Paper 2 370

John Wiley & Sons © 2007 (832 pages) Citation

ISBN:9780470396070

Watch video

https://www.youtube.com/watch?v=sHu5ZDxaMaI&index=10&list=PLTr3GyMgrbb-7MZOfNmiYA7vr66JawWg

Chapter 9: Broadband Network Infrastructure

Poet Giovanni Pascoli, describing his views of the telegraph, 1911, translated from Latin

We appear to have developed a seemingly insatiable appetite for bandwidth in the Wide Area Network (WAN) and Metropolitan Area Network (MAN) as well as the Local Area Network (LAN). A number of broadband network technologies have been developed to address this need for speed. Some of those technologies are specifically oriented toward data applications, and others support the triple play, a marketing term derived from American baseball to describe a full range of voice, data, and video services. (Note: The term quadruple play is used by some service providers who add the wireless component of cellular communications. At this point, the baseball analogy breaks down, however.) Some of these are access technologies that are specific to the local loop while others are transport technologies positioned in the core, or backbone.

In the network core, or backbone, the levels of bandwidth are much greater and DS-1 has little significance. In this context, broadband is more correctly defined as DS − 3 or better and specifically T3 or better. The reason I choose this level is that the T3 rate of 45 Mbps is the foundation for the Synchronous Optical NETwork (SONET), the North American fiber–optic transmission standard, and Synchronous Digital Hierarchy (SDH), the international version of SONET.

Ultimately, of course, the focus has to be on the delivery of end–user services, without which any technology in the core would be totally irrelevant. Manufacturers and carriers are incredibly intent on providing attractive broadband products and services to end users. Governments, in general, and regulatory agencies, in specific, increasingly are centered on ensuring that as many end users as possible have access to broadband services at affordable prices. The FCC, for example, has as a broadband objective, which it has stated as follows [2]:

All Americans should have affordable access to robust and reliable broadband products and services. Regulatory policies must promote technological neutrality, competition, investment, and innovation to ensure that broadband service providers have sufficient incentive to develop and offer such products and services.

· DSL: 61.2 percent

· Cable modem: 32.0 percent

· Other technologies: 6.8 percent [3]

The FCC estimated that, as of the end of 2004, there were 37.9 million high–speed access subscribers in the United States. Of that total, there were 13.8 million DSLL installed, compared to 21.4 million cable modem connections. An additional 2.7 million high–speed connections were provided over satellite, terrestrial wireless, fiber, powerline carrier, or other wireline technologies.

Broadband access and transport technologies, according to the FCC, encompass all evolving high–speed digital technologies that provide consumers integrated access to voice, high–speed data, video–on–demand, and interactive delivery services [2]. This chapter deals with a number of access, or local loop, technologies, including Digital Subscriber Line (DSL), Community Antenna TeleVision (CATV), Passive Optical Network (PON), 802.16 and Worldwide interoperability for Microwave Access (WiMAX), and Access Broadband over Power Line (BPL). This chapter also details SONET and SDH, the standards for optical fiber transmission in backbone carrier networks, Dense Wavelength Division Multiplexing (DWDM), Coarse Wavelength Division Multiplexing (CWDM), and Resilient Packet Ring (RPR).

9.1 ACCESS TECHNOLOGIES

Some of the technologies are incremental in nature. In other words, they are enhancements of existing copper local loop [i.e., Unshielded Twisted–Pair (UTP) or coax] technologies. These incremental approaches substantially improve the performance of the underlying transmission media within the limitations imposed by the regulators and the basic laws of physics. Wireless local loops either overlay the traditional copper network, bypassing it for a variety of reasons that include cost and performance, or are an alternative where wireline loops are either impossible or problematic to deploy. Optical fiber loops generally involve either a complete replacement of trunk facilities in the loop or an overlay. Regardless of whether the access technology is characterized as an enhancement, an overlay, or a replacement, the absolute cost of its deployment is significant.

The cost of UTP local loops in North America ranges from approximately $ 1000 to $ 7500, with the upper end of the range applying where terrain is especially diffi-cult (e.g., rocky soil or high water tables) or where population density is low. The cost of an optical fiber loop is much higher, but they are deployed in support of high–volume business in high–density areas. As telco local loop revenues generally are low and costs are high, there really is no stunning profit potential in the telco local loop business. Rather, the profits are in the capabilities of the networks and services to which they provide access and the company that controls the local loop access to those networks and services. Assuming that the environment is a competitive one, a Competitive Local Exchange Carrier (CLEC) or service provider has two basic choices—buy or build:

· Buy: In order to encourage competition, the regulator may require that the ILEC lease local loops and other network elements to the CLEC, most likely at some rate that is cost based. The ILEC is unlikely to be in agreement with this requirement and almost certainly will not be especially enthusiastic or helpful in its implementation. However, leasing embedded telco local loops certainly is the fastest way to gain access to large numbers of customers.

· Build: Although it is expensive and time consuming to build local loop facilities, the long–term advantages are significant. A local loop built from the ground up, so to speak, likely will be based on the most advanced and appropriate technologies and, therefore, will be optimized to deliver the highest performance at the lowest cost. The CLEC that builds its own network also owns and controls it, which is very significant.

Generic Digital Subscriber Line (xDSL) encompasses a group of digital network access technologies largely intended for the consumer class markets, including residential, Small Office/Home Office (SOHO), and small–business applications. However, several very significant DSL services are oriented toward business class markets. DSL is provisioned over conventional UTP local loop facilities, which comprise the vast majority of network access facilities owned by the ILECs. Beginning in 1880 and continuing through much of the twentieth century, UTP was the only option, or the only viable option, for most local loop applications. UTP inher-ently is the most limited of all transmission media in terms of the dimensions of bandwidth, error performance, and distance, but a two–wire UTP circuit is quite sufficient in support of analog voice and analog voice–grade data through modems. If properly conditioned, it also is quite sufficient in support of ISDN BRI, a digital service running at a total signaling rate of 144 Mbps. UTP even supports Fractional T1, T1, E–1, and ISDN PRI at signaling speeds up to 2.048 Mbps. The main reason that we continue to deploy UTP is that it is relatively inexpensive and simple to acquire and install as an extension of a legacy UTP–based network.

However, the embedded twisted–pair local loop has become strained under the pressure of increased demand for increased speed of information transfer and improved error performance in support of data communications. In large part, this demand is due to the development of the Web and the resulting growth in popularity of the commercialized Internet. As a result of these factors and competition from CATV providers and others, the ILECs have aggressively sought to develop technologies to extend the life of the embedded UTP local loop through improving its general performance characteristics.

