370 (4)

Telephone Almanac, Bell Telephone System, 1959

Many of the early chapters of this book have explored the world of voice communications. During that exploration, it became clear that many of the devices and circuits in the Public Switched Telephone Network (PSTN) are digital in nature. In large part, the contemporary PSTN is a data network, transporting and switching voice data in digital format. After all, voice can be considered as the first in a long string of what, in a contemporary context, are considered to be data applications. This chapter introduces the basic concepts of the communications of computer data. Reflecting the reality of contemporary systems and networks, the balance of this book largely focuses on computer-to-computer data communications.

In order to comprehend the intricacy of contemporary data communications systems and networks, it is necessary to develop a solid understanding of certain basic concepts. These definitions and concepts extend across all technologies and service offerings, from the historical to the most contemporary. The historical is more than a footnote; it is important for purposes of understanding. Indeed, the contemporary and future networks are built on those fundamental concepts.

This chapter first addresses the concept of functional domains. I explain terminal equipment, communications equipment, and communication software and provide a detailed discussion of modems and DSUs/CSUs. I explore the concept of protocols at length, with discussion of basic protocol dimensions and issues. I also discuss computer network architectures, with emphasis on IBM's Systems Network Architecture (SNA) and the Open Systems Interconnection (OSI) reference model. Finally, I pause to consider the importance of security in data systems and networks.

DTE also is in the form of host computers such as mainframes and midrange (mini) computers. Host computers, also known as host nodes, are highly capable devices with substantial processing power and storage memory. Hosts also, at least theoretically, are carefully administered to ensure that they operate successfully and reliably. Hosts also serve as highly effective information repositories, with the data backed up and archived on external storage media such as magnetic tapes and CD-ROMs or on networked storage devices such as RAID (Redundant Array of Inexpensive Disks) systems. Note: In the DTE context, host computers and nodes are end-user endpoint (i.e., transmit and receive) equipment, rather that the various types of computer nodes that might be employed in networking capacity, such as the computers that run the IP protocol suite on the Internet.

Also note that the lines have blurred between categories of host computers. The PC on which I am writing this book is much more capable (and much less expensive) in many respects than was a typical mainframe only 15–20 years ago. Most of you would not even recognize a mainframe if you saw one. I would, but then I come from the days of heavy metal, when a forklift upgrade meant using a forklift to move out the old computer and move in the new one. The old one became a boat anchor or artificial reef. (This is true, except for the part about the boat anchor and artificial reef.)

Also known as Data Circuit Terminating Equipment (DCTE), Data Communications Equipment (DCE) is the equipment that interfaces the DTE to the network and resolves any issues of incompatibility between those domains in the process. Incompatibility issues can include digital versus analog, voltage level, signaling speed, and bit density. DCE includes modems, DSUs and CSUs, and Front-End Processors (FEPs), all of which I discuss in greater detail later in this chapter.

Alternatively, commercial communications management software can control and manage access to the host. IBM's Customer Information Control System (CICS) is such a product. This software resolves contention issues between diverse applications without impacting programs or terminals. It handles polling, selection, and program interrupts, thereby ensuring minimum response time. It also resolves error conditions at both the data and line levels. Random-Access Memory (RAM) maintains CICS and other Terminate-and-Stay-Resident (TSR) software.

Networks provide the connections between computer resources in order to accommodate the flow of information. Networks support the logical transfer of data during a communications session through the establishment of paths, circuits, or channels over a physical medium. The network can be in the form of a Local Area Network (LAN), a Metropolitan Area Network (MAN), or a Wide Area Network (WAN). These networks support communications over areas of increasing geographic scope. Data LANs, MANs, and WANs, which I discuss in later chapters, also can be interconnected.

Developed in support of voice communications, circuit switches serve for the flexi-ble interconnection of circuits. Once the circuit is established, a circuit switch dedi-cates a communications path for the duration of the session in support of a data transmission that is presumed to be in the form of a continuous stream. Because of their ubiquity, Central Office (CO) circuit switches remain widely used in support of dial-up data communications for Internet access. With that exception, most data communications takes place over networks that variously employ packet, frame, and cell switches.

Switches make forwarding decisions independently, link by link, although their activities commonly are coordinated by higher level system. Routers are highly intelligent switches that are capable of making path selection decisions across a network, from end to end and in consideration of perhaps a wide range of factors such as priority, protocol, and underlying application. Routers also can operate at much higher levels than switches, as they can be equipped to accomplish protocol conversions, manage security, and so on. I discuss packet, frame, and cell switching and routing in detail in later chapters. (Note: A switch is beneath a router, and network purists will find my characterization of routers as highly intelligent switches to be offensive, even in a descriptive sense. I reckon they'll get over it, eventually.)

6.2 DCE: EXPANDED VIEW

While I discuss data networks and switches in considerable detail in other chapters, this chapter is a convenient and meaningful place to pause and explore the concept and detail of several types of data communications equipment. Specifically, I want to examine modems, codecs, terminal adapters, CSUs and DSUs, and FEPs.

Modems mo dulate and demodulate signals. In other words, they change the characteristics of the signal in some way. Modems, as discussed in this chapter, are of several basic types: line drivers, short-haul modems, and conventional PSTN modems. The term also is used to describe a wide variety of other devices such as ISDN Terminal Adapters (TAs), ADSL modems, and cable modems used in CATV networks, each of which I explore in subsequent chapters.

Line drivers actually are interface converters, rather than modems in the classic sense. Line drivers extend the distance of a digital connection, within limits, by converting the digital signal to a low-voltage, low-impedance signal that can transmit more effectively and over longer distances on dedicated, specially conditioned twisted-pair circuits. The Recommended Standard (RS) 232 specification [more correctly known as Electronic Industries Alliance (EIA) 232], for example, generally limits the distance between devices to 50 ft at transmission rates of 56 kbps. At lower speeds, line drivers can reshape the digital pulses to extend that distance considerably. At speeds of up to 9.6 kbps, for example, line drivers can extend that limitation to 500–5000 ft over Category 3 Unshielded Twisted Pair (UTP). Line drivers can support speeds up to 19.2 kbps over distances up to 10,000 ft over good-quality Shielded Twisted Pair (STP) or Screened Twisted Pair (ScTP). You can extend the distance further through cascading line drivers. Note: Line drivers are unidirectional and work over simplex circuits. Bidirectional communications requires separate sets of line drivers operating over separate twisted-pair circuits.

Conventional modems allow digital devices to communicate across an analog circuit, accomplishing the digital-to-analog conversion in order to resolve that dimension of incompatibility between the DTE and the network. AT&T set the original de facto standards for modems with the introduction of the DataPhone service in 1959 [5, 6]. Currently, the ITU-T sets modem standards at an international level. Those standards fall into the V series, which includes all standards recommendations for data communications over the telephone network. The digital input to the modem is in the form of a baseband signal of varying electrical voltage levels that represent binary 1s and 0s. The output from the modem is a modulated analog carrier wave, which can be modulated in terms of its amplitude, frequency, phase, or some combination thereof. Through this process, the 1s and 0s of the data stream output by a digital computer can be sent over an analog voice network. While a network that is digital from end to end is preferred for reasons that include error performance and bandwidth, there often are analog components or links involved. In developed countries, of course, the core of the carrier network generally is fully digital in nature. In consideration of lower costs and greater availability, however, the local loop often is analog in residential and small-business applications. If the local loop is analog, the network is presented to the computer as analog. As the lowest common denominator rules, so to speak, the high-speed digital computer must adjust to the voice-grade analog local loop, and that is done through a modem.

It certainly is worth mentioning, and even emphasizing, that modems are unique in the wired world in terms of their portability, as modems work virtually anywhere. Contemporary road warriors largely would be lost without the advantage of modem access to the Internet from hotel rooms, airports, and client sites. While the modem speed may be constrained by the internal processes of an intermediate PBX, the device is absolutely indispensable for remote access to e-mail and other applications through the Internet and World Wide Web. In the United States and many other developed countries since 2000 or so, broadband access has become widely available in hotels, airports, coffee shops, and other public and private places through either hardwired or wireless Ethernet LANs. However, conventional modem access is an essential backup technique in the event that those options are not available. So, I travel with a conventional international modem (the modem standards are the same, but the physical and electrical connections vary from country to country), an Ethernet adapter, and a wireless Ethernet modem.

Also known as Amplitude Shift Keying (ASK), Amplitude Modulation (AM) involves the modulation of the amplitude (i.e., strength or voltage) of the analog carrier sine wave. The transmitting computer outputs a baseband signal, which is the digital transmission of electrical pulses, with 1 bits and 0 bits defined as discrete voltage levels. (Note: A baseband signal is a signal in its original form, without being altered in any way, whether by modulation or conversion.) Using a (single-bit) AM technique, each 1 bit entering the transmitting modem is expressed as a relatively high-amplitude sine wave, or series of sine waves. Each 0 bit is expressed as one or more low-amplitude sine waves, as illustrated in Figure 6.2. The high and low levels are defined in terms of a reference level or by the relative difference between the levels. At 2400 baud, this unibit technique yields a transmission rate of 2400 bps, with one bit transmitted per baud. Note: A bit has one of two possible values, a 1 or a 0. To express one bit with each baud, therefore, there must be 21 = 2 possible signal states.

Also known as Frequency Shift Keying (FSK), Frequency Modulation (FM) is the sole technique used in low-speed, Hayes-compatible modems. FSK involves the modulation of the frequency of the analog carrier sine waves (Figure 6.4). When no bits are transmitted, the carrier is at a reference frequency of 1700 Hz. A unibit FM technique impresses one bit on each baud by shifting the carrier to 2200 Hz when transmitting a 1 bit and to 1200 Hz when transmitting a 0 bit. At 2400 baud, therefore, the transmission rate is 2400 bps. The benefits of dibit transmission can be realized by defining four frequencies, with each sine wave or set of sine waves representing a two-bit pattern (00, 01, 10, and 11). Thereby, at 2400 baud, the transmission rate is 4800 bps.

Through the definition of eight phase shifts separated by 45 ° (0 °, 45 °, 90 °, 135 °, 180 °, 225 °, 270 °, and 315 °), contemporary modems can affect tribit transmission, achieving three bits of data per signal. (Note: 23 = 8.) Thereby, at a signaling rate of 2400 bps, the transmission rate is 7200 bps.

Differential Phase Shift Keying (DPSK) is a variation on the unibit PSK theme. With DPSK, as illustrated in Figure 6.6, each 1 bit triggers a 180 ° phase shift, but 0 bits have no effect. A great number of applications specify various PSK techniques. For example, Wi-Fi5 (802.11a) wireless LAN standards, which have nothing at all to do with the PSTN, call for BPSK at 6 Mbps and QPSK at 12 Mbps.

High-speed modems combine multiple modulation techniques. Quadrature Amplitude Modulation (QAM), for example, splits the carrier into two waveforms that are 90 ° out of phase and specifies two possible amplitude values for each of four phase shifts separated by 90 ° (0 °, 90 °, 180 °, and 270 °). This yields eight distinct signal states, as illustrated in the signal constellation graph in Figure 6.7. Thereby, each signal impulse, or symbol, carries one of eight possible signal combinations and represents three bits. (Note: 23 = 8.) At a signaling rate of 2400 baud, this tribit modulation scheme yields a transmission rate of 7200 bps.

Conventional modems can be characterized along a number of dimensions, including asynchronous versus synchronous, compression, diagnostics, error control, equalization, gain control, and band limitation:

· Asynchronous and synchronous modems both are available. Asynchronous modems transmit one character at a time, with the receiving device relying on start and stop bits to separate transmitted characters. Synchronous modems are much faster, as the signal is synchronized (timed) at the bit rate of the connection by a Transmit Clock (TC) in either the transmit modem or the transmit terminal. The paired modems synchronize on that clocking pulse in order to distinguish between blocks of data being transmitted, rather than identifying each individual character in a transmission and surrounding it with a start and stop bit. Particularly when transmitting large amounts of data, synchronous modems increase the efficiency of data transfer, resulting in increased speed of transfer and lower associated transmission cost. Synchronous modems faithfully transmit any bit sequence, rather than just ASCII characters.

· Diagnostic modems can test their internal clock and transmit and receive circuits. Additionally, such modems may have the capability to monitor their performance and even diagnose certain conditions contributing to performance degradation. Further, they can respond to loopback tests and, therefore, are manageable through higher level Element Management Systems (EMSs). Such management systems typically are located remotely and are capable of managing large numbers of modems and modem pools (i.e., groups of modems to which access is shared among multiple users).

· Error correction capabilities are included in most contemporary modems. The proprietary Microcom Networking Protocol (MNP) was among the first to include error correction and subsequently was incorporated into the ITU-T V.42 standard. While proprietary error correction software remains embedded in certain modems, the ITU-T V.42 and subsequent generations of modems have standardized this function.

V.90 modems are asymmetric in nature, providing a maximum of 56 kbps downstream (i.e., from the network to the modem) and 33.6 kbps upstream (i.e., from the modem to the network). In order to achieve this level of performance, a 56-kbps modem configuration (Figure 6.9) requires that only one transmission link be analog. The end-user dial-up connection is through a V.90 modem over an analog local loop to the PSTN. The connection through the PSTN must be entirely digital, including the originating and terminating COs, all tandem offices, and all transmission facilities. At the terminating end (e.g., corporate intranet site or ISP), the local loop connection must be digital (e.g., T-carrier, E-carrier, or ISDN). Matching 56-kbps modem technology must be in place at the terminating device, typically in the form of an access server or router. This configuration limits the transmissions to only a single D-to-A-to-D (Digital-to-Analog-to-Digital) conversion process, which limits the amount of quantizing noise associated with the D-to-A process. This results in higher speed transmission without sacrificing error performance.

Notably, 56-kbps modems also are V.34+ modems. Assuming that the terminating modem is V.34+ or lesser, the 56 kbps modem "falls back" to that standard, which supports symmetric transmission at speeds up to 33.6 kbps. (Note that the lowest common denominator always determines the maximum level of performance.) Also notably, the fine print on the boxes of the 56-kbps modems indicates quite clearly that they currently are limited to 53.3 kbps because they cannot exceed FCC amplitude (signal strength) limitations. (Note, again, the highest common denominator rule, referring to the network, in this case.) The FCC established this limitation many years ago to minimize the likelihood that transmissions on a cable pair might cause interference on adjacent pairs in the same cable sheath. While the FCC is considering relaxing this restriction, it is clear that the modem manufacturers, rather than the FCC, are responsible for the fine print on the modem boxes [7–11].

It was fully anticipated that V.90 would be the last in a long line of modem standards, but modem technology got a boost with the ITU-T standardization of V.92 (November 2000). Based on work done at Hughes Network Systems for its DirecPC (now HughesNet) satellite Internet service, V.92 offers several key advantages over V.90:

· Faster Connection: QuickConnect is a feature that cuts the time required for handshaking approximately in half, to about 10–15 s. QuickConnect trains the modem on the first call and remembers the characteristics of the circuit. Assuming that the circuit is the same on the next call, the circuit characteristics do not have to be relearned, which results in faster connect times, which has obvious advantages to the end user. It also offers advantages to the ISPs, which may handle thousands, or even millions, of dial-up calls a day—the faster you connect, the faster you take care of your business and disconnect, the shorter the period of time you tie up an expensive port, and the quicker it is available for another user.

· Faster Upstream Speed: Upstream transmission speed is increased from 33.6 to 48 kbps under optimum conditions using a variation of PCM that allows the upstream data stream to use the same clocking source as the downstream data stream.

· Modem on Hold (MoH): A V.92 modem can put a data session on hold when it detects a voice call, either incoming or outgoing, through a call waiting indication and gracefully resume that session when the voice call is terminated. Thereby, V.92 allows a single analog line to be used for both voice and data. MoH requires that the line be equipped with call waiting, a telco PSTN service.

The announcement of V.92 certainly is not as dramatic as that of V.90, which nearly doubled downstream transmission rates. It does, however, offer considerable advantages for dial-up users, including not only those without broadband access via ADSL, cable modem networks, or satellite but also those of us who use modem connections on the road or as a backup in the event of a broadband access failure. Essentially, that means that all users stand to benefit at one time or another once V.92 modems are widely available and widely supported by ISPs. Unfortunately, there seems to be no market driver significant enough to compel the ISPs to expend the capital necessary to upgrade V.90 servers to V.92. Some V.90 modems can be upgraded to V.92 via a network download [12–14].

The reverse conversion of A-to-D is necessary in situations where analog information is to be sent across a digital circuit. Certainly, this often is the case in carrier networks where huge volumes of analog voice are digitized and sent across high-capacity digital circuits. This requirement also exists where high-capacity digital circuits connect premises-based, analog voice PBXs or Key Telephone Systems (KTSs) to Central Office Exchanges (COEs) or to other PBXs or KTSs.

The device that accomplishes the A-to-D conversion is known as a codec. Codecs code an analog input into a digital (data) format on the transmit side of the connection, reversing the process, or decoding the information, on the receive side to reconstitute the analog signal. Codecs are used widely to convert analog voice and video to digital format and to reverse the process on the receiving end. There exists a wide variety of codecs for use in wired and wireless networks of all descriptions. Many of those codecs are discussed in detail in subsequent chapters. [Note: In contemporary systems, a codec commonly is incorporated into a Digital Signal Processor (DSP).]

Channel service units are circuit-terminating equipment that provide the customer interface to the circuit, as illustrated in Figure 6.10. They also permit the isolation of the DTE/CPE from the network for purposes of network testing. CSU functions include electrical isolation from the circuit for purposes of protection from aberrant voltages, serving the same function as a protector in the voice world. Additionally, the CSU can respond to a command from the carrier to close a contact, temporarily isolating the DTE domain from the carrier domain. This enables the carrier to conduct a loopback test in order to test the performance characteristics of the local loop from the serving CO to the CSU and back to the CO. Many contemporary CSUs also have the ability to perform various line analyses, including monitoring the signal level. Such intelligent CSUs also often have the ability to initiate loopback tests, although arrangements must be made with the carrier in advance.

The CSU also serves to provide signal regeneration and generates keep-alive signals to maintain the circuit in the event of a DTE transmission failure. Finally, the CSU stores various performance data in temporary memory for consideration by an upstream network management system.

Smart CSUs increasingly are positioned as Integrated Access Devices (IADs). These multiport devices support interfaces to voice, data, and video devices such as PBXs, routers, and videoconferencing units. The programmable IAD supports bandwidth allocation for the various devices, enabling them to share a single T1 or other digital facility.

Data service units convert the DTE unipolar signal into a bipolar signal demanded by the network. DSU functions variously include regeneration of digital signals, insertion of control signals, signal timing, and reformatting. Some of these functions can be ceded to either the CSU or the terminal equipment [15]. In any case, the functions must be performed, even though the CSUs and DSUs lose their identity.

Front-End Processors (FEPs) combine the functions of concentrators and message switches. In other words, they have the ability to concentrate and switch traffic between multiple terminals and among groups of terminals in order to share a single circuit for access to mainframe resources. They also serve as an interface to Wide Area Network (WAN) circuits to provide remote terminals access to mainframe resources. Most FEPs are midrange computers that have their own databases to support assigned functions such as error detection and correction, queuing, editing validation, and limited application processing. While the mainframe clearly could perform such tasks, it is more cost effective to apply a lower order computer to the performance of such mundane and highly repetitive tasks, thereby reserving the power of the mainframe for more difficult and demanding tasks in support of user-oriented applications.

In a more contemporary client/server environment there is no exact equivalent of an FEP. Although the same functions must be performed, they are spread across multiple and various devices such as an access router with a firewall for security purposes and a Remote Access Server (RAS) or Virtual Private Network (VPN) server for performance of security functions and granting of access privileges.

6.3 PROTOCOL BASICS

· Handshaking is the sequence that occurs between the devices over the circuit, establishing the fact that the circuit is available and operational. The handshaking process also establishes the level of device compatibility and determines the speed of transmission by mutual agreement. Devices accomplish the process of handshaking by passing frames of data back and forth in order to negotiate the basis on which they will communicate, in consideration of the performance characteristics of the circuit. As always, the rule of the lowest common denominator applies and, therefore, communication is limited by the least capable device. Actually, the rule is more like the highest common capability of the least capable device determines the maximum level of performance.

· Line discipline is the sequence of network operations that actually transmits and receives the data, controls errors in transmission, deals with the sequencing of message sets (e.g., packets, blocks, frames, and cells), and provides for con-firmation or validation of data received.

Several hardware and software solutions are designed to specifically deal with protocol analysis and conversion:

· Protocol converters are software-based devices that translate from one native protocol into another [e.g., from ASCII to EBCDIC, from TCP/IP to IBM SNA/SDLC, or from PCM voice to VoIP].

· Gateways in ARPANET terminology, were packet-forwarding devices that, in contemporary data communications terminology, are referred to as routers. Contemporary gateways are hardware/software combinations that connect devices running different native protocols. In other words, gateways are packet-forwarding devices that run gateway protocols for purposes of protocol conversion, perhaps at all layers of the OSI Reference Model (i.e., from the Physical Layer to the Applications Layer). In addition to protocol conversion, gateways provide a point of physical interconnection between incompatible networks, with examples including X.25-to-Frame Relay (FR) gateways and T-carrier-to-E-carrier International Gateway Facilities (IGFs).

· Protocol analyzers are diagnostic tools for displaying and analyzing communi-cations protocols. Analyzers enable technicians, engineers, and managers to test the performance of the network to ensure that the systems and the network function according to specifications. LAN managers, for example, use protocol analyzers to perform network maintenance and troubleshooting and to plan network upgrades and expansions.

A very basic protocol issue involves the manner in which the circuit is set up between devices. There are three alternatives: simplex, Half DupleX (HDX), and Full DupleX (FDX), as illustrated in Figure 6.11 and defined below.

Full duplex (FDX) is a fully bidirectional transmission mode in which communication is supported in both directions simultaneously. The first working FDX communications circuit was invented by Joseph B. Stearns of Boston and installed in 1872 on a one-wire telegraph system using a ground return. This effectively doubled the traffic capacity of the network, and the duplex equipment could be purchased and installed at much lower cost than stringing another wire [1]. Human beings, including telegraph operators, are born fully equipped for FDX communications, but computers require software to enable this capability. While we would rarely employ a one-wire approach these days, there are several different ways to configure an FDX circuit.

· Physical Four Wire: The most commonsense FDX configuration involves a physical four-wire circuit with one pair of wires supporting transmission in one direction and another pair supporting transmission in the other direction. This approach is commonly used for high-capacity, dedicated circuits, of which most are multichannel in nature. T-carrier and E-carrier circuits traditionally were configured in this way. All wideband and broadband circuits are FDX in nature.

· Physical Two Wire: FDX modems use another approach in order to achieve FDX data transmission on a physical two-wire circuit. Modems cause a two-wire circuit to emulate a four-wire circuit through a split-channel approach, using Frequency Division Multiplexing (FDM) to create two frequency channels. This process allows the creation of two carrier waves that are then modulated using Frequency Modulation (FM). Figure 6.12 illustrates this technique over a voice-grade circuit. In this example, one channel uses an unmodulated carrier wave at 1170 Hz, modulating it at 1070 Hz to represent 0 bits and at 1270 Hz to represent 1 bits. The reverse channel uses an unmodulated carrier wave at 2125 Hz, modulating it at 2025 Hz to represent 0 bits and at 2225 Hz to represent 1 bits. Other modulation techniques can be used within these channels as well, including AM, PSK, and QAM.

· Optical: FDX transmission in optical fibers is achieved either by using separate fibers for transmission in different directions or by using Wavelength Division Multiplexing (WDM), which essentially is FDM at the optical level.

FDX circuits sometimes are used to connect HDX terminals in order to avoid issues of line turnaround time. More typical examples of FDX applications include channel links between host processors, channel links between controllers/ concentrators and hosts, and other applications involving the interconnection of substantial computing systems. Carrier services that deliver FDX capabilities include DDS, E/T-Carrier, and broadband services such as Frame Relay, SMDS, and ATM. I discuss all of these in later chapters.

There are two basic methods of data transmission: asynchronous and synchronous. As long as we are on the theme of timing, we will briefly explore the concepts of isochronous and plesiochronous communications as well.

PCs, teletypes, and other devices that make use of asynchronous transmission frame, or surround, each byte of information with start and stop bits, which are interpreted by the receiving terminal and subsequently stripped away in order to get to the actual data payload. The inclusion of start and stop bits adds two or three bits of overhead to the transmission of each eight-bit byte. Additionally, asynchronous transmission commonly involves the addition of a parity bit as an error control mechanism, which happens to be relatively poor, as we shall see later in this chapter. The framing of the data with these three or four bits of control information yields an overhead, or inefficiency, factor of 20–30 percent.

From Latin and Greek origins, synchronous translates as together with time. Such transmission is message framed and overcomes the inefficiencies of asynchronous, start–stop transmission for high-speed data communications applications. Rather than surrounding each character with start and stop bits, a relatively large set of data is framed, or blocked, with one or more synchronization bits or bit patterns used to identify the beginning and end of a logical block of data. Both analog synchronous modems and digital DTE synchronize the receiving ter-minal on the rate of transmission of the data from the sending terminal. Through the receipt of the synchronizing bits, or clocking pulses, the receiving device can match its speed of data receipt to the rate of data transmission across the circuit. Thereby, each bit of data and control information can be distinguished at the physical layer. Higher layers sort out when to expect what information, in which data fields, and in what sequence, based on an agreed-upon protocol such as Frame Relay I P. Because only a few framing bits and synchronizing bits surround a large block of data, the overhead is much reduced, the efficiency of transmission is much increased, and the effective throughput is much improved. As stated in Wonders of the Universe (The Werner Company, 1899):

The Morse printing instrument … is a beautiful, but rather complicated piece of mechanism, for besides the printing and electric arrangements it is furnished with clockwork to keep the paper tape in motion whilst the message is being delivered. But lately these accessories have in many cases been dispensed with, and the operator depends upon his ear for the translation of the message sent.

