IP address.

Routers

Routers are the workhorses of the Internet. They are specialized computers that

connect a user’s LAN to the Internet, and within an IP network they consult routing

tables to determine which of their alternative routes is the most effective one

to carry a packet. They are more expensive than switches, and although they can

be used in a LAN, their more common use is in the WAN.

DATA COMMUNICATIONS APPLICATION ISSUES

Several clear trends are shaping data communications networks. The most significant

is a decline in proprietary protocols in favor of TCP/IP. The shift to open

protocols gives users more vendor choices, which, in turn, lowers costs. Terminals

are disappearing as centralized databases move from mainframes to servers. The

dumb terminal of the past is now a desktop computer that has internal processing

power. As a result, popular office applications run on the desktop machine and

the database runs on a server. Client software on the desktop computer communicates

with the database. In some cases the client is proprietary, while in others it

is a Web browser. All of this renders the central computer–dumb terminal combination

obsolescent. Even operating systems are succumbing to the trend toward

openness as Linux migrates to the desktop and increases in popularity as a server

operating system. Manufacturers and developers, losing their proprietary advantages,

must distinguish themselves by providing additional features, services, and

improved setup routines.

The trend away from mainframes and terminals drives a similar transition

in the network. The applications for conventional packet switching and message

switching have shrunk to insignificance. Point-to-point circuits are still the choice

for many enterprise networks, but where in the past multiplexers were employed

to subdivide dedicated bandwidth, now bandwidths are increasing to support the

demands of the desktop and server environment.

Much of the discussion in this chapter has revolved around analog transmission

and modems. Although private line modems are fading from significance,

dial-up modems are still widely used. They are included with every laptop

and most desktop computers, and are used as backup for data private lines. By far

the bulk of data traffic uses digital transmission, the subject of the next chapter.

Nyamdavaa

Thursday, December 16, 2010

Job ad mon-eng

Saturday, December 4, 2010

Interview questions

Translation method

Chapter 4 Data Communications Principles

Blog Archive

About Me