All xDSL technologies are fairly demanding of the local loop. Some are absolutely unforgiving, and some are adaptive to local loop anomalies. Before examining the specific technologies, I want to examine the problems with electrically based local loops, each of which has definite impact on xDSL performance:

· Splices and mixed gauges can cause echo, or signal reflection, which is a source of noise, or signal distortion. It is not at all uncommon to find mixed 24–and 26–gauge cable in a given local loop. It is not as though the ILEC craftspeople just closed their eyes and reached into the warehouse for a reel of cable, but the effect is the same as if they had. Telcordia Technologies estimates that the average residential local loop involves 22 splices, each of which contributes to the problem [6]. Some of those splices may not be mechanically sound, and some of the splice casings may not be well insulated. The problem is likely to be worse with longer loops, as they generally involve more splices. Echo cancellers and error detection and correction algorithms can compensate for much of this noise, but it is always an issue.

· Bridged taps are sections of a cable pair not on the direct electrical path between the CO and the customer premises. They generally are in place because a previous customer between the CO and the subject building was connected to the CO through the same pair. When that customer disconnected service, the splice was not removed. If a telephone is connected, the bridged tap essentially splits the signal, which roughly halves the signal power to the intended customer. If there is no telephone connected to the bridged tap, the open cable acts as a signal reflector, which causes signal distortion. DSL employs high-frequency signals, which attenuate quickly and require a point–to–point local loop for maximum performance. Bridged taps must be removed for some xDSL services to work optimally and for others to work at all.

· Load coils are passive devices that filter frequencies above 4 kHz in order to tune the copper circuit and optimize it for mid–voice–band performance. As DSL circuits employ frequencies well above 4 kHz, load coils must be removed.

· Electronics such as amplifiers, repeaters, and various types of simple multiplexers and line splitters are installed on some loops. They must be removed. DSL requires a dry copper pair (i.e., a pair with no electronics), also known as a burglar alarm circuit.

· Voice instrument signaling states can impact data communications in a DSL environment. While analog voice circuits are standardized and voice terminals are certified, the wide variety of devices vary enough in their electrical characteristics to cause unpredictable problems when integrating voice and data over a common inside wiring plan. For that matter, the electrical characteristics of multiple phones of the same manufacturer and model number can vary considerably, even within the same manufacturing lot. DSL data communications sessions can be dropped when analog telephones ring (voltages can exceed 100 V), when they go off–hook, and when they go back on–hook. The ultimate solution is an adaptive splitter or modem. The interim solution, at least for some versions of DSL, involves equipping all telephone sets, fax machines, and other connected devices with inexpensive high–pass microfilters that plug in between the set and the jack.

Also, before exploring the specifics of the DSL technology options, it is important to pause just a moment to examine some of the network elements that generally are involved. Those elements include splitters or modems, local loops, and DSLAMs:

· Splitters or modems are interface units that must be installed in matching pairs, with one at the customer premises and one at the edge of the network. The splitter at the network edge generally is embedded in a DSLAM. The premises-based unit may be a centralized splitter in the form of an enhanced network interface unit, which is the point of termination and demarcation between the carrier–provided local loop and the inside wire and cable system. A centralized NIU generally is attached to the outside of the building or is positioned just inside the basement wall, garage wall, or other point of entry. Alternatively, the splitter may be in the form of a set-top box, much like a CATV converter. The splitter functions as a hybrid frequency division multiplexer and time division multiplexer. Additionally, the splitter compresses the data transmissions. The splitter also may function as a codec, digitizing analog voice and fax transmissions as required.

· DSL access multiplexers provide the interface between the DSL local loop and the service provider's Point Of Presence (POP). Most DSLAMs are based on Asynchronous Transfer Mode (ATM), although some earlier models are based on Frame Relay (FR), with the specific technology depending on what the carrier has in place. In any case, the DSLAM and the DSL modem on the user's premises must run matching protocols. In an ADSL scenario, for example, the DSLAM receives upstream DSL traffic and splits, or demultiplexes, the voice and data traffic. The voice traffic then is encoded into PCM format and time division multiplexed over a channelized T1, T3, or, perhaps, SONET link to the PSTN. The data traffic is multiplexed or concentrated in ATM cells over an unchannelized T1, T3, or SONET circuit directly to the Internet backbone or perhaps to an independent Internet Service Provider (ISP). The DSLAM generally represents the first potential point of contention and congestion that affects upstream end–user traffic, as the local loop is a dedicated circuit.

· Single-Point-Of-Termination (SPOT) frames typically are used where a CLEC or ISP is leasing dry copper pairs from the ILEC for purposes of provisioning xDSL data services. The SPOT frame, and other hardware, is collocated in the ILEC CO, generally in separately secured leased space. The ILEC cross-connects the individual leased circuits at the Main Distribution Frame (MDF) and terminates them in the SPOT frame, where the CLEC or ISP connects them to the DSLAM.

The xDSL technologies discussed first are those intended largely for the residential, SOHO, and small–business markets: ADSL, IDSL, G.lite, and VDSL. The technologies discussed last are those intended primarily for the larger business market: HDSL, HDSL2, SDSL, and SHDSL.

The term asymmetric refers to the fact that ADSL bandwidth is provided on an asymmetric (i.e., not symmetric, or unequal) basis, with more bandwidth in the downstream direction than in the upstream direction. This asymmetry is in recognition of the FEXT and NEXT crosstalk issues discussed above and generally suits the applications.

In addition to supporting POTS voice over a separate analog channel running at 4 kHz and below, ADSL supports high–speed data. The upstream data ride in what technically is a bidirectional channel provided in increments of 64 kbps up to 640 kbps in a frequency band from 26 to 140 kHz. Downstream transmission is in increments of 1.536 Mbps up to 6.144 Mbps, based on T1 specifications, in a frequency band that runs from approximately 140–552 kHz. While ADSL is primarily a Layer 1 (Physical Layer) specification, it includes Layer 2 (Data Link Layer) elements. ADSL specifies three modes of operation:

· Synchronous Transfer Mode (STM), which is a bit synchronous mode. In this mode, the ADSL loop is a bit pipe.

· Asynchronous Transfer Mode (ATM).