Isochronous (Isoc) data are synchronous data transmitted without a clocking source. From the Greek isochronos, translating as equal in time, all bits are of equal importance and are anticipated to occur at regular intervals of time. Bits are sent continuously, with no start–stop bits for timing. Rather, timing is recovered from transitions in the received data stream, with a whole number of bit-length intervals between characters. Bit integrity is preserved, with no modifications (i.e., bipolar conventions). The transparent isochronous transmission does not recognize control characters. Some T-carrier nodes operate isochronously on the links, syncing up with several lines operating at slightly different speeds. Isoc often is used in secure military applications that require encryption.

From the Greek plesio, meaning near, and chronos, meaning time, plesiochronous communications involves devices running at nominally the same rate. Plesiochronous networks comprise subnetworks and devices that are free running, although at approximately the same rate and within defined parameters of tolerance for variation. Much like clocks and watches run at approximately the same rate, devices in a digital network are free running, with some running at a slightly faster or slower pace. So, the network is almost, but not perfectly, synchronized across all devices.

Analogous to alphabets, code sets, or coding schemes, are employed by all computer systems to create, store, and exchange information. While code sets vary, they all rely on a specific combination of 1s and 0s of a specific total length in order to represent something of value, such as a letter, number, punctuation mark, or control character (e.g., carriage return, line feed, space, blank, and delete). Contemporary standard coding schemes include Baudot, ASCII, EBCDIC, and Unicode.

The first widely accepted standard coding scheme was Morse code, invented by Samuel Morse sometime prior to 1844 for use in telegraphy, which, of course, is a human-to-human digital data communications method. Friedrich Clemens Gerke invented the International Morse Code in 1848 out of necessity, as some of the spaces in letters created difficulty in radiotelegraphy. The international version was standardized by the International Telegraph Union (ITU) in 1865 and is still used today by amateur radio operators, or hams. Morse code uses series of short and long marks in the form of dots (short marks) and dashes (long marks), with spaces between them, to represent letters, numbers, punctuation marks, and procedural signals (prosigns). The length of the spaces varies, with a short space between letters, a longer space between characters, a longer space between words, and a still longer space between sentences. In order to speed transmission, the fewest number of dots and dashes represent commonly used letters (e.g., E is •, T is –, A is • –). Commonly used words are abbreviated (e.g., calling is abbreviated CG, or – • – • – – •, as are commonly used phrases (e.g., love and kisses is abbreviated 88, or – – – • • – – – • •. Table 6.1 provides the International Morse Code for English letters, numbers, and select punctuation marks and prosigns. (Note: The terms ham and plug originated in tele graphy to describe a telegraph operator lacking in proficiency [2]. Ham alludes to a ham-fisted operator.)

Morse code was the primary communication code for many years, until Emile Baudot invented the Baudot Distributor in the 1870s. That device provided for the transmission of values in a five-bit coding scheme over a line between two synchronized electromechanical devices. The Baudot Distributor soon gave way to the teletype, which also was based on the Baudot coding scheme, subsequently known as International Telegraph Alphabet 2 (ITA 2). Telephone Devices for the Deaf (TDDs) still use ITA 2.

Baudot employs asynchronous transmission, as start and stop bits separate characters. Error detection and correction require human editing. Therefore, Baudot is a human-to-human, rather than a machine-to-machine, communication technique. Detected errors must be corrected through retransmission.

Clearly, Baudot is a highly limited coding scheme. The limited range of letters expression, at 32 characters, is barely enough to accommodate the relatively simple English, French, and Spanish alphabets. To stay within that range, all letters must be in uppercase, which is just as well, as the shift key is used for other purposes. The shift key provides another 32 characters for numbers and control characters, but Baudot is still very limited in its range of expression. Additionally, the asynchronous requirement for start and stop bits makes Baudot overhead intensive. Finally, the error detection and correction technique is far less than desirable. As a result, Baudot currently is limited to use in teletypewriters and very old telex machines.

As a footnote, limited coding schemes are not necessarily overly limiting. Proprietary five-and six-bit codes, for example, have been used in the airline reservation systems (e.g., American Airlines' SABRE System and United Airlines' APOLLO) for many years. The airlines still use the IBM Airline Control (ALC) protocol, or the Unisys version, P1024B, both of which used a six-bit coding scheme. Such applications involve a limited character set that is easily accommodated by a six-bit code; in fact, a seven-or eight-bit coding scheme would be excessive and inefficient.

EBCDIC involves an eight-bit coding scheme, yielding 28 (256) possible combinations and, thereby, significantly increasing the range of expression. As a result, more complex alphabets can be supported, as can upper-and lowercase letters, a full range of numbers (0–9), and all necessary punctuation marks. Equally importantly, if not more so, the eight-bit coding scheme supports a large number of control characters, which is critical in the coordination of communications between complex mainframe computers.

The EBCDIC-based machines communicate on a synchronous basis, thereby improving on the speed of transmission. Since start and stop bits do not surround each character, overhead reduces, efficiency of transmission improves, and more payload bits can transmit per unit of time. Further, a more complex, machine-to-machine error detection and correction technique yields improved performance in that regard. Detected errors may require retransmission, although forward error correction is often employed, with the receiving system identifying, isolating, and correcting the errored bits.

Developed in 1963, the American Standard Code for Information Interchange (ANSCII or ASCII) was specifically oriented toward data processing applications. It was modified in 1967 by the American National Standards Institute (ANSI) to address modifications found in contemporary equipment; that version, originally known as ASCII II, is now known simply as ASCII [16].

ASCII employs a seven-bit coding scheme, supporting 128 (27) characters, which is quite satisfactory for the English alphabet, Arabic numerals, punctuation marks, a reasonable complement of special characters, and a number of control characters, as displayed in Table 6.2. (Note: Actually, the Indians invented numbers 1–9. The Arabs invented only 0 but introduced 0–9 to Europe.) As ASCII was designed for use in asynchronous computer systems (non-IBM, in those days), fewer control characters were required, making a seven-bit scheme acceptable.

Figure 6.13 illustrates ASCII in the context of asynchronous communications, with start and stop bits framing each character, without employing synchronization bits. Asynchronous transmission makes use of a simple error detection and correction scheme known as parity checking, also illustrated in Figure 6.13. Parity checking is error prone with detected errors often going unnoticed or requiring retransmission, although forward error correction may be employed.

Unicode accommodates preexisting standard coding schemes using the same byte values for consistency. In Unicode terms, ASCII, for example, is known as UTF-7. There even is a UTF-EBCDIC, specifically for IBM mainframes. UTF-8 supports any universal character in the Unicode range using one to four octets (eight-bit bytes) to do so, depending on the symbol. UTF-32 uses four octets for each symbol but is rarely used due to its inherent inefficiency. The Unicode standard has been adopted by companies such as Apple, H P, IBM, Microsoft, Oracle, SAP, Sun, Sybase, and Unisys. Unicode is required by modern standards such as CORBA 3.0, European Computer Manufactures Association ECMAScript (JavaScript), Java, Lightweight Directory Access Protocol (LDAP), Wireless Markup Language (WML), and eXtensible Markup Language (XML).

Unicode is developed in conjunction with the International Organization for Standardization (ISO). ISO 10646 defines the Universal Character Set (UCS), into which the UTF code sets map. UCS-4 is a four-octet code set into which UTF-32 maps and UCS-2 is a two-byte code set into which UTF-16 maps. UCS-1 encodes all characters in byte sequences varying from one to five bytes.

· Header: A communications header precedes the data to be transmitted, often serving to establish the fact that the transmission link exists both physically and logically. The header may include synchronization bits that provide for synchronization between the devices and the link. Address fields in the header often include both source address and destination address. The source address identifies the originating device so that a response can be directed correctly and so that a retransmission can be requested if the data suffer errors during transmission. The destination address enables the receiving device to identify data intended for it and enables intermediate devices (e.g., switches and routers) to route the data correctly. Certain data protocols also use fields in the header to identify the length of the text field and the type of data, to indicate the level of tolerance for delay or loss during network transit, and any optional headers that might follow. The user header includes user-definable information such as system access (password), organization or department ID, operator ID, terminal ID, database or application ID, destination address, message sequence number, date/time ID, and message priority.

· Text: The text portion of the data set is the information to be communicated. It may contain either a fixed or a variable amount of information, depending on the specific protocol involved. The text may be preceded by Start-of-TeXt (STX) and succeeded by End-of-TeXt (ETX) control characters so the receiving device can determine the location of the message data. The text field also is known generically as the data field, or the payload.

· Trailer: The trailer, tail, or trace portion of the data set contains information relative to the analysis of the message, including message tracking and diagnostics. Trailing the text, the trailer information may contain the originating ID, the data block number and total number of blocks being transmitted, and identification of system processing points involved in the transmission. The trailer often includes an error detection and correction mechanism to manage the integrity of the transmitted data.

· Recognition and flagging provide for no automatic means of correction of errors. Used primarily in networks involving dumb terminals with no means of buffering or retaining information transmitted; retransmission of errored data is not possible. Detected errors simply are flagged (identified) as such by the receiving device; error correction requires a human-to-machine request for retransmission. Parity checking is an example of recognition and flagging.

Echo checking is one of the earliest means of error detection and correction. The receiving device echoes the received data back to the transmitting device. The transmitting operator can view the data as received and echoed, making corrections as appropriate. But errors also can occur in the transmission of the echoed data, making this approach highly unreliable.

Parity checking is by far the most commonly used method for error detection and correction because it is used in asynchronous devices such as PCs. Parity refers to the number of marks, or 1 bits, in a character. The network can be set for either odd parity or even parity. Once set, the networked devices always create odd or even data values, character by character or set by set. This less than ideal approach is implemented easily and offers reasonable assurance of data integrity. Parity checking has two dimensions: Vertical Redundancy Checking (VRC) and Longitudinal Redundancy Checking (LRC).

Vertical redundancy checking entails the appending of a parity bit at the end of each transmitted character or value to create an odd or even total mathematical bit value. The letter C, for example, in ASCII, is coded as a bit sequence of 1100001, which is an odd number of marks, or 1 bits. So the parity bit would be a 0 if the network or link is set for odd parity, as that would create an eight-bit byte with the sequence 11000010, thereby retaining the odd parity value. Alternatively, the parity bit would be a 1 if the network is set for even parity, as that would create an eight-bit byte with the sequence 11000011, thereby creating an even parity value. The receiving device executes the same mathematical process to verify that the correct total bit value was received—hence the use of the terms redundancy and checking. Speaking in terms of the logical manner in which humans add numbers physically positioned in columns, the two devices sum the bit values vertically, as represented in Figure 6.15—hence the use of the term vertical. While inexpensive and easily implemented in computers employing asynchronous transmission, this approach is highly unreliable, as two errored bits in a character can yield an undetectable error in a character. Further, VRC provides no inherent means of error correction; VRC often is characterized as send and pray. According to Maria Price La Touche (1824–1906):

There is no greater mistake than to call arithmetic an exact science. There are … hidden laws of number which it requires a mind like mine to perceive. For instance, if you add a sum from the bottom up, and then again from the top down, the result is always different.

The technique of block parity improves considerably on simple parity checking. While Spiral Redundancy Checking (SRC) and interleaving improved on the detection of errors due to increased transmission speeds and more complex modulation techniques, they gave way to Cyclic Redundancy Checking (CRC), which is commonly employed today.

CRC validates transmission of a set of data, formatted in a block or frame, through the use of a statistical sampling process and a unique mathematical polynomial, both of which are known to the transmitter and receiver. The transmitting device statistically samples the data in the block or frame and applies a 17-bit generator polynomial based on an Euclidean algorithm. The result of that calculation is a description of the text field, which is appended to the block or frame or text as either a 16-or 32-bit value. The receiving device executes the identical process, comparing the results of its process to the CRC value appended to the data block. The result is an integrity factor of 10-14; in other words, the possibility of an undetected error is 1 in 100 trillion. By way of example and at a transmission speed of 1 Mbps, one undetected error is expected approximately every 30 years!

An unerrored block or frame is ACKnowledged by the receiving device through the transmission of an ACK, whereas an errored block or frame is Negatively AcK-nowledged with a NAK. A NAK prompts the transmitting device to retransmit that specific block or frame, which has been stored in buffer memory. The transmission of an ACK by the receiving device cues the sending device that the block or frame of data can be erased from buffer memory and the next block or frame of data then can be sent.

CRC is relatively memory and processor intensive, but it is easily accommodated in high-order computers that benefit from synchronous transmission techniques. As CRC ensures that data transmission is virtually error free, it is considered mandatory in most sophisticated computer communications environments.

Forward Error Correction (FEC) involves the addition of redundant information embedded in the data set so the receiving device can detect errors and correct them without requiring a retransmission [18]. The two most commonly employed techniques are Hamming and BCH (Bose, Chaudhuri, and Hocquengham).

As the length of the data sets increases, the distances over which they travel increase, and the likelihood of errors in transmission increases accordingly, data compression becomes sensible. Additionally, data compression can significantly reduce the bandwidth required to transmit a set of data. Regardless of the level of bandwidth available in even the most capable networks, bandwidth always has an associated cost. Data compression techniques can include formatting, redundant characters, commonly used characters, and commonly used strings of characters:

· Formatting of the data need not be transmitted across the network. In a basic example, data compression might involve the removal of formatting from a commonly used form, such as an expense report. Such formatting can involve a large amount of redundant data because the receiving device can reformat the data easily, placing the various fields of data in the appropriate places on the form, which resides in memory. An excellent example is that of access to the Internet and the Web through client/server software such as America Online, Netscape Navigator, or Microsoft Internet Explorer. In each case, many of the graphic-intensive screens are stored on the client workstation. When accessing the various Internet portals, therefore, you do not need to download the full set of graphics. This process is extremely bandwidth intensive and, therefore, ultimately translates into long delays and higher costs. Rather, only the updated information must be downloaded.

· Redundant data can be identified easily by the transmitter and communicated to the receiver. This approach is also known as string coding, yielding compression factors of as much as 4: 1. An excellent example is that of fax modem compression algorithms, which use various methods of run-length encoding. As transmitting fax modems scan a document from left to right and from top to bottom, they can quickly sense a run of whitespace. Then, the transmitting modem notes that "nothing" is being transmitted and notes the length of the run of nothing, all in a few bits stored in an internal buffer. Once some "real" data appear, the modem notes this fact and begins to send corresponding bits to the internal buffer. After a specific number of bits are stored in the buffer, the modem packs them into a frame that it transmits. As runs of nothing and runs of real data reoccur, the modem recognizes that fact and adjusts accordingly. Very quickly, therefore, the transmitting modem can transmit a document with lots of whitespace, with that transmission requiring very little bandwidth through the supporting network. (Try faxing a white sheet of paper. Then try faxing a document of very dense text. You will immediately see the difference in transmission time.)

· Commonly used strings of characters similarly can be identified and transmitted in abbreviated form. Such an approach relies on the probability of character occurrence following a specific character (e.g., Q is generally followed by U). Markov source and other techniques address this potential.

Asynchronous Data Link Control (DLC) protocols are used primarily for low-speed data communications between PCs and other very small host computers. Framing occurs at the byte level, with each byte surrounded by a start bit (a 0 bit) and a stop bit (a 1 bit). A parity bit often accompanies each character as well. Telex transmission incorporates an additional stop bit.

Kermit and XMODEM are asynchronous protocols, organizing information into 128-byte packets. Kermit also uses CRC error control. The data also can be blocked at the application level, and adding the technique of LRC can complement VRC for improved error control.

Two general types of data communications protocols exist—byte oriented and bit oriented. While the performance characteristics of byte-oriented protocols are acceptable for many applications, bit-oriented protocols are much more appropriate for communication-intensive applications in which the integrity of the transmitted data is critical.

Byte-oriented protocols require an entire byte to communicate a command signal to the target station. Byte-oriented protocols communicate value strings in byte formats, generally of eight bits per byte. Control characters are embedded in the header and trailer of each byte or block of data. As byte-oriented protocols are overhead intensive, they are used exclusively in older computer protocols at the second layer, or link layer. Byte-oriented protocols generally are asynchronous and HDX, operating over dial-up, two-wire circuits. One example includes Binary Synchronous Communications (Bisync, or BSC).

Bit-oriented protocols can change a single bit with a frame control byte to send a different command to a target station. Bit-oriented protocols transmit information in a much larger bit stream, with opening and closing flags identifying the separation of the text from the control information, which addresses control issues associated with the entire data set. The much less overhead-intensive, bit-oriented protocols are usually synchronous and FDX and operate over dedicated, four-wire circuits. Examples include IBM's Synchronous Data Link Control (SDLC) and the ISO's High-level Data Link Control (HDLC).

The Bisync block consists of synchronizing bits, data, and control characters sent in a continuous data stream block by block. While there are six basic Bisync block formats, the elements of a generic Bisync block, as illustrated in Figure 6.16, are as follows and in sequence [19–21]:

· PAD: A PA D ding character (hexadecimal 55) may be sent as the first character to alert the receiving device of the transmission of a block of data and to ensure that the receiving device is in step with the data bits.

· SYN: SYN chronizing characters (usually two) establish character synchronization between the transmission and the receiving devices. There must be enough bit transitions to allow the receiver to confirm the bit rate.

· SOH: A Start-Of-H eader control character precedes the routing information.

· Header: The Header field contains one or more octets of information indicating the address of the transmitting device.

· STX: A Start-of-TeXt control character indicates the beginning of the data.

· Text: The Text field, aka payload, is the data being transmitted. This field can be up to 512 octets in length.

· ETX: An End-of-TeXt control character indicates the end of the data.

· BCC: Block Check C haracters detect errors. There are one or two BCCs.

· EOT or PAD: An End Of T ransmission character or PADding character (hexadecimal FF) character trails the transmission to ensure the receipt of all previous characters and to indicate the end of the block.

Synchronous Data Link Control (SDLC), developed in the mid-1970s, is at the heart of IBM's System Network Architecture (SNA). SDLC is a bit-oriented, point-to-point protocol that uses bit strings to represent characters. SDLC uses CRC error correction techniques—specifically known as Frame Check Sequence (FCS) here. SDLC supports high-speed transmission and generally employs FDX, dedicated circuits. SDLC works either in HDX or FDX, supports satellite transmission protocols, and works in point-to-point or multipoint network configurations.

Up to 128 frames can be sent in a string, with each frame containing up to seven blocks, each up to 512 characters. Each block within each frame is checked individually for errors. Errored blocks must be identified as such to the transmitting device within a given time limit or they are assumed to have been received error free. As a carefully timed, point-to-point protocol, SDLC depends on high-performance circuits, usually in the form of dedicated leased lines.

The SDLC frame consists of synchronizing bits, data, and control characters sent in a continuous data stream frame by frame. The specific elements of the SDLC frame (Figure 6.17) are as follows and in sequence [19–21]. The same format applies to High-level Data Link Control (HDLC) frames and X.25 packets:

· Flag (F): Flag bits, in a specific eight-bit pattern, alert the receiving device to the transmission of the frame, thereby initiating the error-checking procedures. The most commonly used flag character is 01111110 (7E in hexadecimal).

· Station Address (A): This address field of eight bits identifies the specific target device for which the frame is intended, a group address for multiple target terminals, or a broadcast address to all terminals. This field also can be used to distinguish commands from responses.

· Control (C): The eight-bit control field identifies the type of frame being transmitted. An information frame is used for the transfer of messages, frame numbering of contiguous frames in a message, and so on. A supervisory frame is used for purposes such as to indicate a detected error in transmission, acknowledge that frames have been received without error, request the transmission of specified frames, and order the transmitting device to stop sending.

· Frame Check Sequence: This 16- or 32-bit field contains the CRC character sequence used to check the integrity of the transmitted address and control information as well as the data.

· Flag (F): Flag bits, in a specific eight-bit pattern, alert the receiving device to the end of transmission of the frame, thereby terminating the error-checking procedures. The most commonly used flag character is 01111110 (7E in hexadecimal).

Considering that the information field can contain as many as 4096 octets of payload and that the control fields involve only 6 or 8 octets (8 at full payload), the SDLC frame can be as much as 99.8 percent efficient, at least theoretically. There is some overhead in the information field, of course, but even then the SDLC frame is 95–97 percent or so efficient, which is remarkable.

The ISO developed HDLC as a superset of IBM's SDLC and the U.S. National Bureau of Standards' (NBS) (now National Institute of Standards and Technology, or NIST) Advanced Data Communications Control Procedures (ADCCP). A version of HDLC is the Link Access Procedure—Balanced (LAP-B), which is used in packet-switched networks conforming to the ITU-T X.25 Recommendation. While HDLC was built on SDLC and is very similar, the two generally are not compatible, depending on the framing conventions in the specific HDLC implementation.

6.4 NETWORK ARCHITECTURES

A network architecture defines the communications products and services that ensure the various components can work together. Early on, even the various systems of a given manufacturer did not interoperate, let alone afford connectivity with the products of other manufacturers. While IBM's Systems Network Architecture (SNA) and the Digital Equipment Corporation's Digital Network Architecture (DNA), aka DECnet, architectures solved these internal problems, they still did not interoperate. Truly open-system architectures still remain in the distant future, although great strides have been made in this regard through the Open Systems Interconnection (OSI) model fostered by the International Organization for Standardization (ISO).

Layered models serve to enhance the development and management of a network architecture. While they primarily address issues of data communications, they also include some data processing activities at the upper layers. These upper layers address application software processes, presentation format, and establishment of user sessions. Each independent layer, or level, of a network architecture addresses different functions and responsibilities. The layers work together, as a whole, to maximize the performance of the process. The various functions address the functions of data transfer, flow control, sequencing, error detection, and notification.

Data transfer enables the transfer of data from one node to another. Included are such issues as normal or expedited data flow; packet, block, or frame sizing; and data assembly and segmentation. Flow control controls the pace of packets through the network to prevent a device or link from being overwhelmed with data and to reduce congestion and the resulting degradation of network performance. Sequencing of the data packets is required where packets may take different routes between nodes. Error detection is required to ensure data integrity. Notification provides for the advice from receiver to transmitter of the receipt of packets and their condition.

Developed in 1974 by IBM, SNA was a five-level design architecture that has grown into a seven-layer model. Although similar, SNA is not compatible with the OSI model. SNA comprises software and hardware interfaces that permit various IBM systems and software to communicate. SNA includes network nodes, physical units, and logical units:

· Nodes are physical devices in the SNA network. Nodes can include computers, communications processors (e.g., FEPs), terminal controllers, and terminals.

· Physical Units (PUs) manage the communications hardware and software, participating in the controlling and routing of network communications. All physical devices are assigned a PU type (1, 2, 3, or 5) that identifies the level of the device (i.e., terminal, controller, communications processor, or host node) and its origin (i.e., IBM/SNA or non-IBM/SNA).

· Logical Units (LUs) manage communications software for communications with end users. A logical unit session is an end-to-end communication between an end-user terminal and the originating application residing in the host. LU 6.2, for example, supports peer-to-peer communications between intelligent devices, without requiring the host to assume responsibility for communications support activities. LU 6.2 is also known as Advanced Program-to-Program Communications (APPC).

The OSI model defines a set of common rules that computers of disparate origin can use to exchange information (communicate). As is the case with SNA and other such proprietary architectures, the model is layered to segment software responsibilities, with supporting software embedded in each node to provide an interface between layers. Specific levels of service can be negotiated between nodes.

The transmitting device uses the top layer, at which point the data are placed into a packet, prepended by a header. The data and header, known collectively as a Protocol Data Unit (PDU), are handled by each successively lower layer as the data move across the network to the receiving node. At the receiving node, the data work way up the layered model; successively higher layers strip off the header information. The seven layers of the OSI Reference Model are as follows:

· Layer Six (Presentation Control): Performs functions related to the format and display of received data by terminals and printers. Functions herein include data formatting, code set (e.g., Baudot, ASCII, EBCDIC, and Unicode), code set conversion [e.g., HTML to ASCII], text compression and decompression (e.g., WinZip), and encryption (e.g., DES).

· Layer Five (Session Control): Formats the data for transfer between end nodes, provides session restart and recovery and general maintenance of the session from end to end. This layer is considered by some to be of little or no consequence and is often disregarded and considered to be a function of layer 6.

· Layer Four (Transport Control): Responsible for maintaining the end-to-end integrity and control of the session. Data are accepted from the Session Control layer and passed through to the Network Control layer. Long message blocks are divided into shorter message blocks for transmission, sequence numbers are added, checksums are calculated and appended, retransmissions are generated in the event of errored message blocks or timeouts, and security is added. Example protocols that can be used at this layer include Transmission Control Protocol (TCP), User Datagram Protocol (UDP), and the five classes of the OSI Transport Protocol (TP). These protocols ensure end-to-end integrity of the data in a session. The X.25 packet-switching protocol operates at layers 1–4.

· Layer Three (Network Control): Comprises software that addresses and sequences the PDUs and transports them to the ultimate destination, setting up the appropriate paths between the various nodes. At this layer, message routing, error detection, and control of internodal traffic are managed. The Internetwork Protocol (IP) operates at this layer.

· Layer Two (Data Link Control): Establishes the communications link between individual devices over a physical link or channel. At this level, framing, error control, flow control, data sequencing, timeout levels, and data formatting occur. HDLC, LAP-B, LAPD are at this level. Frame Relay is a layer 2 protocol. LANs operate at layers 2 and 1, and Network Interface Cards (NICs) cut across portions of these two layers.

· Layer One (Physical Control): Defines the electrical and mechanical aspects of the interface of the device to a physical transmission medium, such as twisted pair, coax, or fiber. Communications hardware and software drivers are found at this layer, as are electrical specifications such as EIA/TIA-232 (nee RS-232). T/E-carrier runs at layer 1. Synchronous Optical NETwork (SONET) and Synchronous Digital Hierarchy (SDH) run at layers 1 and 2.

6.5 SECURITY

The only secure computer is one that is turned off, locked in a safe, and buried 20 feet down in a secret location … and I'm not completely confident of that one either.

Beruce Schneier, security technologist and author

Physical security involves access control, that is, control over the individuals who have access to the facilities in which the systems reside. Methods of access restriction include security guards, locks and keys, electronic combination locks, and/or electronic card key systems that require additional input, such as a Personal Identification Number (PIN). The latter is preferable because the system can maintain a record of specific access. Physical security also entails some decidedly low-tech tools such as document shredders and burn bags, which jointly serve to make paper documents and electronic media unusable after they serve their purposes.

Authentication provides a means by which network managers can validate the identity of those attempting access to computing resources and the data they house. Authentication consists of password protection and intelligent tokens. Password protection can restrict individuals on a site, host, application, screen, and field level. Passwords should consist of an alphanumeric value of reasonably long length and should be changed periodically. A current trend points toward the use of dedicated password servers for password management. Intelligent tokens are one-time pass-words generated by hardware devices and verified by a secure server on the receive side of the communication. They often work on a cumbersome challenge–response basis.

Remote Authentication Dial-In User Service (RADIUS) is a highly popular public network authentication service. Developed by Livingston Enterprises and based on a model defined by the Internet Engineering Task Force (IETF), RADIUS comprises an authentication server installed on the user's host computer and client protocols. Remote users are authenticated through a series of encrypted communications between the remote client and the centralized server. RADIUS is an open approach that can be modified easily to work with any security system and virtually any communications device. Access via a RADIUS Remote Access Server (RAS) commonly not only requires a password but also the remote computer must be equipped with a smart card reader into which a smart card must be inserted. The smart card must match the password, user ID, and a PIN before access is granted.

· Kerberos, the best-known authorization software, makes use of private-key authentication. Developed by the Massachusetts Institute of Technology (MIT), Kerberos is available free, although commercial versions exist. IBM's Kryptoknight, for example, is a Kerberos variant. Kerberos was named for the three-headed dog, Cerberus, that guarded the gates of Hades in Greek mythology.

· Sesame (Secure European System for Applications in a Multivendor Environ-ment) was developed by the ECMA (European Computer Manufacturers Association). It is flexible, open, and intended for large, heterogeneous network computing environments. It also is highly complex and not effective for smaller applications.

· Access Manager uses an API for applications, employing scripting. Scripting involves a process of mimicking the log-on procedures of a program, providing basic levels of security for small networks.

Port security is essential to deny unauthorized remote access. Passive devices report on unauthorized access, usage anomalies, and so on. Active devices, which are preferable, act to deny access to unauthorized users and disable ports if user-definable parameters (e.g., number of access attempts) are exceeded.

Transmission security is critical to ensure that unauthorized entities are not permitted to intercept the information as it traverses the network. Transmission of data is especially insecure over analog links because analog transmission does not lend itself to encryption, although encryption can be performed in the Data Terminal Equipment (DTE). Wireless transmission is inherently insecure, although digital wireless systems generally support signal encryption. Transmission security is virtually ensured over coaxial cable and other shielded and screened copper media and, especially, over fiber-optic cable because these media cannot easily be tapped. In order to maximize security, however, it is necessary that the data be encrypted.

Firewalls comprise application software that can reside in a communication router, server, or some other device. That device physically and/or logically is a first point of access into a networked system. On an active basis, the device can block access to unauthorized entities, effectively acting as a security firewall. Firewalls can use one or more basic approaches to access control. A packet-filtering firewall examines all data packets, forwarding or dropping individual packets based on predefined rules that specify where a packet is permitted to go and in consideration of both the authenticated identification of the user and the originating address of the request. Proxy firewalls act as intermediaries for user access requests by setting up a second connection to the resource. That second connection can be established at the application layer (layer 7) by an application proxy or at the session (layer 5) or transport (layer 4) layer by a circuit relay firewall. A stateful inspection firewall examines packets, notes the port numbers that they use for each connection, and shuts down those ports once the connection is terminated. Firewalls are the subject of much continuing interest, especially as organizations seek to protect their data from the ravages of hackers and other less than honorable creatures that prowl the Internet [22, 23].

Chapter 7: Conventional Digital and Data Networks

Wonders of the Universe, The Werner Company, 1899

According to the minutes of a Western Union meeting (ca. 1876):

Bell expects that the public will use his instrument without the aid of trained operators. Any telegraph engineer will at once see the fallacy of this plan. The public simply cannot be trusted to handle technical communications equipment. Bell's instrument uses nothing but the voice, which cannot be captured in concrete form … we leave it to you to judge whether any sensible man would transact his affairs by such means of communications. In conclusion, the committee feels that is must advise against any investment whatever in Bell's scheme.

After all, the telephone was far too technical for the average person. Somehow, the average person was able to overcome that obstacle, and the telephone networks quickly overtook the telegraph networks in terms of size, traffic, revenue, and virtually every other measure. But the telephone networks were limited to voice, and the telegraph networks carried all the data traffic—at least for 100 years or so.

Bell's invention proved itself and his company thrived over the next few years. Western Union continued to lease telegraph lines and supplied its customers with various kinds of telegraph keys, printing telegraphs, and dial telegraphs, some of which could transmit 60 words a minute. Western Union believed that these instruments could never be replaced by a scientific oddity such as the telephone until the Gold and Stock Company, one of its subsidiaries, reported that several of its instruments had been superseded by telephones. Western Union quickly realized its mistake, challenged Bell's patents in court, and formed the American Speaking-Telephone Company. Western Union announced that it had "the only original telephone" and was ready to supply "superior telephones with all the latest improvements made by the original inventors" (Dolbear, Gray, and Edison). Those ridiculous statements lifted Bell's invention from the status of scientific toy to article of commerce as businesses began to take the telephone seriously [2]. Western Union's patent attorney finally convinced the company that it could not win the lawsuit and encouraged them to settle out of court. In the final treaty (1879) between the two companies, Western Union agreed to admit that the Bell patent was the original, to admit that his patents were valid, and to retire from the telephone business. The Bell Company agreed to buy the Western Union telephone system (56,000 telephones in 55 cities), to pay the Western Union a royalty of 20 percent on all telephone rentals associated with that system, and to keep out of the telegraph business. The agreement remained in force for 17 years [2].

· Interconnected: Interconnection is fundamental to the PSTN, which provides for flexible connection between voice terminals, virtually regardless of location. Local Exchange Carriers (LECs), IntereXchange Carriers (IXCs), and inter-national carriers all have been interconnected for many years. Many data net-working applications do not demand full interconnection at this level but do not suffer because of it, issues of security aside.

In the late 1950s, large organizations (initially in North America) desired the ability to move data over telephone lines. This was first accomplished with a variation of the IBM 729 tape drive, which interfaced with the analog PSTN through a matched pair of Bell datasets (1957) [3], or DataPhones (1959) [4], via acoustic couplers and telephone sets. These early modems operated over the analog PSTN at 300 bps and, later, 1200 bps, which was incredibly fast at the time. At that point, voice and data networks began to merge. Datasets quickly spread around the world, rented by the telcos and PTTs to end users until deregulation afforded users the option of acquiring and interconnecting such equipment. The original datasets connected to the PSTN through a Data Access Arrangement (DAA) device that served as a coupler, or protector, to protect the network from high signal levels, out-of-band frequencies, and aberrant voltages. This protection is incorporated into contemporary modems and other devices, which are standardized and regulated by the ITU-T on an inter-national basis and by the U.S. Federal Communications Commission (FCC) and other national and regional regulatory bodies.

The telcos began to digitize their networks in the 1960s, as digital technology became reliable and inexpensive enough to support telecommunications applications and as the requirement surfaced for increased bandwidth in the carrier networks. Digital transmission facilities, in the form of T-carrier (North America) and E-carrier (Europe), increased the traffic capacity of existing facilities. In the 1970s, analog Electronic Common Control (ECC) switches began to be replaced with fully digital switches. Data transmission at relatively high speeds and over fully digital networks became a reality, although a number of years passed before such capability became widely available.

Clearly, digital transmission offers significant advantages, especially for data transmission. Those advantages include increased bandwidth and bandwidth utilization, improved error performance and increased throughput, and enhanced management and control. This chapter focuses on conventional digital data networking options, which include dedicated leased lines, circuit switching, and packet switching. Specifically, those technologies and service offerings include Dataphone Digital Service (DDS), Switched 56, classic Virtual Private Networks (VPNs) on the PSTN, digital carrier systems (T-carrier and E-carrier), X.25 and packet switching, and Integrated Services Digital Network (ISDN). In later chapters I provide a detailed introduction into more recent data networking options, including Frame Relay (FR), Asynchronous Transfer Mode (ATM), and Transmission Control Protocol/ Internet Protocol (TCP/IP).

7.1 DATAPHONE DIGITAL SERVICE

DDS is intended for FDX synchronous communications provided over four-wire circuits between computing systems that communicate intensively (i.e., frequently and passing significant volumes of data). The DDS network provides network timing and synchronization through a master clock, which ensures that all clocks in all slaved network nodes operate at the same bit rate, or clock speed, and at the same clock phase. The Data Communications Equipment (DCE) extracts timing from the received signal. DCE is in the form of a Data Service Unit/Channel Service Unit (DSU/CSU) that operates at the full line rate or on a subrate basis (lower speed), as required. Note: While the DDS circuit operates in FDX, HDX and simplex applications are supported. In fact, a large percentage of the DDS applications were for HDX polling.

Transmission rates vary, within limits, according to the user organization's requirements. Bandwidth generally is available at line rates of 2400 bps, 4800 bps, 9600 bps, 19.2 kbps, 56 kbps, or 64 kbps and digital carrier rates of 1.544 Mbps (T1) and 2.048 Mbps (E-1). Note here that the DDS signals actually are carried inside T-carrier or E-carrier channels in the backbone carrier networks. (I discuss T-carrier and E-carrier later in this chapter.)

While the cost of DDS circuits varies according to specific carrier tariffs and pricing strategies, cost is sensitive to both the distance between the points of termination and the level of bandwidth. Such is the case with all dedicated leased-line services. A traditional rule of thumb is that DDS generally is cost effective in applications that require communications between two locations, for a total of one hour per day or more, at a rate of 56 kbps. That equation clearly is sensitive to local rates and the availability of alternative services, such as Frame Relay.

In such a multipoint network, the head end addresses each remote system con-nected to the circuit on the basis of a unique logical address and in a user-definable and variable polling sequence. The target device recognizes its address and responds across the network, while all other devices remain silent. At that point, the two computer systems conduct a dialog until such time as either the data transfer is complete or the head end truncates the communication in order to address other devices according to its programmed polling schedule. Automatic teller machines traditionally were connected to the central bank in such a manner, and many such networks remain in place. In a typical scenario, the central computer polls the individual automatic teller machines, downloads the user's request for a cash with-drawal, and matches the account number and PIN for authentication purposes. It then queries the centralized database to determine the assigned level of withdrawal privileges and the current account balance and authorizes or denies the cash withdrawal.

DDS provides excellent reliability, which generally is in the range of five nines, that is, 99.999 percent. You should note, however, that all dedicated services are susceptible to catastrophic failure from such causes as cable-seeking backhoes. Therefore, network redundancy must be considered in the form of either backup DDS circuits or some alternative network service such as modem dial-up, Switched 56, or ISDN.

7.2 SWITCHED 56

Switched 56 (kbps service) is the popular term for Digital Switched Access (DSA), even though 64-kbps service is available in some areas. Switched 56 is a circuit-switched digital service intended generally for the same applications as is DDS, although it is more cost effective for less intensive communications. Although the service is switched, rather than dedicated, most of the general characteristics and all of the components closely resemble those of DDS, with the exception that digital COs are involved in setting up the DSA connections. Data Terminal Equipment (DTE) is in the form of computer systems, which connect to digital local loops through DCE in the form of a DSU/CSU. Digital exchanges serve to switch the connections (see Figure 7.2), which are provided through digital carrier transmission facilities on the basis of special routing logic.

While DDS is more cost effective for applications in which communications are intensive between specific physical locations, Switched 56 service is more cost effective for communications between locations that communicate less frequently or communicate lesser volumes of data. As Switched 56 calls are switched through the highly redundant carrier networks, rather than relying on vulnerable dedicated circuits as with DDS, Switched 56 services traditionally were employed as a backup to DDS facilities. They are still used for this purpose in some situations, although alternatives such as ISDN are more common in a contemporary context.

7.3 VIRTUAL PRIVATE NETWORKS: IN THE CLASSIC SENSE

Virtual Private Networks (VPNs), also known as Software-Defined Networks (SDNs) and Software Defined Data Networks (SDDNs), grew out of the voice world, as did the majority of network technologies. In a purely data context, VPNs are incumbent IXC offerings (e.g., AT&T, Cable & Wireless, and Verizon) that operate much like a voice VPN or Switched 56, although the level of bandwidth provided can be much greater in support of intensive data communications, videoconferencing, or multimedia conferencing. Depending on the carrier, VPNs support the following bandwidth levels on a circuit-switched basis:

· 56/64kbps

· N × 64kbps

· 384kbps

· 768kbps

· 1.544 Mbps (T1) or 2.048 Mbps (E1)

· 44.736 Mbps (Ts) or 34.368 Mbps (E3)

7.4 DIGITAL CARRIER SYSTEMS AND NETWORKS

As far back as 1882 Mr. Frank Jacob, Technical Adviser of Messrs. Siemens Bros., designed a plan by which the principle of the Wheatstone Bridge is utilized…. The wires forming the metallic loop should also balance in conductivity, insulation, and capacity. When these conditions are attained, T may speak to T′, and TT to T′T′, without the faintest sound of overhearing being apparent.

William Henry Preece and Juius Maier, The Telephone, Whitaker & Co., 1891

Carrier systems are defined as systems that derive multiple logical channels from a single physical communications path, thereby supporting multiple communications. Initially developed for use within public carrier (i.e., LEC and IXC) networks, the systems provided increased traffic capacity between exchanges without requiring additional transmission facilities. As voice traffic grew dramatically in the post–World War II period, new generations of Central Office Exchanges (COEs) were developed to relieve the strain, and digital carrier systems were developed to relieve the strain on the transmission facilities connecting them [4]. Before exploring digital carrier, we now dedicate a little ink to the history of analog carrier, just to put things in context.

The original N-carrier system (1950) was a short-haul carrier system that employed four-wire twisted-pair cable to deliver groups of 12 frequency division multiplexed voice-grade channels for connecting CO exchanges. ON1-and ON2-carrier systems later provided 16 and then 24 channels, respectively. Technology developed further to provide supergroups of 60 channels and master groups of 600 channels [3, 6, 7].

Type J (1938) systems were intended for long-haul open-wire circuits. Type K (1937) systems were the first long-haul systems developed for cable, rather than open wire; nominal repeater spacing was 17 miles.

L-carrier (1941) was quite an improvement, employing coaxial cable and an analog transmission scheme. L5E (1978), the last L-carrier system, used 22 coaxial tubes, in pairs, to carry a total of 132,000 simultaneous voice-grade conversations. Although this was an impressive improvement over N-carrier, the inherent problems of analog transmission were still present. Additionally, the coax cables were expensive and bulky, and the analog Radio Frequency (RF) amplifiers were expensive and prone to failure [3, 6]. At that point, it seemed as though Bell Labs engineers were intent on using up the entire English alphabet on analog carrier. When they developed the first digital carrier, they skipped a few letters. (That's a joke.)

The United States Bell System turned on the first commercial digital carrier system in 1962 under the streets of Chicago, Illinois, where electrical noise from high-tension lines and automotive ignitions interfered with analog systems. The system was designated T1, with the T standing for Terrestrial to distinguish the land transmission from satellite transmission. Bell Laboratories also launched Telstar I, the first communications satellite, in 1962 [7].

The impetus for development of digital carrier and Time Division Multiplexing (TDM) was due to the following factors:

1. Terminal multiplexing equipment could be made compact at low cost and could take maximum advantage of advances in solid-state digital circuit technology through system software and firmware upgrades.

2. Highly reliable techniques could be employed to provision a rugged transmission circuit that did not require complex design procedures and elaborate adjustments. In other words, the system could truly be standardized and would not require tuning in either the design or implementation phases.

3. A variety of services could share the same circuit and without the requirement to base the design on the most sensitive service.

4. Error performance would be improved considerably and would not be sensitive to circuit length, as regenerative repeaters, rather than amplifiers, would be employed [3].

Before proceeding with the discussion of T-carrier, I have to pause for just a few words to reinforce the fact that digital communications goes back to telegraphy and to mention that time division multiplexing dates at least to 1874, although in a crude form. In that year, Jean Maurice Emile Baudot of the French telegraph administration devised an automatic telegraph that involved synchronized rotating distributor arms that switched the use of the telegraph line between four or six sets of equipment at each end of the line. As the distributor arms revolved two or three times a second, they switched the line between the operators, each of whom had access to the line for a fraction of a second, which was just long enough to transmit a five-bit word in Baudot code. On the receiving end of the line was an electromechanical device that printed the messages in Roman type on a paper tape [1]. With that bit of history to put things in context, we can now fast-forward to 1962, and the next major development in digital transmission technology—T-carrier.

T-carrier is a dedicated, digital, leased-line service offering that employs TDM in order to derive multiple channels from a single four-wire circuit operating in FDX transmission mode. In capsule, T-carrier offers the advantages of digital error performance, increased bandwidth, and improved bandwidth utilization. As is the case with digital services, in general, T-carrier also delivers increased management and control capabilities to the carriers and end users alike. Additionally, T-carrier is medium independent. In other words, it can be provisioned over any of the transmission media (i.e., twisted pair, coax, microwave, satellite, free space optics, or fiber-optic cable), at least at transmission rates of T1 (1.544 Mbps) and below. At the higher rate of T3 (44.736 Mbps), twisted pair is not a suitable transmission medium, except over very short distances, due to issues of signal attenuation.

The significance of T-carrier extends well beyond its practical advantages. Specifically, and as the first digital carrier system, it set the standards for digital transmission and switching, including the use of Pulse Code Modulation (PCM) for digitizing analog voice signals. T-carrier not only set the basis for the North American digital hierarchy but also led to the development of similar standards, such as E-carrier in Europe and J-carrier in Japan. Ultimately, the CCITT (now ITU-T) developed international standards recommendations to ensure interconnectivity of national networks. Although T-carrier, E-carrier, and J-carrier are very different in terms of certain specifics of the protocols employed (e.g., transmission rates, encoding techniques, and signaling and control methods), their basic characteristics are much the same.

The fundamental building block of T-carrier is a 64-kbps channel, referred to as DS-0 (Digital Signal level Zero). Digital carrier is a channelized service, at least in a standard voice implementation. In other words, a single high-capacity digital circuit supports multiple logical channels, with each channel supporting a separate conversation. A T1 circuit, for example, operates at 1.544 Mbps, supporting the standard 24 time division multiplexed information-bearing channels, each with a bit rate of 64 kbps (Figure 7.4). E-1 supports 30 TDM channels of 64 kbps plus 2 separate signaling and control channels; J-1 supports 24 channels, as does T1.

The American National Standards Institute (ANSI) set the T-carrier hierarchy standards (see Table 7.1) in its T1.107 specifications. Beginning at the T1 level, the hierarchy progresses up to T4, which provides bandwidth of approximately 274 Mbps in support of 4032 channels. Most end users subscribe to T1 services, because one or more T1s generally satisfy their bandwidth requirements.

The process of transmitting data (voice, data, video, or image) in designated and consistently repeated channels, or time slots, is known as byte interleaving. For example, each voice conversation to be transmitted is accepted by the multiplexer, assuming capacity is available, and is assigned a time slot. The eight-bit bytes associated with that voice conversation are sent in the designated time slots 8000 times per second, for a total of 64 kbps, which is a voice-grade channel. Those time slots are reserved for that conversation, with the multiplexer providing the transmitting device with regular and repeated access to them for the duration of the communication. Time slots are reserved in both directions, because both real-time voice and T-carrier are FDX in nature.

While T1 is a digital service, it also supports the transmission of data such as voice and video that is analog in its native form. Codecs must be used to convert such analog signals to digital format prior to their being transmitted over a T1 circuit. The standard digitizing technique for voice, known as Pulse Code Modulation (PCM), was developed as an integral part of T-carrier. It also became the standard technique for digitizing voice in Private Branch eXchanges (PBXs) and other devices, for the obvious reason of providing seamless transmission between such devices and the network. The quantizing techniques typically employed in contemporary T1 networks include PCM and Adaptive Differental PCM (ADPCM); there also are a number of nonstandard approaches, including CVSD, VQL, VQC, and HCV.

It should be noted that sampling that is too infrequent results in a phenomenon known as aliasing, in which the digitized points can be used to represent more than one analog signal. As a result, the reconstructed analog voice signal is less than smooth and accurate, or even totally unintelligible. Similarly, sample encoding that yields bit values that are too approximate yields low-fidelity voice. Even PCM yields some amount of quantizing noise, which results from the inexact representation of a smoothly varying analog signal by a digital value restricted to 256 discrete steps. PCM also intentionally introduces some amount of quantizing noise, as it places the greatest emphasis on the amplitude levels in the low and normal volume range, where most speech activity takes place. PCM is nonlinear, much like a sliding-scale or graduated-scale voice ruler with the gradations spaced much closer together near zero volume and farther and farther apart as volume levels increase. This nonlinear approach sacrifices some voice quality at the higher amplitude levels, but the effect is masked by other distortions created by the telephone microphone (transmitter), speaker (receiver), and circuit when the volume is at such high levels. So, PCM can be considered to be a form of compression [3, 11]. (Note: The frequency of signal change, or tonal quality of the voice signal, is automatically taken into account.)

There are two different PCM companding techniques specified in G.711. In North America and areas under North American influence, μ-law, often printed as mu-law, is used. (ASCII does not conform to an expanded character set that includes Classic Greek letters, so you will please understand if the μ does not typeset correctly.) A slightly different nonlinear encoding technique known as A-law is used elsewhere in the world. The two techniques are similar, but different enough to require a gateway to convert from one to the other.

This standard approach of channelized T1, as noted previously, was developed and optimized for voice communications using PCM and TDM. Notably, T-carrier was developed for use in the carrier networks, and subsequently the carriers made it available to end-users. As a carrier-provided service, T1 must conform to all of the expectations of the carrier network, unless special arrangements are made. This is true of leased-line T-carrier networks, as depicted in Figure 7.3. It is particularly true of T-carrier when applied as an access service for purposes of access to, rather than through, the carrier network. For example, an end-user organization that employs a T1 local loop from a PBX to a circuit-switched PSTN must conform to the requirements of the carrier, which typically specifies 24 channels of 64 kbps and PCM-encoded voice. But large end-user organizations with large carrier-provided VPNs may take advantage of more efficient encoding techniques such as ADPCM, assuming that the carrier supports them. The following discussion explains how ADPCM supports toll-quality voice at 32 kbps, thereby yielding more efficient use of available bandwidth.

Differential Pulse Code Modulation (DPCM) makes more efficient use of bandwidth than PCM, as only the changes in signal level are encoded and transmitted. Based on the logical assumption that the change, or differential, in the voice signal occurs relatively gradually, fewer bits can be used to represent each sample. While DPCM will work with various numbers of bits, a four-bit approach generally is used in this technique, which yields a 2: 1 compression ratio. This level of compression enables a T1 circuit to support 48 channels of 32 kbps, rather than the PCM standard of 24 channels of 64 kbps. DPCM generally provides voice quality comparable to that of PCM. However, noise (distortion) may result on occasions when the signal varies significantly from one sample to another. A common example is that of a modem transmission, as the amplitude (i.e., volume) and frequency (i.e., pitch, or tone) levels can vary abruptly. This issue can be resolved through several workarounds. One approach is to run PCM on some channels of an intelligent TDM PBX or multiplexer and DPCM on others. If the PBX or multiplexer is intelligent enough, it will sense the modem transmissions and route those calls through the PCM channels. Another approach is to incorporate modems into the voice interface modules on the multiplexer. The modems will recognize the modem signals and demodulate the analog signal into a digital signal in PCM format transparently [11]. DPCM is unusual, as this set of issues and workarounds is generally troublesome.