· Packet Mode is IP–centric, employing both Ethernet and a variation of the High–level Data Link Control (HDLC) framing convention at Layer 2.

ADSL makes use of Discrete MultiTone (DMT) modulation, which gives rise to yet another moniker, G.dmt. Also known as Orthogonal Frequency Division Multiplexing (OFDM), DMT splits the signal into a stream of data symbols for massively parallel simultaneous transmission over 256 narrowband, low–data–rate subcarrier frequencies, each of which is 4 kHz wide. Within each subcarrier, Quadrature Amplitude Modulation (QAM) is employed to increase the efficiency of transmission. The DMT modem is dynamically rate adaptive, automatically adjusting the QAM constellation from 1 to 15 bits per baud, depending on the subchannel capacity and the line conditions. Under optimal conditions, therefore, a 4–kHz channel can support a data rate up to 60 kbps (4 kHz × 15–bit QAM = 60 kbps). The ADSL modems can detect any impaired subcarriers and shift data transmission away from them and to the unimpaired subcarriers. Note: Variations of the DMT/OFDM technique are used in 802.11 a (aka Wi-Fi5) and 802.11 g, 802.16, and WiMAX wireless systems, as well as Broadband over Power Line (BPL). The OFDM signal is the sum of a number of orthogonal subcarriers. (Note: Orthogonal is defined mathematically as at right angles to or perpendicular to, which in RF terminology means independent and mutually exclusive to the extent that a receiver can recognize and reject an unwanted signal or signal element.)

ADSL is a broadband local loop technology for broadband access to a broadband WAN. The nature of the WAN is not at issue, although interfaces to the circuit-switched PSTN, the IP–based Internet, and other public networks are supported. The TV signals generally travel via broadcast satellite to a head end, which could be in the form of a tandem office, where they are distributed to the lower level COs by microwave or fiber optics. The programming may be stored on tape or disk in a video server at the tandem or CO in support of delayed broadcasting or Video On Demand (VOD). The voice input is handled at each end of the connection through a frequency splitter that ensures it is unaffected by the presence or absence of digital data. At the network side of the loop, the video and data inputs are time division multiplexed into an ADSL signal that originates in an ADSL Transmission Unit—Centralized (ATU-C). The signal then travels the local loop cable plant, terminating in a matching ADSL Transmission Unit—Remote (ATU–R) located on the customer premises, as illustrated in Figure 9.3. The ATU–R acts as a modem, splitting the total bandwidth into three channels: the downstream channel, the bidirectional channel, and a voice channel. The ATU–R can be in the form of a stand–alone unit or a printed circuit board in a PC or other host computer. Alternatively, the ATU–R can be contained within a set–top box in a video or TV application. The Bit Error Rate (BER) of ADSL is 10 -9, which compares favorably even with traditional DS–1 at 10 -7.

The primary interest in ADSL has resided with the incumbent telcos, as it runs over an embedded twisted–pair cable plant. Where a competitive environment exists, CLECs may be allowed to lease the telco loop and terminate it in their collocated equipment. ILECs and CLECs, alike, find ADSL effective in countering competition from CATV companies for the residential voice and data market. As the CATV providers increasingly compete for that business, they are upgrading existing coaxial cable plant to add bidirectional, switched voice, and data capabilities. In the face of that pressure, the telcos have to consider a method of increasing the bandwidth in the local loop to compete effectively. Their options include laying new cable plant (fiber optic or coax), making use of wireless technology, or increasing the capacity of the existing twisted–pair plant. ADSL offers a relatively inexpensive solution, accommodating entertainment TV, videoconferencing, and higher speed data transfer over an existing cable plant of varying gauges. Many telcos in the United States and other countries with highly developed and competitive telecommunications markets view ADSL as a short–term access solution, filling the bandwidth gap until more substantial fiber–optic cable plant is deployed, usually in the form of Passive Optical Network (PON).

Some ILECs (e.g., SBC in the United States) are addressing this issue of video performance over copper loops through a hybrid fiber/copper approach known variously as Fiber-To-The-Neighborhood (FTTN) and Fiber-To-The-Curb (FTTC). FTTN involves one or more high–speed fiber–optic links from the network edge at the CO to an ngDLC that supports both fiber and copper interfaces, as illustrated in Figure 9.4. From the ngDLC to the customer premises, the embedded UTP supports full–rate ADSL. Since this approach reduces the copper portion of the local loop from 12,000–18,000 ft to a few thousand or perhaps a few hundred feet, its performance is improved significantly and it is much more likely to support the full range of voice, data, and video signals. Other ILECs (e.g., Verizon in the United States) consider DSL to be a short–term solution to a long–term problem that they are addressing through PON technology. PON takes fiber optics directly to the customer premises, completely replacing the embedded twisted–pair local loop. Full-rate ADSL development continued well past the introduction of the initial version. Since that time, ADSL2 and ADSL2+ have been released.

ADSL2 supports bonding in ATM, based on the ATM Forum specification for Inverse Multiplexing over ATM (IMA). This allows two ADSL pairs to be bonded together to yield roughly double the single–pair rate across the full range of loop lengths. Bonding has increasing application, as competition from CATV providers and cellular services have led to the disconnection of large numbers of telco-provided copper landlines. The deployment of pair gain technologies such as ADSL and the increasing use of fiber optics have combined to further reduce the pair counts required to support primary and secondary lines and trunks and have added to the excess of available copper loops.

ADSL2 further supports channelization, which provides the ability to split the bandwidth into different channels with different link characteristics in support of different applications. For example, a single ADSL2 loop might support a business class videoconferencing application, which is intolerant of both latency and bit errors, while simultaneously supporting e-mail transfers, which are quite tolerant of both latency and bit errors. Channelized Voice over DSL (CVoDSL) allows multiple derived lines of TDM voice traffic to be transported simultaneously over DSL in 64-kbps channels [9].

The ITU–T reached agreement on G.992.5 in January 2003, adding ADSL2 + to the ADSL suite. This specification doubles the downstream data rate to as much as 24.5 Mbps over shorter loops up to approximately 5000 ft (1500 m) in length. The upstream rate remains at a maximum of 1 Mbps. In order to achieve this enhanced data rate, ADSL2+ increases the downstream frequency range from 552 kHz (ADSL) and 1.1 MHz (ADSL2) to 2.2 MHz (ADSL2+) and increases the number of subcarriers to 512. The analog POTS channel remains at 4 kHz (ADSL and ADSL2), and the upstream data channel remains capped at 140 kHz (ADSL2). Table 9.3 provides a quick comparison of the frequency ranges of the various ADSL technologies.