The following calculation shows that 8000 four-bit samples per second yields a bandwidth requirement of 32 kbps for a DPCM-encoded digital voice signal:

Notably, ADPCM overcomes the deficiency of DPCM with respect to support of modem transmissions over T1. As you undoubtedly notice when dialing into the Internet, modem tones are very different from voice tones in that both the amplitude and frequency shifts are extreme. As noted previously, DPCM cannot accommodate these shifts. ADPCM, however, can distinguish the presence of a modem tone and can adapt by reverting to a channel width of 64 kbps or by forcing the modem to adapt to a lower speed. Although 32 kbps is the most commonly used ADPCM bit rate, ITU-T standards specify bit rates of 64, 56, 48, 40, 32, 24, and 16 kbps.

Digital Speech Interpolation (DSI) is rooted in a voice compression algorithm known as Time-Assigned Speech Interpolation (TASI) developed by Bell Telephone Laboratories in the 1950s for transatlantic telephone cable systems [13]. DSI makes the legitimate assumption that there are predictable pauses in normal human speech. During those pauses, additional voice signals are inserted through a technique known as silence suppression. As DSI works on the basis of statistical probabilities, it is employed effectively only when there are a significant number of voice conversations supported. For example, 72 channels yield additional compression of 1.5: 1, and 96 channels yield an additional 2: 1 (32 kbps). Newer implementations can provide as much as 4: 1 (16 kbps), although voice quality is compromised at this level.

DSI suffers the disadvantage of degradation of the signal quality during periods of heavy use. If the parties in a conversation speak rapidly, with few pauses, the voice signal can be clipped, or truncated, as the system attempts to detect valid speech and struggles to allocate bandwidth. The more conversations supported, however, the more predictable the average pauses and the lower the statistical probability of such degradation. DSI is commonly used in satellite communications, where bandwidth is precious.

Variations in the quantizing method are some times employed, although they are neither generally accepted nor widely deployed. Those variations include the following [11]:

· Variable Quantizing Level (VQL): compression ratio 2:1 (32kbps)

· Continuously Variable Slope Delta (CVSD): compression ratio 4:1 (16kbps), or 8:1 (9.6kbps)

· High Capacity Voice (HCV): compression ratio 8:1 (8 kbps)

T-carrier employs a very specific set of conventions to transmit information. Framing is one example. Using Tl as an illustration, each channel of input is time division multiplexed into a Tl frame, or set of data. In other words, conversation 1 might be allocated time slot 1 (channel 1), conversation 2 might be allocated time slot 2 (channel 2), and so on, through conversation 24 and channel 24. That set of sampled data is inserted into a frame, which is prepended by a framing bit, as illustrated in Figure 7.4, to distinguish it from subsequent frames of data. The process is repeated for frame 2, frame 3, and so on.

The combined processes of voice encoding and framing yield total Tl bandwidth of 1.544 Mbps. Of that total, 1.536 Mbps is available for information transfer, as noted in the following calculation; the remaining 8 kbps is required for framing and other transmission overhead:

There exist several generations of framing conventions, which are designated as Dl, D2, D3, D4, and Extended SuperFrame (ESF). Additionally, the ITU-T has developed an international set of recommendations for framing digital carrier signals:

· Dl framing, developed in 1962, robbed the Least Significant Bit (LSB)—the eighth bit—in each channel of each frame in order to insert a signaling bit in the form of alternating 1s and Os. Although Tl and PCM are designed around the interleaving of eight-bit bytes, bit robbing does not affect the quality of digitized voice because seven bits are satisfactory for reconstructing a high-quality approximation of the analog voice input. By truncating an eight-bit value, however, data are seriously impacted; the integrity of the data stream is violated as well. (Imagine the impact of a bit change that converts a decimal point to a comma during the transmission of a financial transaction.) In order to avoid this impact, the data always avoid the eighth bit in every channel, thereby limiting data transmission to 56kbps. While Dl framing no longer is used, the LSB still is robbed, even in the contemporary D4 framing technique. Therefore, data transmission remains constrained to 56 kbps in many carrier networks.

· D2 framing, was used to create a superframe, a 12-bit pattern in the F (Framing) bit position, that is, for framing locators. Information was transmitted in a 12-frame sequence or superframe. D2 framing is considered obsolete.

· D3 framing, which is still in use, assumes that all inputs—whether voice or data—are analog. It uses a superframe format and sequence bits.

· Extended superframe was originally tariffed by AT&T in 1985 and now heavily used. Extended superframes are 24 frames in length; signaling is performed in frames 6, 12, 18, and 24. It offers the advantages of nondisruptive error detection [six-bit Cyclic Redundancy Check (CRC)] and an Embedded Operations Channel (EOC) for network management using only 8 kbps of overhead [3, 6, 9]. This is accomplished because the highly intelligent ESF channel banks require only 2 kbps for purposes of synchronization. Thereby, the remaining 6 kbps of the framing bits are liberated for other purposed. Specifically, 2kbps is used for error detection (CRC) and 4 kbps is used for end-to-end diagnostics, network management, and maintenance functions.

· ITU-T international framing conventions differ greatly from those described above, which are used in North America and Japan (modified). ITU conventions call for level 1 (E-1) to employ 32 DS-0 channels, 30 for information and 2 specifically designated for signaling and control. The first such DS-0 channel carries the functional equivalent of framing bits, while the sixteenth carries signaling bits [11], as illustrated in Figure 7.7.

The repeaters are line powered, that is, powered from the CO exchange over the same transmission line that they serve at levels up to 100 V. The repeaters maintain their synchronization through the transmission bit stream; therefore, bipolar transmission is critical, as is the ones density rule.

Bipolar transmission refers to the fact that the electrical signal has two active states, positive (+) and negative (-) voltage, plus an inactive state of zero voltage. More specifically, Alternate Mark Inversion (AMI) reverses the polarity of alternate marks, or 1 bits, expressing the first as a positive voltage of +3 V, the second as a negative voltage of-3 V, the third as +3 V, and so on. Between any two successive marks, the signal returns to and pauses at zero voltage in order to maintain a zero reference point for the receiving multiplexer, as illustrated in Figure 7.8 [11].

Although T1 and E1 were developed specifically for voice communications, they are used intensively in support of data, video, and all variety of applications. Some of those applications, and video in particular, can involve long strings of zeros as legitimate data. In order to accommodate this requirement, extended superframe (ESF) for T-carrier and the ITU-T framing conventions for E-carrier both provide clear channel communications without bit robbing. They also support long strings of zeros through a technique known as Bipolar with Eight-Zeros Substitution (B8ZS). When B8ZS encounters a string of eight zeros (00000000), it substitutes a specific bit pattern that intentionally violates AMI, that is, includes an intentional BiPolar Violation (BPV). If the preceding mark (1 bit) was represented as a +3 V, the substituted bit pattern is 000+−0−+, as illustrated in Figure 7.9. If the preceding mark was represented as a-3 V, the substituted bit pattern is 000−+0+−. Since the bit pattern is known to both the transmitting and receiving multiplexer, the receiving multiplexer can restore the original 00000000 bit pattern.

Channel banks were among the first DS-1 devices. Designed for voice-only service in analog applications, channel banks interface analog switches (PBXs and COs) to DS-1 circuits. Channel banks perform two functions in sequence. First, they multiplex up to 24 analog signals on a common Pulse Amplitude Modulation (PAM) electrical bus. Second, they encode the individual PAM channels into a digital format, using PCM, for transmission over a DS-1 circuit [6, 9].

Channel banks also accommodate digital data. As relatively unintelligent devices, channel banks place each conversation on a separate channel; for example, a 9.6-kbps data conversation occupies a 64-kbps channel, just as does a 56-kbps data transmission or a digitized voice conversation. Therefore, channel banks do not make efficient use of available bandwidth. Combined channel banks and CSUs often are in the form of printed circuit boards that fit into PBX card slots for seam-less interface to a network T1 circuit.

Recently, a new breed of multiplexer has emerged in the form of Integrated Access Devices (IADs). These devices, which can be quite small and very low in cost, support multiple interfaces, perhaps to a small PBX for voice and a router for Frame Relay. Substantial economies can be realized through an IAD, which enables a single T1 or E1 circuit to be shared by multiple data types in support of multiple applications and services.

Digital Access Cross-Connect Systems (DACSs, or DCCSs) are nonblocking, ECC switches that serve to cross-connect digital carrier bit streams on a buffered basis by redirecting individual channels or frames from one circuit to another. They provide an electronic common control means of cross-connection that replaces the traditional manual method of physical cross-connection of wires. A DACS can redirect traffic to better manage the capacity and performance of the T-carrier network [13]. A DACS does not switch traffic call by call in the sense that you normally think of a switch. Rather, it switches data circuit to circuit or frame by frame on a preprogrammed basis. Although originally developed for carrier use, DACSs also are deployed in large user organizations to support private digital carrier networks. Smaller versions, residing on a PC, are available for less communications-intensive environments. Typically of significant port capacity, DACSs provide support for DS-0, DS-1, and DS-3 [6] and can accomplish conversions between T-carrier and E-carrier.

While the United States set the theme for digital carrier, the concept was quickly adopted by the CEPT (Committee on European Post and Telegraph) and the CCITT and is now specified in ITU-T G.703 and G. 704. The resulting E-carrier standard differs greatly in its implementation, and there exist some further differences in various national implementations. The Japanese version, J-carrier, resembles T-carrier, but with differences sufficient to cause incompatibility. The DS-0 channel is universal, and some number of DS-0 channels are organized into frames that are repeated 8000 times a second at a precise rate of 125 μs.

J-carrier closely resembles T-carrier, although the hierarchy is slightly different. Line coding and framing also vary considerably from the ANSI approach developed in the United States. Although diversity can be interesting, the advantages of these differences are questionable as, not surprisingly, incompatibility is ensured. The J-carrier digital hierarchy begins at 1.544 Mbps and proceeds to 6.313Mbps, 32.064Mbps (J1), 97.728Mbps (J3), and 397.20Mbps (J4) [6, 8]. Table 7.3 provides a clear comparison of the various DS levels.

While the data rate increases at the TIC and T2 levels, the carrier reference frequency increases, and issues of signal attenuation and crosstalk over copper twisted pair increase. This fact creates special engineering problems, which can be resolved by various means, including spacing repeaters ever more closely together. At the higher levels of T3 and T4 and the higher associated signaling speeds, these issues cannot be resolved satisfactorily. Therefore, these higher levels run over fiber optics, Free Space Optics (FSO), or microwave.

Above Tl, additional multiplexing is required. In the case of T2, the mux is termed an Mil (Multiplex Tl to T2). As mentioned in the above discussion of DLC and SLC-96, a T2 runs at 6.312Mbps and comprises four Tls at 1.544Mbps apiece, plus 132 kbps of overhead and justification, or bit stuffing, to adjust for variations in the clocking rates of the incoming Tls. The following set of simple equations builds the logic for the signaling speed from a single DS-0 channel to a Tl and then a T2:

At this level, overhead includes 555kbps for T3, 882kbps for T2 (136kbps × 7), plus 32 kbps for Tl (8 kbps × 4), for a grand total of 1.469 Mbps, or approximately 3.28 percent. That yields a payload factor of 96.72 percent, assuming clear channel communications. If bit robbing is involved in old North American networks, another 5.376Mbps drops away (8kbps × 672 channels), for a grand total of 6.845Mbps, leaving a theoretical payload of only 37.891 Mbps for an efficiency factor of 84.7 percent. If the payload contains data blocks, frames, packets, or cells with their own overhead issues, it gets even more overhead intensive. Now all of this overhead may seem wasteful, and it is at some level. At another level, however, we must realize that each bit of overhead has a purpose somewhere in some piece of equipment developed at some point in time to be backward compatible with something else that had some legitimate reason for doing what it did some years before, perhaps in a way that we would do entirely differently if we knew then what we know now. So, we deal with it, things interconnect to form networks that interconnect, and life is good. With that bit of wisdom having been painfully extracted from the dark recesses of my mind at this late hour and firmly implanted in this work, let us proceed.

The applications for digital carrier are many. Large user organizations find digital carrier services to be highly cost effective for local loop access, typically replacing multiple, single-channel PBX trunks. Large corporations find T/E-carrier effective for private, leased-line networks or access to VPNs. The ability of T/E-carrier to accommodate voice, facsimile, data, video, and image information on an unbiased basis and, therefore, to eliminate or reduce the number and variety of specialized circuits offers great advantage.

As a result of that study, the Advanced Research Project Agency NETwork (ARPANET) was developed. ARPANET, the first sophisticated packet-switched network architecture, was born in 1971. ARPANET was intended to link computers on a time-share basis in order to share computer resources more cost effectively [15]. Specifically, ARPANET was designed to support various defense, higher education, and Research and Development (R&D) organizations [16].

Packet switching soon was commercialized and made widely available in North America by companies including Telenet, Tymnet, and Graphnet (a facsimile-like service) in the United States and Datapac in Canada. In Europe, packet switching was offered early on by the pan-European Euronet. In fact, packet switching quickly became available in most countries and currently is virtually ubiquitous. The CCITT (now ITU-T) internationally standardized X.25 as the interface into a packet-switched network in 1976 and subsequently revised the standard in 1980, 1984, 1988, and 1993.

In 1983, the majority of ARPANET users spun off to form the Defense Data Network (DDN), also called MILNET (MILitary NETwork), which included European and Pacific Rim continents. Locations in the United States and Europe that remained with ARPANET then merged with the Defense Advanced Research Project Agency Network to become Defense Advanced Research Projects Agency (DARPA) Internet [17].

The wide availability of packet switching has made it consistently popular over the last 20 years or so. Additionally, packet networks are highly cost effective for applications that require many-to-many connectivity and involve relatively low data volumes. That popularity is growing and is ensured well into the future, largely through its historical deployment as the network technology of the Internet. You should note that X.25 is an interface specification, and does not define the internal operational characteristics of the data network.

The basic concept of packet switching is one of a highly flexible, shared network in support of interactive computer communications across a public WAN. Previously, large numbers of users spread across a wide area and, with only occasional communications requirements, had no cost-effective means of sharing access to host computer resources, that is, time share, from their remote asynchronous terminals. The issues were several:

· Analog Technology: Additionally, those early networks consisted of analog facilities in the form of twisted-pair local loops, with various combinations of twisted pair, coax, and microwave in the core. Analog technology offered very poor error performance and relatively little bandwidth.

· Cost: Calling costs were quite high across the WAN. This was especially true of data calls in support of asynchronous communications, as connect times are long even though data volumes tend to be low.

Existing circuit-switched networks certainly offered the required flexibility, as users could dial up the various host computers on which the desired database resided. Through a low-speed modem, which was quite expensive at the time, data could be passed over the analog network, although error performance was less than desirable. But the cost of the connection was significant because the calls were billed based on the entire duration of the connection, even though the circuit remained idle much of the time. Dedicated circuits could address the imbalance between cost and usage, because costs are not usage sensitive and dedicated circuits can be shared among multiple users through a concentrator. But dedicated circuits were expensive, especially in long-haul applications, and involved long implementation delays. Further, users tended not to be concentrated in locations where they could make effective use of dedicated circuits on a shared basis. Finally, large numbers of dedicated circuits were required to establish connectivity between clusters of users and the various hosts.

Packet switching solved many of those problems in the context of the limitations of the networks existing at the time. Packet-switched networks do a very cost-effective job of supporting low-speed, asynchronous, conversational, and bursty communications between computer systems in full-duplex mode. The low-speed, bursty nature of interactive asynchronous terminal-to-host applications enables large numbers of users to engage in simultaneous data sessions across a highly shared network. Rather than establishing connections across physical circuits to transmit data, the various network devices organize user data into packets and interleave them with packets generated by other users in a manner much like TDM muxes interleave bytes, although on a much less rigid basis. As packet-switched network usage can be billed to the user on the basis of the number of packets transmitted during a session, rather than billing for the duration of a call, packet networks are very cost effective for low-volume, interactive data communications. Further, packet-switched networks can perform the process of error detection and correction at each packet switch, or node, thereby considerably improving the integrity of the data from end to end.

Understanding the concept and nature of packet switching requires the examination of a number of dimensions and characteristics of such networks. The following is an exploration of the X.25 protocol suite, including the packet layer, access procedure, and frame format. There is considerable detail on switching and transmission, error control, connectionless service, latency, permanent virtual circuits versus switched virtual circuits, and protocol conversion.

· Packet Layer Protocol (PLP) maps into layer 3, the network layer.

· Link Access Procedure—Balanced (LAP-B) is the X.25 bit-oriented protocol for encapsulating the PLP packet. LAP-B maps into layer 2, the link layer.

· X.21bis defines the mechanical and electrical parameters for the physical, which is at layer 1, the physical layer. Layer 1 options include EIA/TIA-232, EIA/ TIA-449, EIA-530, and G.703.

The X.25 network layer protocol is the Packet Layer Protocol (PLP), which manages packet exchanges between physical DTE across a network of virtual circuits. PLP also can run on LANs and ISDN interfaces running Link Access Procedure—Data channel (LAP-D). PLP is responsible for call setup, synchronization, data transfer, and call clearing (i.e., call teardown). In data transfer mode, PLP transfers data between DTE across both Permanent Virtual Circuits (PVCs) and Switched Virtual Circuits (SVCs). Data transfer mode is responsible for data segmentation on the transmit side and reassembly on the receive side of the communication. This mode also handles bit padding, flow control, and error control.

An X.25 packet network transports and switches data through the network on the basis of packets, each of which is of a finite maximum size and of a specific structure, including a header and a payload. As illustrated in Figure 7.15, PLP packet fields include the following:

· General Format Identifier (GFI) is a 4-bit field that identifies packet parameters, which can include payload type (e.g., user data or control data), windowing information, and whether or not delivery confirmation is required.

· Logical Channel Identifier (LCI) is a 12-bit field that identifies the logical channel group and channel number of the virtual circuit that connects to the destination DTE.

· User or Control Data comprise the payload. If the packet is a data packet, the payload is encapsulated higher layer application information. If the packet is a control packet, the payload comprises various information relating to call setup and clearing, flow control and reset, and so on. The default maximum payload size is 128 octets, which every network must support. Public X.25 networks variously accommodate packets with maximum payloads of 16, 32, 64, 128, 256, 512, and 1024 octets. Airline reservation networks commonly use packet sizes of 1024 octets, although some custom networks can accommodate packet sizes of up to 4096 octets.

· Control: The control field is a one-octet field that identifies the frame type. An Information (I) frame carries upper layer (e.g., application) information and some control data. A Supervisory (S) frame carries control information such as I-frame acknowledgment, request for retransmission, and flow control. An Unnumbered (U) frame carries control data such as disconnection request, acknowledgment frame, and frame reject.

· Data: The data field is a variable-size field that contains upper layer information in the form of an encapsulated PLP packet.

· Frame Check Sequence (FCS): The FCS field is a two-octet CRC field that provides excellent error detection. The CRC provides meaningful information for error correction, which is performed at a higher level.

X.21bis is a Layer 1, or Physical Layer, specification used in X.25. X.21bis defines the mechanical and electrical parameters for cables and connectors in support of FDX transmission at speeds from 9600 bps and 64 kbps for point-to-point connections over four-wire circuits. As a pure Physical Layer specification, X.21bis addresses the movement of electrical bits across a wire that connects DTE and DCE, and that's about all. In addition to moving bits, of course, there are provisions for call control, which entails the devices using signaling leads to signal when they are ready or not ready to receive calls. There also is a provision for signal timing, or synchronization, at 8 kbps. Figure 7.17 illustrates the relationship between the Layer 3 PLP packet, the Layer 2 LAP-B frame, and the Layer 1 X.21bis bit stream. Note that the physical layer circuits connecting nodes in the network core can take any number of forms, from analog to digital, T-carrier to E-carrier, copper to fiber, and anything in between.

X.25 internodal links originally were dedicated analog trunks leased from the various local and long-haul carriers. Over time, those analog facilities were replaced by digital circuits, usually in the form of 56-kbps DDS circuits. Many of those circuits subsequently were replaced with T-carrier and E-carrier facilities. Currently, the facilities generally are high speed and fiber optic in nature, although a variety of analog and digital media are employed in consideration of specific network economics, all of which are sensitive to the region.

Datagram mode aside, X.25 packet switching is a connection-oriented service. That is to say that a call is set up over a shared physical path, or Virtual Circuit (VC), before the first packet is sent and over which all packets may travel in support of a logical connection. In datagram mode, X.25 is connectionless, with each packet perhaps traveling a different path, depending on the availability and performance of the various network links at any given moment in time. In either case, each packet of data is addressed separately and, therefore, is capable of working its way through the network independently of the other packets in a stream of packetized data. This characteristic of packet networks is a critical advantage because the network and all of its elements are shared among a large number of users. Hence, the cost of transmission across such a network is very low in the context of an appropriate application. There are two types of virtual circuits: PVCs and SVCs.

Packet switching supports a large number of transmissions riding over the same previously designated circuit or path. While the individual packets of the typical user may travel different paths, those of large user organizations that use the network intensively commonly travel over permanent virtual circuits (PVCs). In this scenario, all packets always travel the same path between two host computers, as illustrated in Figure 7.19. The path is established on the basis of routing instructions programmed in the routing tables of the involved nodes and is invoked based on the logical channel group and logical channel number contained in the LCI field of the PLP packet. As the path is predetermined and programmed in a routing table, it can be identified and exercised quite quickly.

Alternatively, the network nodes may select the most available and appropriate path on a call-by-call basis using switched virtual circuits (SVCs) that are set up on the command of (i.e., signaling from) the user equipment before the first packet is sent. Once selected, all packets in a given session travel the same path, just as they do in a PVC scenario. SVCs differ from PVCs as the path is set up in consideration of both the condition and the load at the instant the connection is required, rather than being defined well in advance. Therefore, failed and congested switches and circuits are bypassed, and overall performance is improved. This process of automatic load balancing offers benefits to both the user organization and the carrier. SVCs, however, demand a greater level of network intelligence, which adds to total network cost to the carrier and ultimately translates into higher cost to the end-user organization. The establishment of an SVC also involves some level of delay in the call setup process, since the network nodes must examine multiple paths prior to making an optimal path selection.

Latency, or delay, is a troublesome and limiting characteristic of packet networks. As each packet may take a different route though the network in datagram mode, each may travel a route of a different length; therefore, propagation delay may vary from packet to packet. Additionally, each packet may travel through a different number of packet nodes, each of which must act on the packet to read its address, check for errors, request retransmissions of errored packets, and so on. Each of these processes compounds the issue of packet delay. Further, each packet may encounter a different level of congestion in the network and, therefore, can spend a different amount of time in queues. Finally, delay is imposed on each packet if protocol conversion is required; while this process adds value, it also adds to the latency factor. The end result is that some level of latency not only is assured but also is variable (jitter) and uncertain in magnitude [8].

The level of magnitude is measured in milliseconds (ms, or thousandths of a second), which does not seem like much on the surface. But while this characteristic of packet switching does not affect many applications, it renders others ineffective. Many data communications applications, such as e-mail, are not seriously impacted by latency or jitter. Real-time uncompressed audio, voice, and video, however, are affected quite seriously and, therefore, traditionally have not been considered good candidate applications for packet switching.

X.25 actually is the ITU-T recommendation for a standard describing the physical, link, and packet-level protocols between the user DTE/DCE and the network [17, 19]. Host computers connect to the network over an X.25 link. The user data are packetized by DCE in the form of a Packet Assembler/Disassembler (PAD), for which X.3 is the standard. The PAD also may be responsible for password protection and performance reporting [10].

Large user organizations often access the network via a dedicated, leased-line link to the closest network node. Such access often is in the form of an unchannel-ized T1/E1 facility. In either case, the PAD organizes the user data into PLP packets and encapsulates each in an LAP-B frame before presentation to the network [20, 21].

X.25 networks are widely available as a public data network (PDN) service offering, generally using packets of 128 or 256 bytes. Certain applications, however, are supported more effectively by transmission of larger packets. The airline reservation systems (e.g., American Airlines' SABRE and United Airlines' APOLLO), for example, deploy custom packet networks that use packet payloads of 1024 bytes. As this application involves the frequent transmission of relatively large sets of data (e.g., flight schedules, fares, and seating availability), a larger packet size is more appropriate. The larger packet size improves efficiency because the payload is very large, while the overhead information is roughly the same as it is for a smaller packet. As a larger packet is more likely to contain an errored bit, require retransmission, and therefore reduce throughput, the custom reservation networks typically employ digital facilities to minimize this exposure. In the United States and other highly developed countries, the reservation networks largely have shifted to Frame Relay, although X.25 remains heavily used elsewhere, in consideration of poor link quality.

The user of an X.25 packet network may require no hardware other than a PC and modem, with the packetizing performed at the local X.25 node. Alternatively, a hardware-or software-based PAD conforming to the X.3 standard can perform that function at the host. (Note: The advantage of the hardware PAD is that if it misbehaves, you can hit it with a hammer.) The PAD performs the packet assembly (i.e., aggregation of many individual characters) for the transmitting device and disassembly (i.e., disaggregation) for the receiving device in order to reconstitute the data in its native format.

Packet carriers, of course, must invest in packet nodes rather than circuit switches. Such packet nodes are intelligent devices capable of supporting complex routing tables, buffering packets in temporary memory, resolving packet errors, and accomplishing protocol conversions. Where X.25 networks are interconnected or where protocol conversions of any sort are required, gateways must be deployed.