Recent developments in ADSL+ CPE support multiple 10/100 Mbps Ethernet interfaces and optional 802.11 b/g interfaces. Combining these enhancements with the increased bandwidth of ADSL2+ and its channelization capability allows the user organization to derive multiple VLANs from a single ADSL2+ facility, with each VLAN enjoying its own Quality–of–Service (QoS) level. Thereby, separate VLANs can be created for low–latency voice and video as well as latency–tolerant data communications applications in a triple–play scenario. ADSL2 and ADSL2+ certainly relieve the pressure on triple–play bandwidth. MPEG–2 compression reduces the Standard TeleVision (STV) requirement to 3.5 Mbps and that of High-Definition TV (HDTV) to 8.0 Mbps, all of which is downstream, of course. PCM-based voice requires bidirectional bandwidth of only 64 kbps, which leaves plenty of room for high–speed bidirectional data [10].

Also known as ADSL Lite, Universal ADSL, and Splitterless ADSL, G.lite is an interoperable extension of the ADSL specification ANSI T1.413 and has been standardized by the ITU–T as G.992.2. Three deployment options exist, all of which support simultaneous voice and data communications over a single UTP local loop of one physical pair:

· Splittered ADSL resembles the original standardized version of ADSL, in that a single, professionally installed splitter is installed at the point of termination of the local loop at the customer premises. This approach ensures the highest level of data performance, while ensuring that other signals in a multipair cable are not affected by high–frequency ADSL data signals. But this approach also involves an expensive truck roll.

· Distributed Splitter ADSL involves a number of ADSL splitters, some front-ending PCs and others front–ending analog telephone sets. This approach ensures that the voice and data channels do not interfere with each other and avoids the costs and delays associated with the professional installation of a single splitter, as described in the Splittered ADSL option. This version does not require professional installation.

· Splitterless ADSL (see Figure 9.5), which is the most attractive, supports simultaneous voice and data without the requirement for either a centralized splitter or multiple distributed splitters. Rather, high–frequency data communications are supported over an ADSL modem. This version does not require professional installation.

G.lite supports both voice and data over the same circuit, just as does full–rate ADSL, but G.lite uses several mechanisms to optimize their coexistence. Power back-off dynamically reduces the modem power level at the customer premises in order to eliminate the potential for interference (NEXT) with the analog voice channel. Power back–off occurs automatically and reduces the upstream data transmission rate for the duration of the phone call. Fast retrain enables rapid recovery of the upstream when the phone call is terminated. Fast retrain supports the retention in memory of profiles for as many as 16 telephone sets, thereby taking into consideration the specific electrical attributes associated with each. Although early G.lite modems ran Frame Relay as the Layer 2 protocol, contemporary modems mostly run ATM [14]. As G.lite is limited in terms of bandwidth, it never achieved great popularity but did introduce the concept of Splitterless DSL.

IDSL offers several advantages over competing technologies. IDSL is based on ISDN network technologies, which are standardized, well understood, and widely implemented, at least outside the United States. IDSL operates at frequencies that do not interfere with other signals riding over adjacent pairs in a multipair copper cable. IDSL makes use of standard, off–the–shelf Data Communications Equipment (DCE) at the customer premises in the form of a TA with a few modifications. IDSL offers fully symmetric bandwidth, although it admittedly is less than that offered by competing DSL technologies, particularly in the downstream direction. IDSL does not require a truck roll for CPE purposes, although the NIU must be installed and the circuit may require conditioning, as is the case with traditional ISDN. Finally, IDSL is the only DSL technology that operates on local loops provisioned with conventional DLCs. On the negative side again, IDSL is application specific, supporting data only and, therefore, requiring an additional local loop. IDSL never gained much traction in the United States, where ISDN never achieved any appreciable level of success and where other broadband access technologies are widely available. Worldwide, IDSL has a market in areas where ISDN is already in place and other broadband access technologies are not available.

· Long Reach: The long-reach version runs at 12 MHz, divided among 2872 subcarriers, to deliver asymmetric data rates of as much as 55 Mbps downstream and 30 Mbps upstream over loops up to 1000 ft, after which data rates drop precipitously. Considerably reduced rates are achievable at distances up to 4000–5000 ft.

· Short Reach: The short-reach version runs variously at 17.6 MHz and up to 30 MHz, divided among as many as 4096 and 3478 subcarriers, respectively, to deliver symmetric (downstream and upstream) data rates of as much as 100 Mbps over loops up to 500 feet, after which data rates drop precipitously. Again, considerably reduced rates are achievable at distances up to 4000–5000 ft. This version will run in asymmetric mode as well.

VDSL2 employs the same DMT modulation scheme as ADSL but increases the number of subcarriers to 4096 in a typical implementation and up to 3478 in others, as compared to the 512 subcarriers specified in ADSL2+. Trellis–Coded Modulation (TCM) yields higher throughput on long loops where the Signal–to–Noise Ratio (SNR) is low, although data rates drop considerably beyond 500–1000 ft.

There are several application scenarios, all of which are based on FTTN hybrid local loops, with fiber optics from the CO to the neighborhood ngDLC and VDSL to the premises. In North America, Europe, and China, there are plans to deploy VDSL2 at downstream rates up to 30 Mbps and upstream rates of 3 Mbps in support of consumer–class triple–play services, including at least three Digital TV (DTV) or HDTV channels, Internet access at rates up to 5 Mbps, and VoIP. Clearly VDSL has application in servicing a MultiDwelling Unit (MDU), also known as a MultiTenant Unit (MTU), where the bandwidth can be shared among multiple tenants. In this Fiber-To-The-Premises (FTTP) scenario, the ngDLC is positioned on the premises, and the VDSL service runs over the inside wire and cable system in the apartment building or office building, for example [15].

As VDSL matures and implementations become more prevalent, the pressure increases to increase the bandwidth levels and speed ratings, to alter the bandplan toward symmetrical service, and to increase the distances. In combination, these demands will outstrip the capabilities of the specification. As a result, there has been some movement toward the application of Multiple-Input, Multiple-Output (MIMO) technology similar to that used in 802.11 Wireless Local Area Networks (WLANs). In a VDSL application, MIMO treats each binder group of 25 pairs as a MIMO channel and the transceiver treats multiple pairs together, thereby separating signal from crosstalk [16]. According to the DSL Forum, some manufacturers have included MIMO in VDSL2 systems in advance of standardization of MIMO technology.