The ITU-T sets standards recommendations for packet switching. Those standards include the following [8, 10, 21]:

· X.3: Packet Assembly/Disassembly (PAD) functions

· X.25: Interface between DCE and DTE for public packet networks

Packet Layer Protocol (PLP): Network Layer

Link Access Procedure—Balanced (LAPB): Link Layer

X.21bis: Physical Layer

· X.28: Terminal-to-PAD communications formats

· X.29: Host-to-PAD communications formats

· X.31: Packet-mode services over ISDN

· X.32: Defines X.25 synchronous dial-up mode

· X.75: Internetwork call control procedures

X.25 packet switching originally was intended for interactive time sharing, which involves long connect times and low data volumes. While X.25 still supports such applications effectively, contemporary applications include online interactive processing (e.g., reservations systems), messaging (e.g., e-mail), batch file transfer (e.g., data backup), and Internet access.

There are several bright spots in the future of X.25. As it is the packet protocol used in SS7, it is heavily used in the contemporary PSTN for signaling and control purposes. Therefore, it also is used in the ISDN D channel, not only for user-to-network signaling but also for user-to-user signaling in private ISDN networks and for certain user-to-user data communications applications. However, the contemporary PSTN is in a state of decline due to the popularity of Voice over IP (VoIP) and the Internet. ISDN is not exactly on the upswing either. Actually, I reckon these are more like dim spots. I'm sorry. I didn't mean to get your hopes up.

7.6 INTEGRATED SERVICES DIGITAL NETWORK

A CCITT study group first explored integrated services digital network (ISDN) as a concept from 1968 to 1971. A more focused conceptual study took place during the 1981–1984 CCITT study period. The first set of published standards recommendations appeared in 1984 in the form of a CCITT Red Book, which provided the basic framework for the concept, network architecture, UNI (User Network Interface) protocols, and common channel signaling protocols. As a result of the 1985–1988 study period, a Blue Book was published that provided descriptions of supplementary services, rate adaptation, ISDN frame relay, and the initial set of B-ISDN (Broadband ISDN) recommendations. (Note: The color of the books has no significance, other than the fact that a different color represents each study period.)

Rather than being a technology, ISDN is described as a suite of services based on a set of technologies, including transmission, switching, and signaling and control. It is a set of international standards recommendations that permits the provisioning of a wide range of services intended to be available on a ubiquitous basis. Additionally, the ISDN network is accessible through a standard set of interfaces—one for low-bandwidth applications and another for high-bandwidth applications.

Announced to the world amidst great fanfare, ISDN quickly captured the interest of carriers, manufacturers, and user organizations worldwide. ISDN offered the compelling advantages of increased bandwidth, enhanced flexibility, improved error performance, greater reliability, broad availability, and interconnection to a wide range of services. Unfortunately, it then stalled and progressed at a glacial pace for the next few years. Among the many reasons for its slow development are long delays in standards development, lack of adherence to standards, lack of availabil-ity, regulatory hurdles, circuit and equipment costs, and poor marketing, particu-larly in the United States.

Standards development at the ITU-T is infamously slow. Standards traditionally were released every four years, in monsoon fashion and with total droughts in the interim. Over time, the various committees attained the privilege of developing and releasing certain standards recommendations on an intermediate basis. Standards then came in showers. ISDN standards include three layers. The ISDN Physical Layer (Layer 1) addresses mechanical and electrical issues including connectors, signaling rate, and line coding. At the Data Link Layer (Layer 2), ISDN specifies the Link Access Procedure—D Channel (LAP-D). Network Layer (Layer 3) specifications include user-to-user and network-to-network signaling protocols for both circuit-switched and packet-switched networking.

Standards from the ITU-T actually are in the form of recommendations. Individual member nations can implement ISDN options as they see fit or they can deviate from the standards as long as international interconnectivity is accomplished at some reasonable level. The most notable international difference is that of the basic ISDN hierarchy. The North American version follows the T1 hierarchy, with PRI including 24 channels; the European (ITU-T) version is based on E1, providing 30 information-bearing channels. While this difference is understandable in the context of maintaining backward compatibility with existing networks, it also perpetuates issues of basic protocol incompatibility.

Systems manufacturers of COs and PBXs have a strong interest in maintaining the proprietary nature of their systems architectures. Therefore, they implement ISDN in distinctly different ways. ISDN compatibility became ISDN compliance, a decidedly lower level of conformance.

Additionally, carriers have implemented nonstandard versions of ISDN. For example, Pacific Bell (now SBC) initially offered ISDN at a rate of 56 kbps per channel, rather than the standard 64 kbps. This limitation was due to the presence of older channel banks in the Pac Bell ISDN carrier network and the fact that SS7 was not fully deployed. Therefore, in-band signaling and control consumed 8 kbps of channel bandwidth due to bit robbing.

Availability of ISDN was slow to develop in the United States because the carriers were reluctant to invest in the technology unless they were convinced that a market existed for the services or that the technology offered internal cost savings. ISDN is not inexpensive to deploy, a fact that unfortunately has been reflected in relatively high installation charges, recurring circuit charges, and equipment costs.

Regulators in the United States generally have required that the LECs pass on the cost of ISDN infrastructure to ISDN users, rather than averaging those costs across the entire rate base. In other words, they have viewed ISDN as an optional service that must pay its own way, or the carriers must absorb any associated losses. Because the carriers were unwilling to do so, ISDN rates remained high. Coincident with the development of competition in the local exchange and encouraged by the regulators, the ILECs have exercised more freedom in pricing ISDN attractively. Each ILEC, of course, has its own pricing strategy involving installation charges, monthly rates, and usage charges. The CLECs (Competitive LECs) are free to price ISDN (and any other service offering) as they see fit and generally price it at very attractive levels.

Rates for ISDN access historically were not tariffed at attractive levels in the United States, compared with the cost of basic services. Again, the regulators were largely responsible. The ILECs bear responsibility, as well, because they were not willing to absorb initial losses in order to stimulate the growth of the service offering. Currently, the charges for ISDN vary widely, from carrier to carrier and from state to state. Generally speaking, ISDN charges are somewhat higher than those for two analog lines and for monthly recurring charges as well as installation charges. Some carriers also charge for usage per channel and sometimes charge for packet traffic based on packet volume. To encourage ISDN usage, some carriers waive or lower those usage charges during evening and weekends, when calling activity is light.

Equipment costs were high because the manufacturers constantly were investing in R&D to maintain ISDN compliance with developing standards. Additionally, the limited demand for ISDN caused the manufacturing runs to be small, which tends to increase equipment prices. While equipment costs subsequently came down considerably, they remain an additional expense.

ISDN frustrated the industry in the United States, in general, and its less than stunning level of success eroded gradually in the face of V.90 modems. Since 2000 or so, Digital Subscriber Line (DSL) and cable modems have virtually wiped ISDN off the map, at least in the residential and small-business markets. At least one ILEC in the United States has removed all mention of ISDN from its website, closed its ISDN support office, and left ISDN customers with only a telephone number connected to an answering machine on which they can leave messages that (some say) are never returned. (I'll not name that company, in order to save them the embarrassment. At the same time, it will save them the trouble of filing a civil lawsuit against me for damages that they can't prove but that I can't disprove. Note: The United States has approximately 1,000,000 lawyers, which is about 70 percent of the world's lawyers, and a total population of about 300,000,000, which is less than 5 percent of the world's population.)

ISDN matured much more quickly and completely in Europe and certain parts of the Pacific Rim and Africa than in the United States. In those regions, the regulators encouraged deployment. Additionally, marketing was much more effective, focusing on PRI, rather than BRI [18, 23]. Outside the United States, ISDN pricing also has positioned it favorably in comparison to analog lines and modem-based Internet access. In South Africa, for example, ISDN BRI is priced very attractively for low-to-moderate Internet access. In many regions of the world, ISDN also ben-efits from the lack of competition from DSL and cable modems.

ISDN specifications include a number of functional groupings at layer 1, the physical layer. ISDN hardware, at the end-user side of the connection, includes Terminal Equipment (TE), Terminal Adapters (TAs), and Network Termination (NT) devices, line termination equipment, and exchange termination equipment, as depicted in Figure 7.21. The following discussion begins at the subscriber premises and works toward the carrier networks.

Terminal Equipment (TE) is the term for a device that connects a customer site to ISDN services. In more traditional non-ISDN terms, TE includes all Customer Premises Equipment (CPE), which category includes voice terminal equipment and Data Terminal Equipment (DTE). TE also includes premises-based switching equipment, including PBXs and routers. There are two types of TE:

· TE1 is ISDN-compatible equipment, that is, equipment that can interface directly to an ISDN circuit via a four-wire twisted-pair interface. Examples 2, 3, and 4 in Figure 7.21 illustrate TE1.

· TE2 devices do not enjoy native ISDN compatibility. TE2 equipment must connect through a Terminal Adapter (TA) that resolves issues of incompatibility. Example 1 in Figure 7.21 illustrates TE2.

A key function of the TA is that of rate adaption, which operates in several ways. In one case, rate adaption effectively throttles down the transmission rate from 64 kbps to the rate at which the non-ISDN device is capable. Rate adaption also serves to bond multiple B channels into H channels for more bandwidth-intensive applications. V.110 is the ITU-T Recommendation that specifies support for data terminal equipment (DTE) with asynchronous or sunchronous serial interfaces over an ISDN network through rate adaption. V.120 specifies support for data terminal equipment (DTE) with asynchronous or sunchronous serial interfaces over an ISDN network through data encapsulation. V.120 includes specifications for allowing multiple terminals to share a 64-kbps B channel through Statistical Time Division Multiplexing (STDM).

Network Termination (NT) devices are physical devices that operate to interface the four-wire customer wiring to the two-wire UTP local loop. As illustrated in Figure 7.21, there are two types of NTs:

· NT2 operates at Layer 2, the Data Link Layer, and Layer 3, the Network Layer. NT2 is an intelligent device responsible for the user's side of the connection to the network, performing such functions as Layer 2 multiplexing, switching, or ISDN concentration. An NT2 commonly is actually an NT1/2 device, performing the combined functions of an NT1 and an NT2 and operating at Layers 1, 2, and 3 of the OSI Reference Model. Such a device likely would be in the form of a PBX, router, or data switch.

These functional groupings of TE, TAs, and NTs are logically distinguished by reference points, which are points of reference, interface, or demarcation. As illustrated in Figure 7.21, the reference points are as follows:

· Reference Point R corresponds to the interface between TE2 and the TA.

· Reference Point S refers to the point of interface of an ISDN terminal and the NT2, serving to distinguish between terminal equipment and network-related functions. The S interface is defined as a passive bus for up to eight NT2 devices.

· Reference Point T references a minimal point of termination at the customer premises. It is the reference point between the NT1 and NT2 devices.

· Reference Point U is the reference point between NT1 devices and line termination equipment in the carrier CO. The reference point describes the full duplex data signal on the physical two-wire subscriber line, including line coding and framing conventions. The U reference is relevant only in North America, as the NT1 function is provided by the carrier elsewhere.

· Reference Point V is the point of interface at the network side of the connection between the line termination or loop termination and the exchange termination. In other words, it is the point between the circuit-terminating equipment and the ISDN CO. As the V interface exists only if the CO does not have embedded circuit-terminating equipment, it is unusual in contemporary ISDN-compatible COs.

Basic Rate Interface (BRI), also known as Basic Rate Access (BRA) and 2B+D, provides two Bearer (B), or information-bearing, channels, each operating at the clear-channel rate of 64 kbps by virtue of SS7 non-intrusive signaling. Each B channel can carry digital data, digitized voice (PCM encoded at 64 kbps or a lower rate), or a mixture of low-speed (subrate) data as long as it all is intended for the same destination. BRI also provides a Data (D) channel at 16 kbps, which is intended primarily for purposes of signaling and control, messaging, and network management. The D channel also generally is made available for X.25 packet data transmission and low-speed telemetry when not in use for signaling purposes; cost-effective applications include credit card authorization, which involves very small bursts of data [24]. BRI is used primarily for residential, small-business, Centrex, and telecommuting applications that are not particularly bandwidth intensive. The B channels can be aggregated, or bonded, to provide up to 128 kbps to a given conversation, such as a videoconference or Internet experience; additionally, multiple BRIs can be bonded for even greater capacity. Whether bonded or not, ISDN BRI provides multiple channels over a single physical loop, which is a great advantage.

A single BRI line can support up to 16 devices that contend for access to the BRI channels through a Terminal Adapter (TA). The devices can be in a variety of forms, including telephones, facsimile machines, computers, and video cameras. Additionally, ISDN can support up to 64 individual Service Profile IDentifiers (SPIDs), which are equivalent to directory numbers, one per terminal device [25]. The SPID is used in the initialization procedure when the device goes off hook to establish a connection with the network through the CO. While BRI supports as many as three simultaneous calls, the specifics of the carrier offering determine how they can be used. Some carriers offer a confusing array of ISDN BRI packages that either restrict particular B channels to voice or data application or enable them both to be used for voice or data. Some carriers offer application-specific BRI variations such as 1B+D and 0B+D, with the latter targeted at applications such as credit card verification, as illustrated in example 3 of Figure 7.21.

All TDM-based services require framing, and ISDN BRI is no exception. The 2B1Q-coded transmission must be organized and synchronized in some fashion in order for the B and D channels to be interleaved without losing their identity. Signaling and control functions also must be performed. The ISDN BRI involves several framing conventions, with one at the U interface and the other between the S and T interfaces:

· U Interface: The framing at the U interface, as illustrated in Figure 7.24, begins with a synchronization word of 18 bits that provides the receiver with the means to synchronize on the beginning of each frame. That is followed by 12 sets of 2B+D samples, with each B sample comprising 8 bits and each D sample comprising 2 bits, for a total of 18 bits per set. The frame ends with 6 bits of overhead in the M channel for M aintenance and other purposes. Much as is the case with T1 and E1, sets of eight frames are organized into superframes. It is at that level that the M channels find their purpose in addressing, error control, power management, and various network management functions.

Primary Rate Interface (PRI) also is known as 23B+D in the United States and Japan. The European or international version is known as Primary Rate Access (PRA) or 30B+D. PRI offers 23 B (Bearer) channels plus 1 D (Data) channel and is backward compatible with T1 and J1 transmission systems, respectively. PRA offers 30 B channels plus 1 D channel and is backward compatible with E1 transmission. PRI and PRA both provide a full-duplex (FDX) point-to-point connection through an NT2-type intelligent CPE switching device, such as a PBX or router, for interfacing with the carrier CO switch. The DS-0 is the basic building block of both PRI and PRA, as both the B and D channels operate on clear channels at 64 kbps. As is the case with BRI, the B channels can be used individually or can be bonded for voice, data, video, facsimile, any other data and any multimedia combination, but the D channel is reserved exclusively for signaling. As signaling and control functions are fairly light, the standards provide for Non-Facility-Associated Signaling (NFAS), which allows a D channel to support up to five PRI connections. The first PRI in a PBX trunking application, for example, would be provided at 23B+D, and the next four PRIs would be delivered at 24B+0D. Some manufacturers and carriers have stretched this limit to seven PRIs. In a con-figuration involving multiple PRIs, the typical recommendation is that a backup D channel be provisioned in order to maintain signaling and control functionality in the event that the primary D channel fails. While this backup approach diminishes the B-channel count, it is generally considered to be the most prudent approach, for B channels are useless without a D channel. NFAS is required for Switched-1536 data service; because all 24 channels of the T1 PRI line carry user data, the D channel must be on another line.

While designed for transmission over a standard DS-1 trunk, PRI is a significant improvement over T1 or E1, because the channels can be allocated dynamically. In other words, each channel can act as an incoming, outgoing, combination, or DID trunk, as the need arises. The nature of the channel can be determined as required, based on user-definable parameters. Additionally, multiple B channels can be aggregated to serve bandwidth-intensive applications, such as videoconferencing. On the negative side, PRI does not compare favorably with T1 in terms of the raw number of B channels, and PRI can be considerably more expensive, depending on tariff specifics.

PRI requires the same line coding as contemporary T1 and E1. Specifically, the coding technique is Alternate Mark Inversion (AMI) with Bipolar with Eight-Zero Substitution (B8ZS).

H channels (High-speed channels) are functionally equivalent to B channels but provide greater aggregate bandwidth in PRI applications. In this N × 64 mode, any number of B channels can be aggregated, or bonded, on a dynamic basis to enable multirate communications through inverse multiplexing. The ISDN approach does have a drawback, however, when compared to traditional inverse muxes, as the connection must be torn down and reinitiated when channels are added or dropped. The feature is known variously as multirate ISDN, N × 64, channel aggregation, and bonding. H channels find application in fast faxing (Group IV), videoconferencing, high-speed data transfer, high-quality audio transmission, and Frame Relay. Defined H channels include the following:

· H0 channels have an aggregate bit rate of 384 kbps, which is the equivalent of six B channels (6 × 64 kbps = 384 kbps). This is a common port speed offered by Frame Relay service providers.

· H1 is a full DS-1, with no framing overhead. This channel is sensitive to the specifics of the DS-1 implementation.

· H10 operates at 1.472 Mbps, which is the sum of the 23 B channels (23 × 64 kbps = 1.472 Mbps) in a baseline PRI implementation in which channel 24 is devoted to the D channel. H 10 applies in North America and Japan and is based on T1 and J1, respectively.

· H11 operates at 1.536 Mbps, the sum of all 24 B channels for the North American and Japanese versions, which is based on T1. H11 relies on Non-Facility-Associated Signaling (NFAS) to provide a D channel an H10 facility for signaling and control.

· H12 operates at 1.920 Mbps, the sum of all 30 B channels for the European version, which is based on E1.

Offered by some manufacturers, inverse muxes enable multiple BRIs to bond, or link, for greater aggregate transmission over multiple BRIs. For example, four BRIs can be linked to support a 512-kbps data transmission. Such an approach competes effectively with Fractional T1 (FT1) where it is available.

D-channel contention devices, also known as ISDN routers and offered by some manufacturers, permit as many as eight devices to share a BRI circuit, contending for access to the B channel. The individual devices identify themselves to the network through contention for the D channel. The ISDN router also serves traditional switch functions within the context of the small-user domain.

LAP-D is a balanced protocol that operates in Asynchronous Balanced Mode (ABM), referring to the fact that the devices have a balanced, rather than a master/ slave, relationship. Therefore, a device at either end of the link can initiate a dialogue at any time. As illustrated in Figure 7.25, the LAP-B frame format includes the following fields:

· Flag: The flag field is a one-octet field that delimits (i.e., establishes the limits or boundaries of) the beginning and end of the frame. The flag is always the specific bit pattern 01111110 (7E in hexadecimal), which is known to all transmitters and receivers.

· Address: The address field is a two-octet field known as the Data Link Connection Identifier (DLCI), which is divided into two addresses. The first octet is the Service Access Point Identifier (SAPI), which identifies the destination service access point, each of which can support multiple terminal devices. The second octet is the Terminal Endpoint Identifier (TEI), which is the address of the destination terminal device.

· Control: The address field is a one-or two-octet field that identifies the frame type. An Information (I) frame carries upper layer (e.g., application) information and some control data. A Supervisory (S) frame carries control information such as I-frame acknowledgment, request for retransmission, and flow control. An Unnumbered (U) frame carries control data such as disconnection request, acknowledgment frame, and frame reject.

· Frame Check Sequence: The FCS field is a two-octet field that provides excellent error detection.

ISDN is unusual, if not unique, in that it is undoubtedly the most carefully planned, well-coordinated, and best-documented network technology in history. Despite this fact, or perhaps because of it, ISDN's popularity has lagged due to the previously mentioned issues of cost, availability, and applications. The key characteristics of ISDN include its end-to-end digital nature, which is unusual for a public circuit-switched network. Through a small family of interfaces, a wide range of services can be accessed through the LEC, IXC, or CAP. Rate adaption and channel aggregation permit bandwidth-intensive applications to be supported on a dynamic basis.

ISDN does have some drawbacks. These include limited availability, standards variations, and cost. As discussed earlier in this chapter, availability has always been limited in most nations because ISDN and SS7 software is costly for the carriers to deploy; while ISDN capability can be extended to non-ISDN CO exchanges, that incremental cost is not trivial.

Not surprisingly, cost–benefit considerations dictate the success or failure of technologies, regardless of how compelling they appear at first glance. The cost of an ISDN BRI circuit often is more than twice that of an analog line, and installation costs can be considerably higher. These costs tend to discourage ISDN to some extent, particularly in voice-intensive environments where an ISDN BRI configuration may limit voice to a single channel.

Additionally, many carrier tariffs impose a usage surcharge in the form of a flat rate per minute for circuit-switched connections and a packet surcharge for packet data. The usage charge applies to local as well as long-distance calls. While these additional usage charges tend to discourage ISDN usage, the faster speed of data transfer may serve to reduce call connect time significantly, at least in comparison to dialup modem connections over analog lines.

Notably, ISDN hardware depends on local power. In other words, the phones do not work when the power goes out. Large end-user organizations are accustomed to providing power backup in the form of an Uninterruptible Power Supply (UPS) for voice [e.g., PBX, ACD, and KTS) and data (e.g., workstations, servers, hubs, access points, switches, and routers) systems. Residential and small-business users often do not employ UPS systems. Therefore, ISDN BRI is not recommended as a full replacement for analog telephones. Rather, it should be considered a supplemental service to support a mixture of voice and data applications. Also, note that ISDN adds no real value to most voice communications, other than enhanced voice quality and extending functionality from an ISDN-based PBX system. Call centers, however, prefer PRI trunks, as they deliver Calling Line IDentification (CLID) information at no additional cost and are useful in transferring calls within the PSTN.

Also note that ISDN is required on both ends of the connection in order to provide any benefit. An ISDN BRI, for example, is of no value for access to the Internet or Web unless ISDN is in place at the ISP site. Similarly, ISDN BRI is of no value to the telecommuter for corporate intranet access unless the host location also supports ISDN in the form of either BRI or PRI.

It also is worth noting that ISDN can be wasteful of network resources in an Internet or intranet data application. Remember that ISDN is a circuit-switched service. Also you might remember that circuit switching provides temporary, continuous, and exclusive connectivity. Now, consider that interactive data applications such as the Internet, the Web, and intranets are best served by packet-based networks, as discussed earlier in this chapter. Some years ago, several carriers, including Pacific Bell and US West, sought to stimulate IDSN through tariffs that negated usage charges during off-peak hours. Those pricing plans proved highly successful because they encouraged the use of ISDN for purposes of Internet access. However, the increased usage of the PSTN, which is highly inefficient for bursty, low-volume packet data applications, caused substantial network congestion. Since then, manufacturers of carrier-class equipment have developed devices that recognize the telephone number of an ISP and shunt that traffic around the circuit switch to a packet-switched network, thereby eliminating this issue in properly equipped COs.

ISDN certainly has a lot of benefits to offer in comparison to the PSTN, and it has some limitations. Ultimately, the greatest limitation is that of bandwidth. While ISDN offers much greater bandwidth than the PSTN, plus the advantage of digital communications, IDSN BRI bandwidth pales in comparison to DSL, cable modem, and Fiber-To-The-Premises (FTTP) technologies, all of which are capable of running in the Mbps range. At the PRI level, ISDN does not compete well against fiber-optic-based technologies.

As mentioned earlier in this chapter, ISDN standards are voluminous. While still under development in some respects, current ITU-T standards for ISDN include the following:

· I.515: Parameters for ISDN internetworking

· Q.700: Signaling System Number 7 (SS7) specifications

· Q.921: Layer 2 specification for D channel; Link Access Protocol, D Channel (LAPD)

· Q.931: Layer 3 User Network Interface (UNI) specifications

· V.110: B-channel procedures (Europe) for Terminal Adapters (TAs)

· V.120: B-channel procedures (North America) for TAs

In addition to the ITU-T, other organizations actively develop and promote ISDN standards. For example, ANSI (United States) and ETSI (Europe)each lobby the ITU for the acceptance of their parochial ISDN variations.

The applications for ISDN are broad in range. While ISDN was long phrased a technology in search of an application, it later opened to applications developers, and aggressively so. There is no killer app for ISDN. Rather, there are a number of applications that, in total, enhance its future. A host of applications that benefit from the improved quality of digital networking and are bandwidth intensive are well served by ISDN. Further, ISDN offers an affordable circuit-switched alternative to DDS, Switched 56, and T-carrier, which simply cannot be cost justified in many cases.

Personal office internetworking, remote office internetworking, and telecommuting (or Telework) all are facilitated by the increased bandwidth and error performance offered by ISDN BRI. In such applications, file transfers and facsimile transmission are accomplished much more quickly and with much greater clarity; the improved quality of the voice communications presents an added benefit.

ISDN also is used for access to packet data networks, including X.25, Frame Relay, ATM, and IP, with users benefiting from the faster call setup and teardown time made available by virtue of SS7. Because either the B channels or D channels can be used for packet data transfer in a BRI implementation, ISDN offers additional flexibility and bandwidth utilization. Additionally, some manufacturers of Terminal Adapters (TAs) offer built-in X.25 PADs and Frame Relay Access Devices (FRADs) for end-to-end error correction [27].

In terms of vertical markets, ISDN is of particular interest in the health care and education sectors, largely because of its ability to support imaging and video through rate adaption. TeleMedicine, for example, enables specialists to diagnose and treat patients in remote areas based on video examination and transmission of X-ray images across error-free and high-speed ISDN links.