Bellcore (now Telecordia Technologies) also developed High Bit–Rate Digital Subscriber Line (HDSL) at the request of the RBOCs as a more cost–effective means of providing T1 local loop circuits over existing UTP. Standardized by the ANSI T1E–1.4 committee, HDSL eliminates repeaters in the T1 local loop for distances up to 12,000 ft, which can be extended another 12,000 ft through the use of a line doubler, which essentially is an HDSL repeater.

HDSL2 is an HDSL variant that supports T1 speeds of 1.544 Mbps over a single twisted pair (two conductors) and over 24 AWG loops up to 13.2 kft, 768 kbps at up to 17.7 kft, and 384 kbps at up to 22.5 kft. The international E–1 version supports signaling speeds up to 2.3 Mbps. An HDSL2 regenerator can double the distance for each speed rating. A variation known as HDSL4 can run over two copper loops (four physical conductors) in order to extend the maximum transmission span to as much as 16,500 ft. HDSL2's level of performance is achieved through an advanced coding technique known as Trellis-Coded Pulse Amplitude Modulation (TC-PAM). This technique places three bits on a baud, which is an improvement over the two bits per baud realized through the 2B1Q technique used in HDSL. While the bandwidth required to support the HDSL signal is sensitive to the implementation, TCM features inherent Forward Error Correction (FEC) to overcome line impairments and issues of Signal–to–Noise Ratio (SNR), thereby offering excellent performance and the same reach as HDSL but requiring only a single pair. With respect to applications, HDSL is essentially a Layer 1 bit pipe that replaces leased full–rate and fractional T1 and E–1 local loops in support of any bit stream, including TDM voice, ATM, and Frame Relay, for example. Although the standard version does not support telephone service over the same line, some equipment manufacturers offer a passive POTS splitter [18].

Symmetric DSL (SDSL), also known as Single-line DSL, is an umbrella term for nonstandard variations on the HDSL theme running over a single pair. SDSL uses the same 2B1Q line coding as HDSL and runs at rates from 128 kbps to 2.32 Mbps. Example speeds are 384 kbps for loop lengths of up to 18,000 ft and 768 kbps for loop lengths of up to 12,000 ft (see Figure 9.8). Providers generally limit SDSL to transmission rates of 768 kbps, which is half the payload of a full T1. At that data rate, the signaling rate is 784 kbps, including 16 kbps of error control overhead. At a data rate of 384 kbps, which is one–fourth a full T1, the signaling rate is 400 kbps.

SHDSL can transport T1, E–1, ISDN, ATM, and IP signals, with the specific framing requirements negotiated during the handshaking process defined in G.994.1. Although SHDSL is intended for data–only applications, newer techniques will support voice as well [19–21]. Targeted applications include Web hosting, videoconferencing, Virtual Private Network (VPN) services, and remote LAN access. Note: SHDSL is specified for North America in ANSI T1E–1.4/2001–174, for Europe in ETSI TS 101524, and worldwide in ITU–T G.991.2.

Cable television, formally known as Community Antenna TeleVision (CATV), was born in the mountains in the state of Pennsylvania in the United States. In the late 1940s, there were only a few TV stations, all located in major cities. As TV transmit and receive antennas operated exclusively over the air, signal quality was sensitive to Line–of–Sight (LOS), ElectroMagnetic Interference (EMI) and Radio Frequency Interference (RFI), weather conditions, and other factors. Reception was poor, even in the cities, and awful elsewhere. In remote rural areas and particularly in mountainous areas where LOS was not possible, reception was next to impossible, even on the best of days. John and Margaret Walson, owners of the Service Electric Company, a retail appliance store in Mahanoy City, were having a difficult time selling TV sets as reception was so poor in the valley where the town was situated, 90 miles from Philadelphia. In order to demonstrate TV sets to their best advantage, Mr. Walson placed an antenna on top of a tall utility pole on a nearby mountaintop and ran antenna wire to the store. In June 1948, he built some amplifiers to bring the signal to customers who had bought his TV sets, thereby creating the first CATV network. Walson also was the first to use microwave to import TV signals from distant stations and the first to use coaxial cable to improve reception. In 1972, Walson also was the first to distribute Home Box Office (HBO), which marked the beginning of pay TV and the start of the explosive growth of the CATV industry [28, 29].

Beginning in the mid–1990s, a few large CATV providers began to upgrade their aging coaxial cable systems. As many of those networks were installed in the late 1960s and early 1970s, they were in awful shape. Further, those networks were strained for capacity as the CATV providers sought to increase revenues and profits through the introduction of premium movie channels and Pay-Per-View (PPV). The upgrades, in some cases, went beyond simple coax upgrade and replacement to include optical fiber in the trunk facilities from the head end to the neighborhood, where they terminated in an optoelectric conversion box that interfaced with the existing coax for ultimate termination at the premises.

There were several problems with this scenario. Most of the CATV operators were overburdened with debt and, therefore, could not afford to make the necessary capital investments in network upgrades. The requisite investment in the necessary circuit and packet switches added greatly to the problem. Further, the CATV providers lacked the necessary skills, network management systems, and billing systems to support transaction–based switched voice and data services. Also, no standards existed for the set–top boxes, or much of anything else actually. Also, the CATV networks generally were not interconnected—not even those owned by the large Multiple-System Operators (MSOs).

Major telecommunications companies began to acquire a number of CATV providers in the late 1990s, addressing many of the cash and management problems associated with this scenario. Most notably, AT&T in early 1999 acquired TCI (TeleCommunications, Inc.) and MediaOne, two of the largest CATV providers in the United States, and invested billions of dollars in system upgrades. The lack of standards for set–top boxes was resolved in March 1997 by the Data Over Cable Service Interface Specification (DOCSIS), developed by the limited partnership known as the Multimedia Cable Network Systems Partners Ltd. (MCNS) and now administered by CableLabs, which has relabeled the initiative CableLabs Certified CableMo-dem. Through matching DOCSIS Cable Modems (CMs) at the head end and the customer premises, two–way cable paths are provided over a Hybrid Fiber/Coax (HFC) system. The head–end portion of the network is in the form of a Cable Modem Termination System (CMTS), which supports a packet data connection to an IEEE 802.3 10/100–Mbps Ethernet port on a router. The system runs IP at Layer 3 (Network Layer) in Ethernet frames at Layer 2 (Data Link Layer). Associated with the CMTS are various servers for security, address translation, data caching, video caching, and so on. A CMTS can support as many as 2000 cable modem users on a single 6-MHz channel (8 MHz in Europe), with issues of congestion for shared bandwidth becoming more severe as the number of active users increases. The modem on the customer premises is in the form of a TV/data set–top box, which supports traditional coax connections to multiple TV sets and a 10/100Base–T Ether-net connection to a PC or to a hub serving multiple PCs. Figure 9.10 illustrates a typical CATV network configuration.

DOCSIS is both always on and generally asymmetric in nature, as are most xDSL technologies. DOCSIS 2.0 specifications address downstream and upstream channels:

· Upstream: In early versions of DOCSIS, transmissions upstream from the user premises to the head end are supported either through the contentious CSMA/ CD access protocol used in Ethernet LANs or through Time Division Multiple Access (TDMA), depending on the specifics of the CATV provider's implementation [30, 31]. According to DOCSIS 1.0 standards, traffic in the upstream direction is supported over 6-MHz channels in the range between 5 and 42 MHz for U.S. systems and 8-MHz channels in the range between 5 and 65 MHz for European systems, each of which can operate at about 12 Mbps through use of the Quadrature Phase Shift Keying (QPSK) modulation technique. Again, framing and FEC overhead reduce the level of usable bandwidth, in this case to about 10 Mbps [32–35]. In a field implementation, many systems use QPSK across a 1.6-MHz channel, yielding 2.56 Mbps (2 bits per symbol × 1.28 megasym-bols per second). DOCSIS 1.1 added 16-QAM to the mix, roughly doubling the data rate and doubling the channel width to 3.2 MHz to double it yet again (4 bits per symbol × 2.56 megasymbols per second), for a four fold increase to over 10 Mbps. QAM increases sensitivity to noise, but DOCSIS deals with that through sophisticated noise cancellation techniques. DOSCIS 2.0 triples the upstream rate of DOCSIS 1.1 over a channel of 6.4 MHz using either TDMA in combination with 64-QAM or Synchronous Code Division Multiple Access (S–CDMA) in combination with 128-QAM Trellis-Coded Modulation (TCM). S–CDMA transmits 128 orthogonal codes simultaneously (i.e., in the same time slot) [36].

There also are telco-return modems that use the circuit–switched PSTN for the upstream path, although this approach generally proves less than totally satisfactory. Not only does the telco–return approach require the use of a telco–provided local loop, which adds costs to the equation, but it slows the upstream path to modem speeds. Also, the widely mismatched speeds on the upstream and downstream paths cause the downstream server to slow down, therefore running at less than optimum efficiency. DOCSIS 3.0 standards currently under development are expected to increase upstream data rates to 120 Mbps and downstream rates to 160 Mbps.

Wireline local loops are problematic. Cabled media are expensive and time consuming to deploy. Right–of–way must be secured, trenches must be plowed and conduits placed, poles must be planted and crossarms must be hung, conductors must be spliced, amplifiers and repeaters must be connected, and so on. The vast majority of loops are copper twisted pair, which tends to offer relatively little in terms of bandwidth and error performance and is highly distance limited, especially at the higher frequencies required to maximize data rates. The ILECs own most of the conventional local loops, CATV providers own the coaxial cable networks, and neither of them is the least bit inclined to share them, except as required by the regulators. As I discussed earlier in this chapter, the decision boils down to one of build or buy (i.e., lease), and buy is not always an option.

In building local loops, the advantages of Wireless Local Loop (WLL), also known as fixed wireless, are fairly obvious at this point in our discussion of communications systems and networks:

· WLL systems often can be deployed much more quickly and much less expensively that wireline systems.

· The aggregate bandwidth of a WLL system can be apportioned to end users on a channelized basis and often very flexibly so.

· The system can be configured and reconfigured remotely, often without a truck roll

The disadvantages of WLL also are fairly obvious:

· Interference always is an issue, and one that in many ways is out of the control of both the service provider and the user organization. While licensed frequency bands are protected from direct interference from others who might covet the same frequencies, unlicensed bands are available for all to share. Regardless of whether the band is licensed or not, Radio Frequency Interference (RFI) and Electromagnetic Interference (EMI) are always issues. Such interference can be caused by electric motors, radio systems transmitting out of their assigned bands or in excess of their prescribed power levels, and forces of nature such as lightning, static electricity, and solar flares. Additionally, the quality of the airwaves is always an issue, as the quality of the signal can be negatively affected by precipitation, fog, humidity, dust, smoke, pollution, and temperature.

· Distances of terrestrial systems are limited due to factors such as quality of the airwaves and the frequencies used. The systems operating in the higher frequency ranges (e.g., 28–31 GHz) suffer from attenuation to a much greater extent than those in the lower frequency ranges (e.g., 2.5–2.7 GHz).

· Line–Of–Sight (LOS) is always preferable and is absolutely required at the higher frequencies.

· Security is always an issue with RF–based systems.

The WLL options available are numerous and include nonstandard and standard approaches and both licensed and unlicensed frequency bands. This discussion of WLL options includes the most significant and what I consider to be some of the most interesting. Regardless of the specific technology, RF–based WLL configurations include centralized antennas, or base stations, that either are located at the edge of the carrier network or are connected to the network edge via optical fiber, DSL, T1/E–1 facilities, or perhaps microwave. The base stations connect to matching antennas at the customer premises, as illustrated in Figure 9.11. Free Space Optics (FSO) solutions involve infrared (IR) transmitter/receivers.

As noted above, LMDS generated a lot of excitement in 1998 and again in the 1999–2000 time frame, and hundreds of millions of dollars were spent securing LMDS licenses at spectrum auctions in the United States. Most of the successful bidders (e.g., Teligent, WinStar, and Advanced Ratio Telecom) have since declared bankruptcy. XO Communications (previously Nextlink) invested a total of approximately $ 900 million in those licenses and now holds licenses in 73 major cities in the United States. XO installed a small number of systems over the intervening years but never was able to achieve any real success with LMDS due to technological problems and high equipment costs. In early 2006, however, XO announced a new LMDS initiative targeted at cellular backhaul as well as enterprises and government agencies looking for access alternatives, redundancy, and disaster recovery solutions [40–42]. [Note: In this context, backhaul refers to the transporting of traffic between distributed sites, such as cellular base stations, and a backbone or centralized network or site, such as a Mobile Traffic Switching Office (MTSO), or other point of concentration or switching.] Most of the current interest in WLL is on IEEE 802.16, aka WiMAX, which I discuss later in this chapter.