The applications for ISDN are virtually unlimited, at least in terms of the network services that you can access. Through a single ISDN local loop, voice, facsimile, data, video, and image information can be accommodated. Additionally, simultaneous access to multiple networks and network services can be accomplished, perhaps including circuit-switched voice, X.25 packet, and Frame Relay. From a user's perspective, ISDN is highly flexible. From a carrier's perspective, ISDN offers the advantage of consolidating access to multiple networks, thereby relieving the strain on local loop, switching, and transport facilities. In other words, the LECs can market ISDN as a single network access solution—sort of a one-stop shop.

Notably, ISDN often is used as a backup to Frame Relay. In the event of a Frame Relay network failure, such as the total failure experienced in the AT&T network in 1998, the ISDN circuit responds immediately without dropping the data session. This approach offers diversity of both networks and services, offering greater protection than a backup Frame Relay permanent virtual circuit (PVC), which also would be affected by a total network outage.

Worldwide ISDN has experienced differing levels of success due to various marketing approaches, pricing strategies, and, in some cases, aggressive government support. The Japanese government, for example, has lent strong support to the development and deployment of high-technology networks and network services.

Telecom Australia achieved much success in marketing PRI to large user organizations, in part as an alternative to leased lines. An unusual offering is that of semipermanent circuits within PRI, priced at approximately 50 percent of the cost of a dedicated circuit. In the competitive Australian telecom environment, this approach was successful in countering leased-line networks offered by alternative carriers such as Optus.

Deutsche Bundespost Telekom offers ISDN on a widely available and low-cost basis. In excess of 80 percent of Germany's population has access to ISDN within six weeks. Pricing is very attractive compared to leased lines.

An interesting variation on the ISDN theme is Always On/Dynamic ISDN (AODI), which enables the user to establish a LAN-like always-on ISDN BRI connection to an ISP server, corporate intranet server, or corporate video server, for example, through special equipment and using only the D channel. The D channel maintains the always-on logical link between the client and the server systems, enabling the transfer of data (e.g., e-mail, stock quotes, or news bulletins) at rates of up to 9.6 kbps over an X.25 switched virtual circuit and without the call set up time required for a circuit-switched connection. As illustrated in Figure 7.26, a portion of the D channel on the client side operates as a link to a X.25 packet network, with the Multilink Point-to-Point Protocol (MPPP) used for access to the network from the TA. The D-channel AODI service employs the X.25 packet protocol that is used for signaling and control purposes in the SS7 network and establishes an SVC over the packet network. On the client side, the TCP/IP suite is encapsulated within the X.25 logical channel carried by the D channel, in support of connectionless data transfer. Packet handlers route the data packets around the circuit-switched network and over the X.25 packet-switched network, thereby maximizing efficiency and avoiding the unnecessary use of the circuit-switched PSTN for such an inappropriate application. On the server side, the SVC can be terminated on a BRI D channel, a designated PRI B channel, or a high-speed serial link. Should the need arise, the equipment automatically activates one or both B channels for transfer of large sets of data or to establish a videoconference. Once the need for the B channel(s) has ceased, the equipment automatically terminates those channel connections, the cost of which typically is usage sensitive. The cost of D-channel usage in an AODI application generally is based on a flat rate per month, with surcharges for kilopackets or megapackets applying above a defined usage threshold [28, 29].

Chapter 8: Local Area Networks—Connectivity and Internetworking

Robert M. Metcalfe and David R. Boggs, Xerox Palo Alto Research Center,

Ethernet: Distributed Packet Switching for Local Computer Networks,

Association for Computing Machinery, 1976

Over time, it became possible for multiple users at dumb terminals to input information to the mainframe. The terminals were connected to ports on the mainframe through terminal controllers, or cluster controllers. Controllers essentially act as traffic concentrators to enable multiple inputs from clusters of slow terminals to share one of a limited number of very expensive ports on the host computer. As time ticked away, Remote Job Entry (RJE) was developed to enable groups of users seated at remote clusters of dumb terminals to enter data from a remote location, connecting through a cluster controller to the mainframe over a Wide Area Network (WAN), perhaps over a Dataphone Digital Service (DDS) circuit.

Parallel to the development of data networking, the computers themselves began to change. Computers became more powerful as processor speeds increased with the development of ever-faster microprocessors on silicon chips. Memory became more available as chip technology and hard-drive technology both improved. Additionally, computers became ever smaller and less expensive, to the point that the current typical desktop Personal Computer (PC) is equivalent to an early mainframe that would have filled a moderate-size office building. The PC was legitimized by the introduction of the IBM PC in 1983. By 1993, an estimated 75 percent of professionals in the United States had a workstation on the desktop. In this day and age of the mobile professional, many of those PCs are laptops, and tablet and hand-held PCs are increasing in power by the day.

It was logical that all of this computing power and storage capability on all of these desktops would lead to a need to network those devices within the workplace. And it seems as though data traffic is more or less in line with the Pareto principle, with an estimated 80 percent of data transfer confined to the workplace, and only 20 percent traveling across the WAN. Whether or not that figure is accurate, PC users clearly have a requirement to share access to hosts, applications, databases, and printers. They also require a means to share access to WANs. Local Area Networks (LANs) provide a solution to those requirements.

Ethernet clearly remains the most popular LAN standard. In part, that popularity is due to the fact that 802.3 was the first standard. In part, it also is due to the inherent simplicity of Ethernet, as compared to other standards such as Token Ring. Also, DEC's chip design team sourced chip manufacturing to Intel, Advanced Micro Devices, and Mostek, thereby creating a highly competitive environment that quickly led to low chip prices. Token Ring is considerably more complex and costly. Developed by IBM, Token Ring chips were sourced exclusively to Texas Instruments. The impact of this decision was that of higher prices due to the lack of competition [4]. According to Metcalfe's estimates, in 1994 there were 50 million Ethernet-connected computers, 5 million of which were on 10-Mbps networks [5]. Further, 500,000 Ethernet networks were TCP/IP registered and 50,000 were connected to the Internet. No doubt those numbers have increased considerably in the past dozen years, as have the speeds at which Ethernet runs.

This discussion of the basic concepts of LANs and LAN internetworking serves as the launching pad for discussion of the network technologies of the future. In this chapter, I address the definition, origin, and evolution of LANs and their application. Dimensions of LANs to be explored include media alternatives, physical and logical topology, baseband versus broadband, Medium Access Control (MAC), and standards and standards bodies. I define and illustrate bridges, routers, hubs, switches, and gateways as well as LAN operating systems. This chapter concludes with a discussion of LAN internetworking, remote LAN access, and recent developments in the realm of high-speed LANs and wireless LANs and Personal Area Networks (PANs).

8.1 LANs DEFINED

Generally, LAN specifications are the province of the Institute of Electrical and Electronics Engineers (IEEE), although the American National Standards Institute (ANSI) and other standards bodies are involved, and the regulators are very much involved in spectrum allocation in the Wireless LAN (WLAN) domain. LANs operate at Layer 1, the Physical Layer, and Layer 2, the Data Link Layer, of the Open Systems Interconnection (OSI) Reference Model. Raw bandwidth ranges up to 10 Gbps, although actual throughput often is much less. LANs are limited to a maximum distance of only a few miles or kilometers, although they often operate within a much more confined area measured in feet or meters. LANs support the transmission of data in frame format, with the frames varying in size within specified minimums and maximums.

LANs are used almost exclusively for data communications over relatively short distances such as within an office, office building, or campus environment. LANs enable multiple workstations to share access to multiple host computers, other workstations, applications and databases, printers and other peripherals, and WAN connections. LANs traditionally are used in computer data applications, although they increasingly support video and voice communications as well.

8.2 LAN DIMENSIONS

LANs can be characterized along a number of common dimensions, for ease of understanding. Those dimensions include transmission medium, physical and logical topology, baseband versus broadband, and medium access control method.

In retrospect, perhaps the use of coaxial cable lessened the resistance of DP/MIS managers to the concept of LANs. Those who lived in the mainframe world (most did) regarded PCs with disdain and sneered at twisted pair, which they referred to as telephone wire.

The advantages of coaxial cable include high bandwidth and exceptional error performance over relatively long distances as the thick inner core conductor results in fairly modest signal attenuation. Further, the outer shield rejects ElectroMagnetic Interference (EMI) and Radio Frequency Interference (RFI) as well as providing excellent security. Coax is also highly durable, but the costs of acquisition, deployment, and reconfiguration are high. While the disadvantages of coaxial cable have been mitigated to a large extent through the development of new coax designs, those designs also affect system performance. By way of example, consider three variations on the coax theme: ThickNet, ThinNet, and Twinax.

· ThickNet: Thick Ethernet, also known as 10Base5, was approved by the IEEE in 1983. 10Base5 uses traditional thick coax, often referred to as goldenrod, referring to its high cost, high value, and the yellow cable sheath used by some manufacturers. Other manufacturers used orange cable sheaths for thick coax, giving rise to the term orange hose. 10Base5 translates to 10Mbps, Baseband (one transmission at a time over a single, shared channel), and 500m maximum segment length. While individual devices can be separated by much greater distances across the network, issues of signal attenuation limit each segment, or link, in the network to approximately 500 m.

· ThinNet: Thin Ethernet, also known as 10Base2, was approved by the IEEE in 1986. 10Base2 uses coax of thinner gauge. The thinner cable is less costly to acquire and deploy, although its performance is less in terms of transmission distance. 10Base2 translates to 10 Mbps, Base band, and 200m maximum segment length (actually 185 m, rounded up).

· Twinax: Twinaxial cable, resembles ThinNet coax, but with twin coaxial conductors, rather than one. Twinax is used in older IBM midrange systems such as Systems 34, 36, and 38 as well as the younger IBM AS/400 and RS/6000. More recently, the IEEE has developed the 10GBase-CX4 standard in support of 10-Gigabit Ethernet (10GbE). Based on the Infiniband high-speed cable assemblies, the specification calls for twinax assemblies operating over distances up to 50 ft. The standard calls for four transmitters and four receivers operating differentially in simplex mode over a bundle of eight twinax cables, with each simplex transmission occurring at 2.5 Gbps at a frequency of 3.125 GHz per channel with 8B/10B line coding. The cost of this patch cord technology is expected to be approximately th that of comparable 10GBase-optical solutions.

Category 3 (Cat 3) UTP also is used for 4-Mbps Token Ring LANs. Category 4 (Cat 4) UTP, developed for 16-Mbps Token Ring LANs, has a bandwidth of 20 MHz. In addition to its application in Cat 6 and 7 cables, STP sometimes is used in high-noise environments in which UTP data transmission is especially susceptible to EMI or RFI. Examples include manufacturing environments where there are large numbers of powerful machines, power plants, and old buildings (e.g., hospitals, or government or military facilities) where it might be impossible when installing a LAN to avoid placing wires close to electric motors or older fluorescent light fixtures.

Because of its outstanding performance characteristics, optical fiber also is used extensively in contemporary LAN applications. Its cost and fragility, however, generally relegate it to use as a backbone technology in Fiber Distributed Data Interface (FDDI) networks, for example.

The advantages of fiber certainly include the combination of high bandwidth and excellent error performance. Additionally, fiber performs well over long distances and offers excellent security. The disadvantages of fiber transmission systems include their high cost of acquisition, as compared to UTP systems. While the fiber, itself, is not significantly more expensive than Cat 5e UTP, the light sources and detectors are considerably more expensive than the metallic interfaces used with UTP. As fiber is very fragile, it must be protected carefully, and redundancy is always a good idea.

Wireless LANs (WLANs) offer the obvious advantage of avoiding much of the time and cost associated with deploying wires and cables. This is especially important in a dynamic environment where portability is desirable, such as an office where cubicles are frequently reconfigured. WLANs have found acceptance in providing LAN capabilities in temporary quarters, where costly cabling soon would have to be abandoned, and in older buildings, where wires are difficult or impossible to run. WLAN technologies include both RF and IR.

Most WLANs use unlicensed frequencies in the 2.4-and 5-GHz ranges. This approach avoids the expensive and time-consuming spectrum licensing process. However, the use of unlicensed frequencies creates significant potential for interference from other systems in proximity that use the same frequencies. A wide variety of other devices (e.g., garage door openers, bar code scanners, industrial microwave ovens, and cordless telephones) run in the same frequency ranges, which are in the Industrial/Scientific/Medical (ISM) bands. The risk works in both directions, of course, as there is the potential that you could click your mouse and cause garage doors to open and close all over the neighborhood. (That's a joke, just to keep things in perspective, although I suppose it's possible.) The systems that use licensed frequencies avoid the potential for interference but require that the manufacturer carefully police the deployment of such systems on a site-specific basis and under the terms of an omni-license.

A good WLAN example is the IEEE 802.11b Ethernet standard, aka Wi-Fi, which begins at theoretical maximum transmission rates of 11 Mbps, with optional fallback to 2 Mbps in noisy environments. The large number of products supporting that specification generally deliver actual throughput of considerably less, as they are sensitive to factors such as background noise, distance, physical obstructions, and competing systems running in the same band. The more recent 802.11a specification runs at much higher speeds in the 5-GHz range. The 802.11 standards use the same collision-based access protocol as is optional with more traditional wired Ethernet. I discuss WLANs in more detail toward the end of this chapter.

The physical topology, or layout, of a LAN is in the form of a bus, ring, or star. As trains, ovals, planets, and constellations are not defined, you should avoid vendors promoting such topologies.

The original, classic Ethernet was based on a bus topology comprising coaxial cable segments that could be a maximum of 500m in length. Each segment supports as many as 1024 (210 = 1024) network addresses, each of which is associated with an attached device. The maximum segment length was due to issues of attenuation at the relatively high carrier frequency. Ethernet segments could connect through bridges, which function as signal repeaters. The total route length of the entire Ethernet was limited to 2.5 km, which is a function of both signal propagation time and MAC mechanisms. I discuss CSMA/CD and CSMA/CA, the Ethernet MAC standards, later in this chapter.

A tree topology is a variation on the bus theme, with multiple branches off the trunk of the central bus. Bus networks also suffer from the vulnerability of the bus—should the bus be compromised, the entire network is compromised. Similarly, tree networks are dependent on the integrity of the root bus [2].

Star topologies (Figure 8.4) consist of a central node, or point of interconnection, to which all other devices are attached directly, generally via UTP or STP. The central node is in the form of a hub, switch, or router. Transmission rates vary, with 10Base-T operating at 10 Mbps and 100Base-T at 100 Mbps. Perhaps the greatest advantage of a star is that individual devices can connect to the node via UTP or STP. Another key advantage of a star is the ability to isolate a dis-ruptive or failed station, thereby eliminating any negative effect it may have on overall LAN performance. Additionally, multiple attached devices simultaneously can share the full bandwidth of the LAN, at least in a switched environment. The primary disadvantage is that a hub failure is catastrophic; as all connectivity is provided through the central hub, its failure affects the entire LAN. Examples of star configurations include 10Base-T and 100Base-T. AT&T's StarLAN and DataKit and 100VG-AnyLAN are now considered obsolete. Asynchronous Transfer Mode (ATM), also based on a star topology, enjoyed the LAN switching spotlight briefly but was overwhelmed by switched Ethernet at 100 Mbps and 1 Gbps.

There are single-cable and dual-cable broadband LANs. A dual-cable LAN involves one cable in support of upstream transmissions from the stations to the head end and a second cable in support of downstream transmissions from the head end to the stations. In a single-cable system, specific frequency channels are designated for upstream transmissions and others for downstream transmissions. The stations connect to the cable through multiport Media Access Units (MAUs) that house the RF modems and that connect to the coax cable, as illustrated in Figure 8.5. All transmissions go through the head end. Broadband LANs commonly use 75-ohm CATV–type coax and use CATV-style connectors, taps, filters, and amplifiers in a tree-and-branch architecture, which essentially is a variation of the bus with multiple branches off of a main root bus, much as there are branches off of the main trunk of a tree.

Baseband LANs are digital and single channel in nature, supporting one transmission at a time (see Figure 8.6). The full range of media options are available, including both wired (i.e., coax, UTP, STP, and fiber-optic cable) and wireless (i.e., RF and IR). Distance limitations depend on the medium employed and the specifics of the LAN protocol. Baseband LAN physical topologies include ring, bus, and star.

Medium access control (MAC) describes the process that is employed to control the basis on which devices can access the shared medium. Some level of control is required to ensure, or at least improve, the ability of all devices to access the network within a reasonable period of time, thereby resulting in acceptable access times and response times. It also is important that some method exist to either detect or avoid data collisions, which are caused by multiple transmissions placed on the shared medium simultaneously, and to recover from them.

Medium access control takes place at Layer 1, the Physical Layer, and Layer 2, the Data Link Layer, of the OSI Reference Model. Medium access control is accomplished at the Network Interface Unit (NIU), or Network Interface Card (NIC). A NIU or NIC is at the board level, with the boards typically fitting into an expansion slot of an attached device (e.g., workstation, printer, or server). Alternatively, multiple cards may be contained within a multiport device that supports multiple workstations on a pooled basis. Each NIU/NIC has a unique logical address for purposes of identification, with the address hard coded on a silicon chip at the time of manufacture. Medium access control can be accomplished on either a centralized or decentralized basis and can be characterized as either deterministic or nondeter-ministic in nature.

Decentralized control is somewhat anarchistic, as each station assumes responsibility for controlling its access to the shared network. Additionally, each station must assume responsibility for detecting and resolving any data collisions that might occur in the quite possible event that its access to and transmission over the shared medium overlaps with that of other devices. Decentralized control networks generally use a nondeterministic, or contentious, MAC mechanism. By way of example, Ethernet control is decentralized.

Deterministic access is a MAC convention that enables both the centralized master station, which commonly is in the form of a server, and each slaved station to determine the maximum length of time that passes before access is provided to the network. In other words, each station can be guaranteed the right to communicate within a certain time frame. Additionally, the system administrator can assign access priorities. Deterministic access is also known as noncontentious because the devices do not contend for access; rather, access is controlled on a centralized basis.

Deterministic access employs token passing. The token, which consists of a spe-cific bit pattern, indicates the status of the network—available or unavailable. The token is generated by a centralized master control station and transmitted across the network. The station in possession of the token controls the access to the network. That station may either transmit or require other stations to respond. Transmission is in the form of a data packet of a predetermined maximum size, determined by the number of nodes on the ring and the traffic to be supported; oversized transmissions are segmented, or fragmented. After transmitting, the station passes the token to a successor station in a predetermined sequence. While the process is complex and overhead intensive, its high level of control over the network avoids data collisions.

Manufacturing Automation Protocol (MAP) is a good example of a deterministic protocol. Developed by General Motors (GM) in the early 1980s for the interconnection of computers and programmable machine tools in factory or assembly line operations, MAP is based on Token Bus (IEEE 802.4) running at 1, 5, 10, and 20 Mbps. GM developed MAP as a multivendor solution using off-the-shelf parts, including coax cable, taps, connectors, amplifiers, splitters, and terminators. MAP sometimes is referred to as Manufacturing Automation Protocol/Technical and Office Protocol (MAP/TOP). (Note: The 802.4 committee disbanded in 2004 as the Token Bus standard has been in hibernation for some years. Token Bus is yet another casualty of Ethernet.)

CATV providers in a convergence scenario are applying deterministic protocols for Internet access, which is provided to large numbers of residences, businesses, and schools over a two-way coaxial cable system terminating in cable modems at each customer premises. As much as 500+ Mbps is reserved for such applications, with as much as 10 Mbps available to an individual user. In order to manage contention over such a network, which extends over fairly significant distances, a token-passing MAC technique is employed, with the master station positioned in a Cable Modem Termination System (CMTS) physically located at the CATV provider's head end.

General characteristics of token-based networks include a high level of access control, which is centralized. Access delay is measured and ensured, with priority access supported. Throughput is very close to raw bandwidth, as data collisions and therefore retransmissions are avoided. Throughput also improves under load, although absolute overhead is higher than with nondeterministic access techniques. Deterministic access standards include Token-Passing Ring, IBM Token Ring (Figure 8.7), and Token Bus.

Token-based LAN technologies are somewhat overhead intensive, due to the token-passing and management processes. But they can more than compensate for that fact by avoiding data collisions and the retransmissions required to recover from them. Token Ring, for example, comes in 4-and 16-Mbps flavors; in each case, bandwidth utilization is virtually 100 percent under full load [11].

CSMA is a decentralized, contentious MAC method used in Ethernet and other bus-oriented LANs. The carrier frequency is sensed by each of multiple stations, or nodes, to determine network availability before accessing the medium to transmit data. Further, each station must monitor the network to determine if a collision has occurred because collisions render the transmission invalid and require a retransmission.

CSMA works much like an old telephone party line, where there are multiple subscribers with individual logical addresses in the form of telephone numbers, all of which are connected to a single physical circuit. When placing an outgoing call, the subscriber must pick up the telephone to monitor the line for a short while to sense the level of activity. If there is no activity, a call can be placed. If there already is a call in progress, another call attempt cannot be made without causing interference (and hard feelings). Rather, the subscriber must hang up the telephone and subsequently monitor the circuit on some basis in order to determine its availability. Incoming calls are addressed to each party on the party line by varying the number of rings, indicating the unique logical address of each party sharing the circuit, and only the target party can answer the call without violating security and thereby creating ill will. In other words, MAC protocols govern the manner in which the circuit is managed to the satisfaction of all parties. As a result, collisions are less likely. Variations on the theme include Nonpersistent CSMA, 1-Persistent CSMA, and P-Persistent CSMA:

· 1-Persistent CSMA also allows a machine to transmit data whenever it senses an idle channel. If the channel is in use, the machine will continuously sense it until the channel becomes free. The protocol gets its name because the machine is persistent, that is, tenacious or obstinate, in its monitoring of the channel and transmits with a probability of 1.0, that is, 100 percent certainty of access success, whenever the channel is idle. If the network includes a large number of stations persistently monitoring the network, a great many of them might sense the availability of the network and begin to transmit simultaneously, virtually guaranteeing a collision. In such a scenario, 1-Persistent CSMA can be characterized as eager, if not downright greedy.

· P-Persistent CSMA allows a machine to transmit a frame during an idle time with probability P or lower, based on the length of the idle time as measured by a time slot. A time slot is the maximum packet transmission time for a station at one extreme end of the network to send a packet to a station at the opposite extreme end of the network and is based on the physical length of the cable, the physical size of the frames, and the speed of signal propagation through the wire. If a machine senses an idle condition on the channel, it transmits with probability P for one time slot. The machine then delays for worst-case propagation delay for one packet with probability 1-P. If the channel is busy, the machine listens persistently until the channel becomes idle and starts over. For example, P.01 means that there is a probability of 1 percent that the transmission will be unsuccessful. If P is set very low (e.g.,.01), throughput is nearly 100 percent, but transmission delays will be very long, as the machine will wait a very long time between idle periods. This highly cautious approach introduces useless delays at low loads but certainly improves the rate of successful transmissions at high loads.

CSMA is implemented in two standard means: CSMA/CD and CSMA/CA. In either case, latency and throughput degrade under heavy loads of traffic; for example, a classic Ethernet network running at the theoretical speed of 10 Mbps typically delivers throughput of no more than 4–6 Mbps, and often much less. Note that CSMA is half-duplex (HDX) in nature, and only one transmission can take place at any given time. While it is less costly than Token Ring networking, it also delivers less efficient bandwidth utilization.

CSMA/CD is the most common MAC method used in bus networks (Figure 8.8). In an Ethernet environment, for example, the transmitting station sends a data packet in both directions of the bus. The 802.3 Ethernet frame (i.e., packet) takes the following form:

· Start Of Frame (SOF): The SOF delimiter is a single octet that ends with two consecutive 1 bits that serve to synchronize the receiving stations on the rate of transmission.

· Destination and Source Addresses: The destination address and source address fields are the addresses of the target station and the originating station, respectively. Each address comprises each six octets, the first three of which are speci-fied by the IEEE on a vendor-dependent basis and the last three of which are assigned by the vendor. The address is hard coded on the NIC at the time of manufacture. Theoretically, each NIC and therefore each station have a unique address.

· Length: The length field of two octets indicates the number of octets of data in the data field.

· Data: 802.3 Ethernet frames have a lower limit of 64 octets and an upper limit of 1518 octets. In consideration of the fact that 18 octets are consumed with layer 1 and layer 2 processing, the data field, or payload, must comprise 46–1500 octets. In the event that the payload is less than 46 octets, padding bytes are inserted.

· Frame Check Sequence (FCS): The FCS consists of a 32-bit Cyclic Redundancy Check (CRC) that is appended to the frame trailer for purposes of error control [12].

Several factors help to determine the maximum packet size. First, and as the Ethernet is shared, it is necessary to fragment a file transfer into smaller subsets of data so one transmission does not speak to the entire level of bandwidth available across the network for a long period of time. Second, the maximum frame size is a trade-off between raw efficiency and throughput. In other words, a stream of unfrag-mented, unframed data is most efficient because it requires little overhead. But a single bit error in a data stream associated with a bulk file transfer might require that the entire file be retransmitted in order to correct for the error. This process might have to repeat an infinite number of times, with the file never making it from transmitter to receiver without error. By fragmenting the file into subsets of data in the form of frames, the level of overhead increases, as each frame requires 18 octets of overhead. An errored bit, however, is confined to a single frame, which easily can be retransmitted without significant likelihood of error and without serious degradation of overall throughput across the LAN.