Operating in five bands in the range 2.15–2.68 GHz in the United States and Canada and in the 3.5-GHz range elsewhere, Multichannel Multipoint Distribution Service (MMDS), also known as wireless cable, initially was developed for one–way TV transmission. As the name suggests, MMDS is a point–to–multipoint technology that operates from an antenna typically placed on a hilltop or other location in order to maximize Line–Of–Sight (LOS) connectivity. The antenna may be omnidirectional or may be sectorized in order to improve spectrum efficiency, and distances of up to 31 miles (50 km) can be achieved at allowable power levels due to the excellent signal propagation characteristics in this relatively low range of the microwave spectrum.

Today, the world is still enjoying the results of new scientific discoveries that are constantly being made, and the one that seemed most marvelous when they were first announced, become commonplace after a few years have passed…. It was a triumph when his [Marconi's] experiments resulted in communications at will without wires over distance of 250 miles.

W. J. Jackman et al, The World's Workshop, The C. F. Beezley Company, 1911

Worldwide Interoperability for Microwave Access (WiMAX) is a Broadband Wireless Access (BWA) solution that has a maximum range of approximately 30 miles, which would hardly impress Marconi's fans but offers bandwidth that they would not have appreciated at the time. WiMAX is based on the standards recommendations from the IEEE 802.16 Working Group and the European Telecommunications Standards Institute (ETSI) HiperMAN group. WiMAX is promoted by the WiMAX Forum, a special–interest group with members from the manufacturing, carrier, service provider, and consulting communities. Although WiMAX and 802.16 are not exactly the same, technically speaking, most people, including myself, use the terms interchangeably. WiMAX is just easier to say and it looks better in print. The same goes for Ethernet and 802.3

· 802.16 (2001) standardized LMDS. This first specification focused on fixed wireless solutions in both point-to-point and point-to-multipoint configurations. 802.16 specified frequencies in the range 10–66 GHz and required LOS. This first standard garnered little support.

· 802.16a (2003) was based on MMDS and the European HiperMAN system. This extension operates in the range 2–11 GHz, which includes both licensed and license–exempt bands. In the lower end of the range, LOS is not a requirement, although it is always preferable. 802.16a specifications include both point-to-point and point-to-multipoint configurations.

· 802.16d, aka 802.16–2004 (2004), is a compilation and modification of previous versions and amendments 802.16a, b, and c. 802.16d specifies frequencies in the range 2–11 GHz and includes point-to-point, point-to-multipoint, and mesh topologies. This specification recommends, but does not require, LOS and includes support for indoor CPE.

· 802.16e (October 2005), formally known as Mobile WirelessMAN, adds hand-off capability, thereby supporting portability and pedestrian speed mobility for users of laptop, tablet, and hand–held computers. Operating in the range 2–6 GHz, it is designed for point–to–multipoint applications and does not require LOS.

WiMAX standards address both LOS and NLOS scenarios. Where LOS can be achieved, WiMAX cell radius is as much as 50 km (31 miles). Under NLOS conditions, the maximum cell radius is approximately 9 km (5.6 miles). The fixed wireless standards provide for aggregate raw bandwidth up to about 70 Mbps per Base Station (BS), although the data payload is much less due to overhead. The level of actual throughput depends on LOS, distance, air quality, EMI and RFI, and other signal impairments. Mobile network deployments (802.16e) are expected to provide up to 15 Mbps of aggregate raw bandwidth within a cell radius of up to 3 km. Whether the specification is fixed or mobile, WiMAX data rates and distance are tightly and inversely related. In other words, the maximum signaling rate of 70 Mbps and the maximum throughput of approximately 40 Mbps can be realized over the shortest distance between the BS and the user antenna under LOS conditions. Over the maximum distance of 50 km under LOS conditions, or the maximum distance of 9 km under NLOS conditions, throughput drops considerably. WiMAX specifications address NLOS performance issues through a number of mechanisms, including the following:

· Orthogonal frequency division multiplexing subdivides the spectrum into a number of independent, narrowband subcarriers, across which it sends the signal in parallel fashion. The receiving antenna monitors all subcarriers, reassembling the multiple parallel signals into one high–speed serial output.

· Antenna design clearly impacts signal strength. The fixed wireless specifications call for directional antennas in order to reduce multipath fading and, thereby, improve signal strength and cohesiveness. The directional antenna at the customer premises may be in the form of an adaptive, passive array pizza box, so called because it is about the size and shape of a pizza box. Such an antenna possesses beamforming properties that permit it to adjust its logical focus to maximize the strength of the incoming signal from the BS at the network head end. These adjustments in focus are accomplished passively, as no physical (i.e., mechanical) reorientation is required.

· Multiple Input, Multiple Output (MIMO) antennas employ space/time coding to compensate for multipath fading over long loops. Copies of the same RF signal are transmitted by multiple antennas separated by physical space and received by multiple antennas also spatially separated. These intelligent antenna systems can compensate for multipath fading and realize diversity gain, that is, increase in signal strength. This same technique is specified in 802.11n and VDSL and has been used in long–haul microwave systems for many years.

· Rate-adaptive modulation dynamically adjusts the signal modulation technique of each carrier, individually, to compensate for variations in signal quality at that carrier frequency. When the signal is strong, an intelligent WiMAX system can use the most sophisticated modulation scheme, which yields the highest transmission rate. As the signal fades due to factors such as temperature and interference, error performance degrades and link stability is jeopardized. In such a situation, the system can shift to successively more robust modulation schemes, but slower and more overhead intensive, to compensate for those conditions. While transmission rates drop accordingly, at least the link is maintained. The modulation schemes employed, from the most efficient to the least, are specified as 256–QAM, 64–QAM, 16–QAM, QPSK, and BPSK.

· Reed–Solomon Forward Error Correction (FEC) is employed to deal with issues of signal quality. While FEC inherently involves some degree of bit–level redundancy, it provides the receiver with enough data to reconstruct a large percentage of frames errored in transit. Automatic Repeat reQuest (ARQ) is employed to request retransmission of any remaining errored frames.