The minimum packet (frame) size is a direct function of the design of the classic Ethernet and the CSMA/CD control mechanism. In the most extreme case, an Ethernet comprises a number of segments, each of which is up to 500m in length, supports as many as 1024 addresses, involves a great number of transceivers, and connects to one or more other segments through a bridge. The total route length of the classic Ethernet is up to 2.5 km. Given issues of propagation delays across the segments and through the bridges, it takes a certain amount of time for a frame to traverse the network from the originating device to the target device, which is 2.5 km away in the most extreme possible case. If that frame of data encounters a data collision at the distant end just before it reaches the target device, it takes an identical amount of time for a collision notification to be received by the transmitting device across the subcarrier channel. Then a retransmission is used to adjust to that fact. If such notification is not received in time, the originating device assumes that the data were received in good form when that is not at all the case. Therefore, the minimum frame size in classic Ethernet is 64 octets, which includes 46 octets of payload and 18 octets of overhead.

Wireless LANs, as standardized in IEEE 802.11, employ CSMA/CA. The 802.11 standard uses a positive ACKnowledgment (ACK) mechanism which requires that the transmitting station first check the medium to determine its availability. The transmitter sends a short Request-To-Send (RTS) packet that contains the source and destination network addresses as well as the duration of the subject transmission. If the shared medium is available, the destination station responds with a Clear-To-Send (CTS) packet. All devices on the network recognize and honor this acknowledged claim to the shared network resources. If the source station does not receive an ACK packet from the destination station, it retransmits RTS packets until access is granted.

8.3 LAN EQUIPMENT

In addition to the attached transmit and receive devices, aka nodes or stations, LANs may make use of other devices to control physical access to the shared medium, extend the maximum reach of the LAN, switch traffic, and so on. Such hardware is in the form of NICs/NIUs, transceivers, MAUs, bridges, hubs, routers, and gateways. As is true of much of the technology addressed in this book, the lines increasingly blur between these devices. Therefore, I focus on the classic definitions, expanding on those concepts and introducing discussion of multifunctional devices as appropriate.

Also known as Network Interface Units (NIUs), Network Interface Cards (NICs) are chipsets on printed circuit boards that provide physical access from the node to the LAN medium. The NIC is responsible for fragmenting the data transmission and formatting the data packets with the necessary header and trailer. A standard IEEE NIC contains a unique, hard-coded logical address, which it includes in the header of each data packet it transmits. The NIC typically has some amount of buffer memory, which enables it to absorb some number of bits transmitted by the associated device, form the packets, and hold them until such time as the network is available. In the context of the OSI Reference Model, NICs function at the lower two layers, the Physical and Data Link layers. The NIC also may contain a microprocessor that can relieve the attached device of some routine computational functions.

The NIC (refer to Figure 8.9) can take a number of forms, including a circuit board that fits into the expansion slot of a desktop PC, a PCMCIA card, or a stand-alone device. A NIC commonly is embedded in a desktop, laptop, tablet, or hand-held computer at the time of assembly. Transceivers (transmitter/receivers) are used in LANs to receive a carrier signal and then transmit it on its way. They are embedded in NICs/NIUs and MAUs. MAUs (Medium Access Units, or Multistation Access Units) are stand-alone devices that contain NICs in support of one or more nodes. MAUs are very unusual in contemporary networking.

It also is possible for a bridge to support multiple LANs of disparate origin. For example, Ethernet-to-Ethernet and Token Ring-to-Token Ring connectivity can be provided [13]. It also is possible to interconnect disparate LANs, such as Ethernet to Token Ring, through the use of an encapsulating bridge. Such a bridge encapsulates the native LAN frame, surrounding it with control information appropriate to the LAN to which the target device is attached (Figure 8.11).

Medium Access Control (MAC) bridges are more sophisticated, still. MAC bridges have the ability to connect disparate LANs (e.g., Ethernet to Token Ring). This is accomplished through the process of encapsulation, or translation, as depicted in Figure 8.10. When operating in this mode, the bridge alters the frame format by encapsulating, or enveloping, the original frame with control data specific to the protocol of the destination LAN supporting the target device. Such an approach might be used to connect an Ethernet LAN to a Token Ring LAN or where two Ethernet LANs are connected via an intermediate FDDI backbone [15].

Specific bridge protocols include Spanning Tree, Source-Routing Protocol, and Source-Routing Transparent:

· Spanning Tree Protocol (STP) bridges, also known as learning bridges, are defined in IEEE 802.1 standards. Spanning tree bridges are self-learning, filtering bridges for use in connecting LANs on a point-to-point basis. The bridge is programmed or teaches itself the addresses of all devices on the network; subsequently, the network tree of the bridge provides only one span (link) for each LAN-to-LAN connection. Some spanning tree bridges also have the capability to provide security by denying access to certain resources based on user and terminal ID. Bridges that support the spanning tree algorithm have the ability to automatically reconfigure themselves for alternate paths if a network segment fails, thereby improving overall reliability [16].

· IBM Source-Routing Protocol (SRP) packets are programmed with specific routes (i.e., lists of bridges), based on considerations such as the physical location of the nodes and the capacity of the links involved. The maximum number of bridges hopped is 13. SRP bridges are most commonly used in Token Ring networks.

· Source-Routing Transparent (SRT) is defined in the IEEE 802.1 standard. It is effectively a combination of STP and SRP. The SRT router can connect LANs by either method, as programmed [2].

Hubs reflect the trend toward star and away from bus configurations. Hubs can be either active or passive. Passive hubs act simply as cable-connecting devices, while active hubs also serve as signal repeaters [2]. The first generation of hubs (1984) acted as LAN concentrators and repeaters, with a single internal collapsed backbone bus for connecting like LANs. The second generation accommodated multiple LAN architectures (e.g., Ethernet and Token-Passing Ring) over separate ports, with rudimentary network management and configuration capabilities included as well [15].

This collapsed backbone is analogous to the method by which electrical wiring is run in a home or office. One approach is that of running a thick electrical bus cable all through the walls, floors, and ceilings, splicing in outlets as required. The better approach is that of collapsing the bus and placing it in a centralized circuit breaker box, where the connection to the wide area electrical grid is made and common electrical ground is established. Connections to outlets are made through circuit breakers that snap into the common electrical bus, with some circuit breakers being for power users (e.g., washing machines and dryers) and others for devices drawing less power (e.g., lamps, hair dryers, and food processors). The appliances plug into outlets that are connected to the common electrical bus by relatively thin gauge electrical wiring that is fairly inexpensive and easily installed. In the event that a device causes a problem (e.g., shorts out or draws too much power), it is isolated automatically when the circuit breaker trips.

Intelligent hubs, the third generation, provide multiple buses for multiple LANs of either the same or disparate architectures, in much the same manner as encapsulating bridges, which are illustrated in Figure 8.11. They can support multiple media (e.g., UTP and fiber), multiple speeds (e.g., 10/100 Mbps), and multiple LAN protocols (e.g., Ethernet and Token Ring). (Note: Support for multiple LAN protocols is unusual in contemporary networking, as Ethernet has overwhelmed its competition.) As addressable devices, intelligent hubs can be managed centrally via Simple Network Management Protocol (SNMP) or another appropriate network management protocol. Intelligent hubs also provide bridging and basic routing capabilities [15].

Regardless of the generation of the hubs, they serve, at minimum, as central points of interconnection for LAN-attached devices. Additionally, they serve as concentrators of LAN traffic and as repeaters, with multiple hubs interconnected through high-speed media [17]. A number of hub manufacturers offer stackable hubs, which offer the advantage of scalability; in other words, the hubs can be physically stacked and interconnected to increase port and traffic capacity [18]. Hubs, like bridges, do a good deal to reduce congestion through LAN segmentation and at very low cost.

LAN switches are intelligent hubs with basic packet store-and-forward cap abilities that can support multiple simultaneous transmissions. Switches have the ability to read the target addresses of the packets and forward them only and directly to the appropriate port associated with the target device. That device may be directly attached to the switch, may be attached to a lesser workgroup switch, or may be connected to a hub that connects to the switch, as illustrated in Figure 8.13.

The LAN switch architecture can take several forms:

· Shared Bus: A shared bus switch has a single high-speed bus that is shared by all incoming and outgoing ports on a Time Division Multiplexing (TDM) basis. This is a relatively low cost approach commonly used in smaller workgroup-level switches where issues of congestion typically are relatively modest.

Along another dimension, switches may use several methods of operation:

· Cut-Through: A cut-through switch quickly reads the address of the data packet and quickly flows the frame through the switching matrix bit by bit.

· Store and Forward: A store-and-forward switch temporarily buffers, or stores, the frame as it is presented to the incoming switch port, examining the entire frame for errors through a CRC check before forwarding it through the switching matrix to the output port. While cut-through switching is faster and less expensive, it carries with it the risk of the propagation of errored data and the resulting potential for negative impact on overall throughput, as errored frames ultimately must be retransmitted. Therefore, store-and-forward switching generally is preferred over cut-through switching.

· Fragment Free: A third, and less common, method is fragment-free switching, which is similar to cut-through except for the fact that the switch stores the first 64 octets of the frame before forwarding it. As most errors occur at the beginning of a frame, this approach eliminates the possibility that runt frames, that is, truncated frames, will be transmitted [20]. Note: Recall that the minimum size of an Ethernet frame is 64 octets. The small frame size is used for VoIP over switched Ethernet.

Switches operate at the Physical and Data Link Layers of the OSI Reference Model—Layers 1 and 2, respectively. Switches read the destination addresses of the packets, filtering and forwarding as appropriate, based on MAC addresses (Layer 2). Switch logic is relatively simple and is in the form of firmware at the chip level. Therefore, switches are fast and relatively inexpensive. Some switches make routing decisions based on IP addresses (Layer 3). Layer 3 switching (admittedly an arguable term) involves a combination of switching and routing. This involves more complex routing decisions that are made in the context of the network as a whole yet not at the level of complexity that characterizes a router.

As illustrated in Figure 8.13, LAN switches may be positioned either at the workgroup level or in the backbone, or core.

· Workgroup Switch: A workgroup switch commonly has multiple 10/100-Mbps ports in support of workstations, perhaps connected through hubs, printers, and servers of various kinds.

· Backbone Switch: A backbone, or core, switch serves to interconnect workgroup switches and often provides access to large servers or server clusters. A core switch typically runs at 1 Gbps or perhaps 10 Gbps.

A switch does a great deal to reduce congestion and in a number of ways. First, a switch can support multiple simultaneous transmissions, whether through a matrix, a shared bus, or a combination. Second, switches serve to segment a network through filtering, as they forward traffic only to the port associated with the link to which the target device is connected. Thereby, that traffic does not contribute to congestion on other links or segments. Third, a switch can be equipped to buffer incoming packets until internal bus resources are available to process them. A switch also can be equipped to buffer outgoing packets until such time as the link to the next switch becomes available. Fourth, a switch can exercise a flow control mechanism, whereby it can advise a device to stop transmitting when its buffers are in danger of overflowing and then advise the device to resume transmission when the pressure on resources has been relieved. Fifth, store-and-forward and fragment-free switches variously eliminate or reduce the number of errored frames. Finally, a switch supports full-duplex transmission, thereby reducing or eliminating data collisions associated with CSMA in an Ethernet environment, assuming that the station is directly connected to the switch rather than through a hub. This approach is the current best practice.

The cost of switches has dropped dramatically in recent years to the point that they often compete effectively against hubs. But switch costs are sensitive to factors such as the type and speed of the transmission media interfaces, the number and speed of the ports, the number and size of the buffers, the number and speed of the internal buses, the complexity of the internal switching matrix, and the complexity of the switching or routing logic.

In addition to supporting filtering and encapsulation, routers route traffic based on a high level of intelligence that enables them to consider the network as a whole. This is in stark contrast to bridges, hubs, and switches, which view the network simply on a link-by-link basis. (Note: Layer 3 switches cross the line into the routing domain, although at a relatively simple level.) Routing considerations might include destination address, payload type, packet priority level, least-cost route, minimum route delay, minimum route distance, and route congestion level. Routers also are self-learning, as they can communicate their existence to other devices and can learn of the existence of new routers, nodes, and LAN segments. Routers constantly monitor the condition of the network as a whole, thereby dynamically adapting to changes in the condition of the network from edge to edge. Routers are multiport devices with high-speed ports running at rates up to 155 Mbps or more and with high-speed internal buses that can be on the order of 1 Gbps in the aggregate. Additionally, routers typically provide some level of redundancy so they are less suscep-tible to catastrophic failure [15].

Routers commonly are capable of alternate routing and inverse multiplexing. As noted by Charles Darling [14], what dedicated WAN links lack in cost efficiency they make up for in lack of reliability (e.g., backhoe fade). Routers are available that sense a network failure and reestablish the connection via an alternate means, which may include a totally different network and service. While a backup ISDN Basic Rate Interface (BRI) link may be painfully slow compared with a T1 connection, for example, Darling suggests that slow is better than stop, and I suspect that most would agree. Additionally, some routers offer inverse multiplexing capabilities over ISDN and T-carrier facilities.

Router protocols include interrouter protocols, serial line protocols, and protocol stack routing and bridging:

· Interrouter protocols are router-to-router protocols that can operate over heterogeneous networks. These protocols pass routing information and keep-alive packets during periods of idleness. Routing Information Protocol (RIP) is an example of in interrouter protocol. The Internet Engineering Task Force (IETF) specified RIPv1 in RFC 1058, RIPv2 in RFC 2453, and RFCng for IPv6 in RFC 2080. The IETF specified Open Shortest Path First (OSPF) in RFC 1131 (1989), and there have been multiple revisions since.

Gateways can perform all of the functions of bridges and routers as well as accomplish protocol conversion at all seven layers of the OSI Reference Model. Generally consisting of software residing in a host computer equivalent in processing power to a midrange or mainframe, gateway technology is expensive but highly functional.

8.4 LAN OPERATING SYSTEMS

A LAN Operating System (OS), or Network Operating System (NOS), is software that provides the network with multiuser, multitasking capabilities across the network. The OS facilitates communications and resource sharing, thereby providing the basic framework for the operation of the LAN. The OS consists of modules that are distributed throughout the LAN environment; some NOS modules reside in the servers and other modules reside in the clients.

I want to digress for just a moment. The client/server model originated with the development of the U.S. Department of Defense ARPANET in the 1960s. As the cost and size of computer systems decreased and as the capabilities of those systems and the networks increased, the embodiment of client/server changed. A contemporary client is an application that generally resides on a microcomputer. Example applications include word processing, spreadsheet, and database software. The client runs against a server, which is a multiport computer containing large amounts of memory and enabling multiple clients to share its resources while performing certain functions independently. Servers are database engines capable of processing client requests for information and managing the resident data. For example, client/server continues to be used extensively in the Internet. When accessing America Online, Prodigy, CompuServe, or another service provider, you make use of a Graphical User Interface (GUI) and browser software that resides on the PC. When initiating an Internet session, that software runs against software installed in the service provider's communications server. Through this approach, the two devices communicate effectively without requiring that the software be downloaded from the server as a part of every Internet session, which, because the graphic files are huge, would cost a great deal of time, and bandwidth would be wasted. Once connected to the communications server, you subsequently can access a large number of database servers on a point-and-click basis, courtesy of the GUI, and with only the target data being transmitted across the Internet.

8.5 VIRTUAL LANs

Virtual LANs (VLANs) are software-defined LANs that group users by logical addresses into a virtual, rather than physical, LAN through a switch or router (refer to Figure 8.15). The LAN switch can support many VLANs, which operate as subnets [21]. Users within a VLAN traditionally are grouped by physical ports on switches and routers, TCP port address, MAC address, or IP address. Each node is attached to the switch port via a dedicated circuit. A variation on the theme is a policy-based VLAN, which can base VLAN membership on such factors as protocol, location, user name, and workstation address [22]. Users also can be assigned to more than one VLAN, should their responsibilities cross workgroup domains.

The LAN switches can be networked, thereby extending the reach of the VLAN. The networking is generally provided through FDDI, 100Base-T, or GbE over optical fiber links. VLANs also can be extended across the WAN through access routers and various services such as dedicated leased lines, Frame Relay, and ATM.

8.6 REMOTE LAN ACCESS

Remote LAN access is the ability to access a LAN from a remote location. The need for remote LAN access is increasing worldwide, especially in the United States and Western Europe. Telecommuting and telework now are growing at a fast pace. Various market research firms suggested that 4 million to 9 million employees worked at least 8h per week from home in 1992. In 1998, an estimated 20 million people worked from home, and an additional 9 million worked at home after normal business hours [27]. Telecommuting estimates now range as high as 40 million in the United States and 16 million in the European Union.

The concept of remote LAN access is one of providing access from remote locations to one or more host computers, which typically are LAN attached. In support of telecommuters, contractors, remote offices, and the Small Office/Home Office (SOHO), remote LAN access often is essential to the operation of the enterprise. Additionally, remote access often is provided to customers, suppliers, trading partners, and so on. The yield is that remote users are provided access to resources with the same level of privilege as though they were on-site [17]. Key components to be examined are the network, the equipment, and the applications supported. Other issues include security management and network management.

The network is clearly a developing enabler. The network can assume a variety of forms, depending on issues such as availability, cost, and bandwidth (as illustrated in Figure 8.16). Network options literally run the full range of conventional data networks, including modem-based communications over the analog PSTN, ISDN, and IP-based Virtual Private Networks (VPNs) over ADSL and cable modems. Public Wi-Fi hotspots are now widely available and are often used by traveling employees for access to corporate networks through the Internet. Finally, and in order to support truly mobile remote LAN access, cellular radio data networks often are employed [28, 29].

8.7 LAN STANDARDS AND STANDARDS BODIES

Standards developed by the Institute of Electrical and Electronics Engineers (IEEE) largely have governed the world of LANs since the formation of Project 802 in February 1980. Project 802, working within the scope of the OSI Reference Model, was chartered to deal with the two lower layers. Notably Layer 2, the Data Link Layer, was divided into two subgroups: Medium Access Control (MAC) and Logical Link Control (LLC) [2]. The initial development was based on Ethernet, the embellished version of which was finalized in December 1982 as 802.3, which is commonly referred to as Ethernet. The IEEE membership, the U.S. National Bureau of Standards, and the European Computer Manufacturers Association (ECMA) accepted that first release of three standards. The International Organization for Standardization (ISO) issued correlating international standards—known as 8802 LAN standards. Since then, the IEEE has continued to develop a broad range of LAN and MAN standards. IEEE standards include the following:

· 802.1: Architecture and internetworking (high-level interface). Defines architecture layers and rules for interconnection of disparate LAN protocols. Includes data formatting, network management, and internetworking.

· 802.2: Defines equivalent LLC services, including protocol for data transfer. Largely addresses bridges.

· 802.3: Defines CSMA/CD Access Method and Physical Layer Specifications. Commonly referred to as the Ethernet standard.

· 802.4: Token Bus Access Method and Physical Layer Specifications.

· 802.5: Token-Passing Ring Access Method and Physical Layer Specifications. Includes Token Ring.

· 802.11: Wireless Local Area Networks (WLANs). This is a family of standards describing the over-the-air interfaces for a number of RF-based WLANs.

· 802.12: 100+ Mbps LANs using demand priority access. The focus was on 100VG-AnyLAN, which standard is considered obsolete.

In addition to the IEEE, other standards bodies are involved in the establishment and promotion of certain LAN and computer networking standards. ANSI (American National Standards Institute), for example, developed the following standards:

· X3T9-3: HIgh Performance Parallel Interface (HIPPI)

· X3T9-5: Fiber-Distributed Data Interface (FDDI)

Table 8.3 compares Ethernet, Token Ring, and FDDI across a number of critical dimensions.

8.8 LIFE IN THE FAST LAN: THE NEED FOR SPEED

FDDI largely is deployed as a campus and sometimes a Metropolitan Area Network (MAN) backbone technology for the interconnection of major computing resources such as hubs, switches, routers, and servers. While FDDI can be extended to the device levels, the cost of optoelectric termination is high. The advantages of FDDI, however, can be extended to the workstation through a concentrator that accomplishes the optoelectric conversion process for multiple attached devices. The connection from the concentrator to the workstations is accomplished via UTP over distances of 100m or less, based on a standard known variously as Copper Distributed Data Interface (CDDI) and Twisted-Pair Distributed Data Interface (TPDDI).

The fragility of the fiber is a deterrent to the application of FDDI as well. The FDDI specifications provide for a dual counterrotating ring, which provides a measure of redundancy. Should the primary ring fail, a Dual Attached Station (DAS) or Dual Attached Concentrator (DAC) can communicate with any other device by transmitting in the opposite direction through the secondary ring, which typically is collocated in the same cable sheath as the primary ring (see Figure 8.18). If there is more than one physical failure in the cable plant, however, the ring segments and the network all fail. There are dual-homing solutions to this dilemma, although they involve considerable additional expense, with the designated stations connected via fiber to multiple servers to provide redundancy [35].

The standard for Gigabit Ethernet (GbE or GigE) was finalized and formally approved on June 29, 1998, as IEEE 802.3z. Although fully compatible with both 10-and 100-Mbps Ethernet, most equipment has to be upgraded (i.e., replaced) to support the higher transmission level. GbE addresses the bandwidth problem in 10/100 Mbps Ethernet networks, which began to feel the stress of bandwidth-intensive, multimedia-based Internet and intranet applications as well as scientific modeling and data warehousing and data backup.

GbE is available in shared and switched versions (see Figure 8.19), both of which support multiple ports that can run at 1 Gbps in full-duplex (FDX). Shared GbE essentially is a much higher speed version of 10Base-T and 100Base-T. Shared GbE is a high-speed hub that uses CSMA/CD for Medium Access Control over the shared bus. Therefore, you can characterize shared GbE as a brute-force attack on congestion. Switched GbE addresses the congestion problem through buffering incoming Ethernet frames and passing them to the output port when the shared bus becomes available. The shared bus can run at a speed of several Gbps. The more substantial switched GbE products offer nonblocking switching through a crossbar switching matrix, which may run at an aggregate of tens of Gbps [37]. The cost of a GbE switch is greater than that of a GbE hub and is sensitive to such factors as port density, buffer placement and capacity, switch matrix complexity, and throughput.

Clearly, the primary application for GbE in the near future largely will be in the backbone, for interconnecting lesser Ethernet hubs, Ethernet switches, and high-performance servers, rather than connecting individual nodes. However, GbE already is used occasionally in certain bandwidth-intensive desktop applications such as Computer-Aided Design (CAD), and GbE hubs and switches commonly offer port speeds of 10/100/1000 Mbps. Gigabit Ethernet supports HDX and FDX interfaces, with FDX offering the advantage of virtual elimination of issues of data collisions. The HDX and FDX declarations are made on a port-by-port basis.

Notably, GbE is not limited to the LAN domain. A number of service providers now offer GbE as a metropolitan area network service offering intended for multi-site enterprises confined to a metropolitan area. This service involves centrally positioning one or more GbE switches in the MAN and providing the enterprise with access via a variety of technologies. Although fiber optics clearly is the most attractive access technology from a performance standpoint, alternatives include unchannelized T-carrier, Very-high-data-rate Digital Subscriber Line (VDSL), and various wireless options. The real advantage to GbE in the MAN is that all traffic is carried in native Ethernet format, with no requirement for introducing SONET, Frame Relay, ATM, or other Layer 1 or Layer 2 protocols that can increase both complexity and cost.

· Infrared (IR) light systems currently require no licensing. The potential for interference between systems is very limited as LOS generally is required. IR is seldom employed in contemporary WLANs.

Note that IR generally requires LOS. RF systems generally do not require LOS but certainly benefit from it, particularly at the higher frequencies. Where RF signals must pass through walls, floors, ceilings, and windows, care must be taken with respect to the construction materials used. Metallic foil-backed insulation, for example, can have disastrous effects on the RF signal, as can certain glass windows with UltraViolet (UV) protection afforded by embedded metallic film. In advance of installing a WLAN, it is highly recommended that a site survey be conducted. A number of manufacturers offer highly sophisticated site survey systems that will ensure a satisfactory level of WLAN performance.

The IEEE 802 standards committee formed Working Group 11 to develop a set of specifications for over-the-air (i.e., wireless) LAN/MAN standards. The resulting family of standards, variously referred to in the vernacular as Wi-Fi and Wireless Ethernet, include IR and RF solutions, although there appear to be no practical applications for IR. The RF standards variously fall into the 2.4-and 5-GHz ISM bands and offer raw bandwidth up to 54 Mbps, at least theoretically. The original 802.11 standard, released in 1997, supported data rates theoretically up to 2 Mbps in the 2.4-GHz band. This early standard included a great number of options, which made interoperability of products of disparate origin difficult, or at least uncertain. As a result, 802.11 never gained any real traction in the market. Soon afterward, however, much improved extensions to 802.11 were finalized, and WLANs quickly gained in popularity.

Dubbed Wi-Fi5 (Wireless Fidelity 5 GHz) by the Wireless Ethernet Compatibility Alliance (WECA), now the Wi-Fi Alliance, 802.11a supports speeds up to 54 Mbps in a 300-MHz allocation in the 5-GHz range, which the FCC allocated in support of U-NII (Unlicensed-National Information Infrastructure). Specifically, 200 MHz is allocated in the band 5.15–5.35 MHz for in-building applications and 100 MHz in the band 5.725–5.825 MHz for outdoor use.

· QPSK (Quadrature Phase Shift Keying) at 250 kbps per channel for a total of 12 Mbps and 375 kbps per channel for a total of 18 Mbps

· 16QAM (16-level Quadrature Amplitude Modulation) at 500 kbps per channel for a total of 24 Mbps and 750 kbps per channel for a total of 36 Mbps

· 64QAM (64-level QAM) at 1 Mbps per channel for a total of 48 Mbps and 1.125 Mbps per channel for a total of 54 Mbps.

The data rates actually are raw theoretical signaling rates. The actual data rates are more on the order of 40–60 percent of the theoretical due to issues including overhead and medium access control. The symbol rate (i.e., the rate of transmission of a symbol, or set of bits) is slowed down enough that each symbol transmission is longer than the delay spread, that is, the variation in timing between receipt of the signals associated with a given symbol, with the delay spread being caused by mul-tipath propagation, which is the phenomenon by which the RF signals carrying a given data symbol arrive at the receiver at slightly different times as a result of their having taken slightly different paths as they bounced off of various physical obstructions.

While the 5-GHz spectrum is relatively clear in the United States, it is not nearly so readily available elsewhere as military and governments use portions of this band overseas. In Japan, only the 5.15–5.25-MHz spectrum is available. In Europe, the 5.725–5.825-MHz spectrum is already allocated for other uses, including HiperLAN, which competes with 802.11a. In any event, the 5-GHz spectrum is highly susceptible to attenuation [46, 47].

802.11b was the first of the 802.11 family to be released. Although the standards development process of 802.11a began first, technical and regulatory difficulties arose and development slowed. WLANs conforming to the IEEE 802.11b specification now are by far the most common, due not only to the early release of the standard but also to the signal propagation characteristics of the 2.4-GHz band. As a side note, 802.11b has been dubbed Wi-Fi (Wireless Fidelity) by the Wireless Ethernet Compatibility Alliance. The term Wi-Fi also has been attributed to the IEEE 802.11 Working Group, with Wi referring to the fact that a wire traditionally served as the physical medium for LANs and the homonym Fi referring to PHY, the PHYsical layer. So, Wireless PHY became Wi-Fi. Now, let's get back to work.

The FCC limits power output to 1 watt Equivalent Isotropically Radiated Power (EIRP). At this low power level, the physical distance between the transmitting devices becomes an issue, with error performance suffering as the distance increases. Therefore, the devices adapt to longer distances by using a less complex encoding technique and a resulting lower signaling speed, which translates into a lower data rate. A system running at 11 Mbps using CCK and QPSK, for example, might throttle back to 5.5 Mbps by halving the signaling rate as the distances increase beyond 30m or so and error performance drops. As the situation gets worse, it might throttle back to 2 Mbps using only QPSK and 1 Mbps using BPSK. At this lowest rate, link quality generally is acceptable at distances of up to 100m or so. 802.11b divides the available spectrum into 14 channels, each of which has a width of 25 MHz. In the United States, the FCC allows the use of 11 channels. Four channels are available in France, 13 in the rest of Europe, and only 1 in Japan. There also is overlap between adjacent channels as each has a width of 25 MHz and all share a band that is only 83 MHz (2.4–2.483 GHz) wide, which fact further affects performance. In the United States, for example, only 3 of the 11 available channels are nonoverlapping. Therefore, any given system must maintain maximum channel separation and physical separation from other systems in proximity.

Wi-Fi can operate in two modes. Ad hoc mode allows devices such as laptops to discover each other and communicate directly, without the involvement of an AP. This approach is certainly convenient for ad hoc communications, such as a spur-of-the-moment file transfer when out of range of an AP. Infrastructure mode requires that an Access Point (AP) be involved to support a connection. This approach is by far the most common, as it is truly a LAN mode of operation. Access points can be either fat or thin:

· Fat Access Point: A fat AP is intended to act independently and contains sufficient program logic and processing power to allow it to enforce policies relating to access and usage. Traditionally, multiple APs are hardwired to a switch that serves to interconnect the APs and to provide access to other internal and external resources.

· Thin Access Point: A thin AP is intended to act under the supervision of a centralized controller that configures, manages, and secures the environment. The centralized controller provides a single point of administration for all APs. This tends to be the preferred approach in contemporary Wi-Fi networks [48–50].

Speaking of power, all devices comprising a LAN, whether wired or wireless, require electrical power. That includes all clients (even those that can operate on battery power for short periods), all servers, and all APs. If the building was designed with plenty of electrical outlets in just the right places, power is no problem. Since LAN-attached clients now include security cameras, tablet and hand-held computers, and even telephones (more on that later), LAN connectivity and electrical power issues can reach not only into every room, every hallway, and every corner of every building but also into parking garages and outdoor areas. A network designed to provide full coverage at that level can involve considerable costs for electrical wiring, whether it is full-mesh wireless or not. That fact increasingly prompts reexamination of full-mesh Wi-Fi, particularly in consideration of the fact that it can be much less expensive to run Cat 5e LAN cables than to run electrical wiring.

PoE clearly solves a power problem where Cat 5e or Cat 6 is in place and properly installed and over which PoE-compliant endspan or midspan power injectors can connect to compliant client devices. It does, however, also pose a bit of a cabling problem, as it consumes two pairs of what is usually a four-pair cable. Usually Cat 5e cable is run in a four-pair configuration, with two pairs for data and one or two pairs often used for voice. Where PoE is deployed, there is the potential for something of a conflict between voice, data, and power as they all compete for limited pair count. There also can be issues with respect to the total electrical load a PoE system might place on electrical circuits and wiring closet cooling systems in a sce-nario involving large numbers of PDs.

That said, in November 2004, the IEEE formed the 802.3at Working Group to develop the next level, known as PoE Plus, with the goal of increasing power to 30 W or more. The standard will address Cat 5 or better cable and is expected to support 10Base-T, 100Base-T, 1000Base-T, and 10GBase-T. The 802.3at specification is expected in the 2007–2008 time frame [52–61].

Two security mechanisms are specified in 802.11b. The most basic is the Extended Service Set IDentifier (ESSID, or SSID), which is in the form of an identifier code established by the system administrator for each device set up to gain access through each access point. At the next level is Wired Equivalent Privacy (WEP), a stream cipher that uses a 40-or 128-bit encryption key to protect data in transit. After WEP was compromised by hackers in 2001, it largely was replaced by Wi-Fi Protected Access (WPA), based on the more secure Advanced Encryption Standard (AES), which employs a 128-bit block cipher. WPA is included in the 802.11i standard (2004) for Wi-Fi security.

802.11b has become incredibly popular, not only in enterprise, SOHO, and even personal applications but also in public settings. Wi-Fi hotspots increasingly are installed in public venues such as airports and coffee shops and even on airplanes, usually on a pay-per-minute basis. A number of municipalities have installed free Wi-Fi access in downtown areas to encourage people to frequent those areas and patronize local merchants, and a great many more municipalities plan to offer such services in the future. A number of Internet Service Providers (ISPs) quite naturally are battling to keep Wi-Fi from attaining the status of a public utility, but their success has been mixed, to say the least.

It was only a matter of time until someone came up with the idea of using a Wi-Fi network for voice communications. After all, Private Branch eXchange (PBX) technology has moved to VoIP over switched Ethernet, wireless voice is commonplace through cellular and cordless technology, Wi-Fi tech-nology offers plenty of bandwidth, and Wi-Fi mesh networks are possible through fat clients and high-speed layer 2 switches. It is not quite that simple, of course. Voice over Wi-Fi (VoWi-Fi) presents a number of technical challenges, including handoffs between APs as the user moves from cell to cell, Quality of Service (QoS), and security.

The IEEE addressed QoS issues in 802.11e (2005) through a new coordination function that provides a station with high-priority traffic such as voice with more frequent network access than a station with low-priority traffic such as e-mail. Further, the station with the high-priority traffic is granted a longer transmit opportunity, that is, time window, in which to transmit as many frames as possible. In all, 802.11e defines four access priority classes, which the Wi-Fi Alliance terms to be Wi-Fi MultiMedia Extensions (WMMs or WMEs). Those classes are as follows:

· Voice priority, the highest level, is defined in support of low-latency voice.

· Video priority, the second highest level, prioritizes video relative to other data traffic. One 802.11a/b channel can support three-to-four Standard-Definition TV (SDTV) or one High-Definition TV (HDTV) data streams.

· Best effort priority is intended to support traffic from legacy devices and from applications or devices that lack QoS capabilities. Web browsing is an example of best effort traffic.

QoS also requires smooth handoffs in order to avoid lapses in conversation and dropped calls as the user moves between cells. This is more than a matter of making a connection with one AP before dropping the connection with another. It is a matter of security as well. When a client seeks access to a Wi-Fi network, it does so through an AP that, either independently if a fat AP or under the direction of a switch if a thin A P, is responsible for authenticating the identity of the client and either granting access or denying it. This process takes a few seconds, which is not noticeable in a data application but is a huge QoS issue for voice calls. As Wi-Fi cells tend to be quite small at a maximum diameter of 200m or so, these handoffs can be frequent. To mitigate this issue, Wi-Fi switches assume control of large numbers of thin APs, coordinating and controlling their activities much as a Mobile Traffic Switching Office (MTSO) controls a number of dumb cell sites. As long as the user remains within the domain of a single switch, the handoffs can be handled fairly smoothly. If the user moves between switch domains in a large Wi-Fi environment, however, QoS issues develop as the switches struggle to coordinate handoffs. There is an even more complex handoff issue if a user moves from a private VoWi-Fi domain to a public one. The IEEE is addressing these problems through the 802.11r initiative, which it expects to finalize in early 2007. Also, the IEEE 802.21 Task Group is in the early stages of developing specifications to support interoperability and handoff issues between heterogeneous networks, including 802 and non-802 network types. The expectation is that this effort will result in specifications supporting smooth interconnectivity and interoperability between Wi-Fi and cordless telephony and cellular, WiMAX, and other wireless networks. Some dual-mode and even trimode handsets already exist. For example, several ILECs in Africa that also are cellular service providers offer dual-mode phones that support VoWi-Fi and cellular. Since a single service provider owns both the PSTN and cellular networks and both installs and maintains the Wi-Fi LAN, interconnectivity can be achieved more easily than if multiple network operators were involved.

The physical and logical layout of the Wi-Fi network has a real impact on its ability to support voice. A Wi-Fi network in support of laptop users is relatively simple to configure. As computing on such a platform tends to require a flat and stable surface, one looks for places with tables or at least chairs or benches for users with fairly flat laps. Configuring a network for users of tablet and hand-held computers is more challenging, as they can compute in hallways and other unusual places and while on the move. Not only do the coverage areas expand, but cells must overlap, handoffs must be made, and frequency assignments must be carefully administered so that the same channels are not used in adjacent cells. As 802.11b/g offers only 11 channels in the United States, spectrum management can be quite a challenge, particularly where user density is high and, therefore, APs must be tightly spaced and cell sizes must be small. While one obvious solution is that of reducing the power levels of the APs, VoWi-Fi network configuration issues can be tough to solve.

HiperLAN (High performance radio LAN) is a high-speed LAN standard running in the 5-GHz range. Approved by the European Telecommunications Standards Institute (ETSI) in February 2000, HiperLAN grew out of efforts to develop a wireless version of ATM and a European alternative to 802.11. HiperLAN1 operates at rates up to 20 Mbps and HiperLAN2 at rates up to 54 Mbps. As noted in the discussion of 802.11a, in Europe the 5.725–5.825-MHz spectrum is already allocated for HiperLAN. Therefore, ETSI requires that two additional protocols be used in conjunction with 802.11a in order to protect incumbent applications and systems running over previously allocated shared spectrum. Dynamic Frequency Selection (DFS) allows the 802.11a system to dynamically shift frequency channels, and Transmission Power Control (TPC) reduces the power level. In combination, these protocols serve to eliminate interference issues with incumbent signals. HiperLAN uses Orthogonal Frequency Division Multiplexing (OFDM) as the signal modulation technique [46].

HiperLAN has a very small following, as it has been overwhelmed by the popularity of the 802.11 family. HiperLAN is unlikely to survive the next few years as the pressure increases with the release of 802.11n.

In 2006, Exmovere unveiled a Web-based Bluetooth-enabled biosensor wrist-watch designed to provide elderly care assistance. The wearable Exmocare sensor system monitors the elderly person's vital signs, including pulse, heart rate, and motion, and allegedly can assess up to 10 different emotional states, including relaxed, worried, and agitated. The device collects information every 30 min and automatically alerts caregivers of any abnormal status. The wristwatch connects via Bluetooth to an enabled PC, cell phone, or GPS-car kit, which monitors the wearer's location and vehicle speed.

Bluetooth makes use of Frequency Hopping Spread Spectrum (FHSS), with devices employing a pseudorandom hop sequence that makes data collisions highly unlikely. Bluetooth operates in the 2.45-GHz range of the ISM band, hopping through a set of 79 (United States and Europe) or 23 (Spain, France, and Japan) channels spaced 1 MHz apart at a rate of about 1600 hops per second, with each hop lasting 62.5μs. At each hop, the baseband signal is modulated using Gaussian Frequency Shift Keying (GFSK), a technique that smoothes out the signals and limits their spectral width in order to reduce the potential for crosstalk in tightly packed radio channels. Gaussian Phase Shift Keying (GPSK) involves a positive frequency shift to represent a 1 bit and a negative frequency shift to represent a 0 bit. GPSK is not highly efficient, yielding a maximum raw link speed of 1 Mbps; however, its simplicity allows for low-cost device implementations. Depending on the transmitted power, Bluetooth devices can be organized into three classes. Class 1 equipment (maximum 100mW) has a nominal link range of as much as 100m, and Class 3 devices (maximum 1 mW) are limited to about 10m [71].

Bluetooth technology supports both Synchronous Connection–Oriented (SCO) links for packet voice and Asynchronous Connectionless Links (ACLs) for packet data. Bluetooth supports an asynchronous data channel in asymmetric mode of up to 721 kbps in either direction and 57.6 kbps in the reverse direction. Alternatively, the data channel can be supported in symmetric mode of up to 432.6 kbps. As yet another alternative, Bluetooth supports up to three simultaneous synchronous packet voice channels, or a channel that simultaneously supports both asynchronous data and synchronous voice.

Bluetooth supports FDX communications using Time Division Duplex (TDD) as the access technique. Voice coding is accomplished using the Continuously Variable Slope Delta (CVSD) modulation technique. CVSD yields voice compression of 4: 1 (16 kbps) or 8: 1 (8 kbps).

As mentioned above, each slot in a Bluetooth connection lasts 62.5 μs and corresponds to a single hop from once device to another. The master always transmits in even-numbered slots and the slave in odd-numbered slots, with one packet transmitted per slot. As illustrated in Figure 8.23, the Bluetooth packet takes the following format:

· Access Code: The 72-bit access code is defined by the master and is unique for each piconet. The access code serves to identify incoming packets associated with a given piconet. Devices on the piconet will accept packets with the proper access code and will reject all others. This field also allows synchronization purposes.

· Medium Access Control (MAC) address: 3 bits, with 000 identifying a broadcast packet. Note: As 23 = 8, a piconet is limited to one master and seven slaves.

· Packet type: Four bits indicating the 16 packet types. There are 4 types of control packets and 12 types of data packets. There are 3 types of voice packets, each running at a rate of 64 kbps, including overhead and, in some cases, an FEC mechanism.

· Flow control: One bit.

· Automatic Repeat reQuest (ARQ): One bit. This bit is used to request a repeat packet transmission in the event that there is a detected error in the payload. ARQ is not used in SCO links, as error control would increase latency and retransmissions would create jitter.

· Sequence number: One bit.

· Header Error Correction (HEC): Eight bits.

· Payload: The variable-length packet is limited to 366 bits, although the theoretical limit is 625 bits (62.5 μs × 1 Mbps = 625 bits). The limit of 366 provides the transmitters and receivers with enough time to hop to the next frequency and stabilize. As the access code and packet header consume 126 bits, the payload cannot exceed 240 bits, or 30 octets. There is a provision for multislot packets, which can support a larger payload. The payload includes an error control mechanism in the form of a CRC [71].

ZigBee is a specification from the ZigBee Alliance for a set of high-level communications protocols based on the IEEE 802.15.4 standard for Personal Area Networks (PANs). ZigBee is intended to be simpler, more flexible, and less expensive than either Bluetooth, which also is based on 802.15, or Wi-Fi. ZigBee is designed for connecting devices in ad hoc mesh networks over very short distances with very low power consumption. There are three network topologies possible—star, peer to peer, and mesh. The high levels of redundancy and network resiliency offered by mesh networking make it the preferred approach in consideration of the applications, which include building automation and industrial, medical, and residential monitoring and control. Should a device fail or be removed, the autodiscovery feature of such a mesh network will recognize and register that fact and exercise an alternate path.

ZigBee is hierarchical in nature, including ZigBee End Devices (ZEDs), which are terminal devices, such as sensors, that can perform only a single monitoring or control function. ZEDs communicate with ZigBee Routers (ZRs), which can serve as ZEDs in addition to functioning as routers to pass messages to other ZRs or to the ZigBee Coordinator (ZC). The ZC initializes the network, coordinates its operation, and is responsible for security.

A ZED comprises a low-cost microprocessor, RAM and ROM memory, a battery, a radio and controller, and the IEEE and ZigBee protocol stacks. The device is small enough to be embedded in a light switch, smoke or carbon dioxide detector, thermostat, security sensor, utility meter, or medical sensor. As battery life is so critical, a ZED goes into sleep mode when not actively performing its functions. Therefore, a battery could last for years if used in a low-duty-cycle application such as meter reading or alarming [75, 76].

The term ZigBee refers to the technique, known as the ZigBee principle, that a domestic honeybee uses to communicate the location of a new food source to other members of the colony. The bee dances in a zigzag pattern that communicates information such as distance and direction, at least according to the ZigBee Alliance [76]. Others suggest that it is just a made-up name chosen at random, since independent research, including my own, fails to confirm that ZigBee has anything to do with honeybees. (That's a shame, because it's a really cool story.)

The FCC has defined Software-Defined Radio (SDR) as a generation of radio equipment that can be reprogrammed quickly to transmit and receive on any frequency within a wide range of frequencies and using virtually any transmission format and any set of standards. Theoretically, a device such as a WLAN NIC, cellular telephone, or PDA with an SDR chipset could seek out various available frequency bands and native protocols supported by the networks, lock in on the signals, and negotiate access to the desired network, downloading any necessary supplemental software required to effect network compatibility. In the process, SDR-equipped devices would resolve any conflicts between networks sharing a given band (e.g., 802.11b and Bluetooth overlap in the 2.4-GHz ISM band). The FCC began hearings on SDR in March 2000, with the intent that the development of SDR could promote more efficient use of spectrum, expand access to broadband wireless communications, and increase competition among service providers.

8.10 MINDING YOUR Ps AND Qs

Ps and Qs are the mechanisms by which layer 2 switches support voice and video as well as LAN traffic in the converged LAN domain. Within the customer premises, these specifications encourage the development of the convergence scenario via such relatively inexpensive technologies as switched Ethernet.

8.11 IEEE 1394 AND FIREWIRE

The IEEE 1394 specification, known as FireWire in Apple Computer terminology, is for a data transport bus between a host computer and peripherals (e.g., high-density storage devices and high-resolution still and video cameras) and is designed to eliminate the bottleneck at the serial port of the LAN-attached PC. While 100-Mbps LANs address the bottleneck of the shared medium, they can overwhelm the attached workstation (sort of like drinking out of a fire hose). As videoconferencing and multimedia applications increase in popularity, it is necessary to increase the speed of the Small Computer Systems Interface (SCSI) to support them and take full advantage of the speed provided by the LAN. 1394 addresses that requirement through standards for 100, 200, and 400 Mbps. A single 1394 port can support up to 63 peripherals over a six-conductor cable up to 4.5m in length and as many as 16 cables can be daisy chained to extend the total length to as much as 72 m.

8.12 NONSTANDARD LANs

In addition to standard LANs, there exist a number of nonstandard options. Some are proprietary standards that serve certain applications and vertical markets, such as the oil refining industry described earlier. Others are more widely accepted speci-fications, published by other than a recognized standards body, with ARCnet (Attached Resource Computer net work) being a prime example. Developed in 1977 by Datapoint Corporation, ARCnet is a highly reliable, low-cost token bus system based on a physical star and logical ring and a character-oriented protocol. ARCnet supports data rates up to 2.5 Mbps and as many as 255 attached devices. More recent versions deliver 20 and 100 Mbps, although they have never been in great demand and are not widely available. ARCnet resembles but does not adhere to the IEEE 802.4 specification. RXnet was Novell's implementation of ARCnet.

8.13 BROADBAND OVER POWER LINE

Broadband over Power Line (BPL) is a set of specifications for Power Line Carrier (PLC), a technology that has been used for certain telco local loop applications since 1928. In-house PLC technologies have been used in key telephone and intercom systems since at least the early 1980s, although not particularly successfully. Standards for In-house BPL, a premises networking technology, are a relatively recent development, with HomePlug standards being the most prevalent. Loosely based on Ethernet LAN standards and using a variation of CSMA/CA, In-house BPL allows any device to connect to the LAN directly through the low-voltage electric lines (110 V at 50–60 Hz or 220 V at 50 Hz). HomePlug 1.0 supports up to 16 nodes sharing bandwidth up to a theoretical maximum of 14 Mbps. Some proprietary systems support raw signaling rates up to 85 Mbps, which comes very close to 100Base-T performance.

HomePlug-compatible devices include PCs, routers, bridges, switches, and any other devices that use RF-45 (Ethernet) or USB physical interfaces. The devices plug into a HomePlug adapter that is about the size of a typical low-voltage transformer or power adapter and that plugs into any electrical outlet on the premises. Thereby, every electrical outlet becomes a port into an Ethernet LAN.

HomePlug uses of a version of Orthogonal Frequency Division Multiplexing (OFDM) specially tailored for powerline environments. OFDM splits the signal into a stream of data symbols for massively parallel simultaneous transmission over a number of narrowband, low-data-rate subcarrier frequencies. [Note: OFDM is the transmission technique used in 802.11a (aka Wi-Fi5) and 802.11g, 802.16, and WiMAX wireless systems. Also known as Discrete MultiTone (DMT), the technique is used in ADSL service as well.] HomePlug 1.0 specifies 84 equally spaced subcar-riers within each of which several differential modulation techniques are employed. Security is through 56-bit DES.

Attenuation in HomePlug networks is influenced not only by the propagation of the signal through the copper conductors (commonly 12 or 14 gauge) but also by splices and various components such as fuse boxes, surge suppressors, and circuit breakers. HomePlug currently offers a range of as much as 300m without repeaters, which compares favorably with the 100m supported by 10/100BaseT.

The next step in the specification is HomePlug AV, which is intended to support entertainment applications such as HDTV and home theater. HomePlug AV will run in the range of 2–28 MHz, offering a raw signaling speed of up to 200 Mbps through the use of OFDM. Throughput will be more in the range of 100 Mbps, given TCP/IP and other overhead considerations. HomePlug AV will offer inherent Grade of Service (GoS) considerations through the use of IEEE 802.1Q Virtual Local Area Network (VLAN) tags for marking high-priority traffic such as VoIP (Voice over Internet Protocol) and the streaming audio and video components of HDTV. With regard to security, HomePlug AV makes use of the highly secure AES. Note: The IEEE is working on the competing P1901 specification for BPL communications, which will address both Access BPL and In-House BPL [80–85].

8.14 STORAGE AREA NETWORKS

As electronic commerce (e-commerce) has exploded along with the growth of the Web, so has the requirement for storage. Other applications also are increasingly storage intensive, and it has become clear over the last few years that it is absolutely necessary to maintain highly secure database backups. A number of network storage technologies, which can be quite complex, have developed to satisfy this requirement. The simplest approach is that of Network-Attached Storage (NAS), which simply is one or more storage devices (e.g., disk arrays) associated with a server that exists as a node on a LAN. The storage server assumes the responsibility for all data storage and for making the data available to all users on the network who have appropriate access privileges.

The InterNational Committee for Information Technology Standards (INCITS) began work on Fibre Channel (FC) in 1988 as a replacement for the HIgh Performance Parallel Interface (HIPPI) technology, a highly distance-limited technology that involved 50-pair cable and huge connectors. ANSI approved the resulting specification in 1994. FC is connected at Layer 1, the Physical Layer, by fibre, a term the Fibre Channel industry coined to refer to a network comprising a close-knit fabric of access including both optical fiber and copper for large data transfers with low overhead, low-latency switching, and minimal interruptions to the flow of data [92]. The physical media and transceivers are the same as those used in LANs and telecommunications networks. The physical medium of choice is optical fiber, which can be MultiMode Fiber (MMF) of either 62.5 km (300 m, maximum distance) or 50 km (500 m, maximum distance) or Single-Mode Fiber (SMF) (50+ km, maximum distance). The recently developed Fibre Channel over IP (FC/IP) technology extends FC to operate through secure tunnels over public IP networks at WAN distances. FC operates at four link speeds (and FDX throughput rates, as measured in bytes per second) as follows: 1 Gbps (200MBps), 2Gbps (400MBps), 4Gbps (800MBps), and 10 Gbps (2400 MBps). The line coding technique is 8B/10B, which now also is used in GbE and 10GbE. FC equipment comprises hubs, switching hubs, switches, and routers. Gateway routers are responsible for protocol conversion to support interconnection to telecom networks such as ATM and SONET as well as SCSI SANs and Ethernet LANs.