· Power control is the responsibility of the base station, which analyzes signal strength and advises the remote terminals. On the basis of that information, the remote terminals can dynamically adjust their transmission levels to conserve power and to minimize the likelihood of cochannel interference with other CPE in proximity.

802.16 specifications include several multiplexing options. Time Division Duplex (TDD) supports Half–DupleX (HDX) communications and Frequency Division Duplex (FDD) supports both HDX and FDX.

Note that the transmission rate in Table 9.6 is stated in terms of throughput, in consideration of overhead and assuming reasonable signal performance. In a 3.5-MHz channel and at a maximum of 5 bits/Hz, the raw signaling rate is 17.5 Mbps. In consideration of overhead factors such as framing and error control, the maximum actual bidirectional data throughput is anticipated to be approximately 11 Mbps for a standard system operating at short range and 8 Mbps for a full-featured system operating at long range, as illustrated in Figure 9.12. 802.16 includes provisions for improved efficiency through header suppression, concatenation, and fragmentation.

The 802.16 security protocol is built on enhancements to the Privacy-Key Management (PKM) developed for cable modem communications. The protocol uses X.509 digital certificates with Rivest–Shamir–Adleman (RSA) encryption for authentication and key exchange. Traffic encryption options are Data Encryption Standard (DES) and Advanced Encryption Standard (AES).

The 802.16 specifications include convergence sub layers designed for mapping services to and from 802.16 connections. The ATM convergence sublayer is for ATM services and the packet convergence sublayer is for packet services such as IPv4, IPv6, Ethernet, and Virtual LAN (VLAN). As 802.16 is connection oriented, all services, including those inherently connectionless (e.g., SMTP and UDP) in nature, are mapped to a connection. This approach pro vides a mechanism for requesting bandwidth, negotiating service parameters, and establishing Quality of Service (QoS) levels. Downstream transmission is on the basis of a TDM signal, with individual stations allocated time slots serially. As the link is under the control of the BS, downstream QoS is straightforward. Upstream access is by TDMA and QoS is the function of a set of scheduling schemes that the BS has at its disposal in order to optimize performance. An important feature of the Medium Access Control (MAC) layer is an option that allows the BS to grant bandwidth to an intelligent subscriber station, rather than to the individual connection it supports. This allows the station to manage its bandwidth allocation among the users and applications it supports, which provides for more efficient bandwidth allocation in multiuser and multitenant applications [49]. 802.16 defines four polling schedules, as follows:

· Unsolicited Grant Service (UGS) is designed for services that periodically generate fixed units of data. TDM services T1 and E–1 are examples.

· Real-time polling service is designed for services that are dynamic in nature but require periodic dedicated request opportunities to meet real–time demands. Examples include real–time compressed voice services such as VoIP and IP-based streaming audio and video.

· Non-real-time polling service is identical to real–time polling service, except that connections may use random transmit opportunities. Internet access with a minimum guaranteed connection rate is an example of such an application.

· Best effort service provides neither throughput nor latency guarantees.

WiMAX has its share of weaknesses as well. EMI and RFI are always issues with RF–based wireless systems, which places WiMAX at somewhat of a disadvantage compared to DSL and PON. DSL and PON also offer dedicated bandwidth, while WiMAX bandwidth is shared, much like that of cable modems. Also, PON, cable modems, and VDSL all offer higher levels of bandwidth. Competition from the IEEE 802.20 Working Group will also be an issue. That group is chartered to develop similar standards for wireless access systems operating in licensed bands below 3.5 GHz. The technical goal is that of optimizing IP–based data transport, targeting peak data rates per user at over 1 Mbps and supporting vehicular traffic at speeds up to 250 km/h (155 mph). There also will be competition from the IEEE 802.22 Working Group, which is developing a standard intended for wireless data over UHF and VHF spectra currently used for broadcast TV. Also known as Wi-Fi TV, 802.22 targets the Radio Area Network (RAN), which it defines as having a range of up to 30 miles. Competition extends to Europe and Asia as well. ETSI chartered the Broadband Radio Access Networks (BRAN) project. HiperAccess is for frequencies above 11 GHz and HiperMAN for below 11 GHz. ETSI and IEEE 802.16 cooperate to some extent. South Korea's Electronics and Telecommunications Research Institute (ETRI) developed Wireless Broadband (WiBro) for 100 MHz of spectrum allocated in the 2.3-GHz band by the Korean government. WiBro offers aggregate throughput of 30–50 Mbps and has a reach of 1–5 km. Spectrum issues around the world exist, even though WiMAX includes both licensed and license-exempt bands. WiMAX-targeted spectra include license-exempt 2.4- and 5.8-GHz bands, which invite interference. Sprint controls much of the licensed 2.5-GHz band in the United States and that same band already is used for fixed wireless and cable transmission in Mexico [50–54].

While there certainly are a number of issues swirling around WiMAX, there is a tremendous amount of interest in it. The numerous applications include full–rate and fractional T1 and E1 services, especially for remote rural or developing areas where broadband telco or cable access is unavailable. ILECs and CLECs, alike, have shown interest in WiMAX for provisioning circuits in urban areas where telco or cable upgrades are not easily, quickly, or cost effectively implemented. Backhaul applications for Wi-Fi hotspots and cellular net-works are a natural application, as is disaster recovery. Several manufacturers have announced multimode chips that will allow portable computers to connect via both Wi-Fi and WiMAX, and 802.16e supports portability and mobility for users of laptop and hand–held computers. The 802.16e air interface also likely will be included in trimode voice handsets that will interoperate with Wi-Fi and cellular networks.

In the frequency ranges of 24 and 38 GHz, microwave has been licensed by the FCC to a number of service providers for point–to-point WLL applications. The licenses are for aggregate channel capacity of 100 MHz. Some carriers hold multiple licenses, which yield aggregate bandwidth of 400–500 MHz or more. In these microwave frequency ranges, LOS is required and distances are limited to approximately 5 miles. Typically, the cells are much smaller, in consideration of the advantages of frequency reuse. Error performance for these digital microwave systems is in the range of 10 -13, which compares very favorably with UTP-based T1 at 10 -7, and under optimal conditions, it compares with optical fiber. The paired transmit/receive antennas typically are about 1 ft in diameter. Larger antennas and variable power levels are employed to compensate for rain fade in areas where rainfall is heavy and prolonged. As is the case with LMDS, licensed microwave generated a lot of excitement in the 1999–2000 time frame, and hundreds of millions of dollars were spent securing spectrum at FCC auctions in the United States. Most of the successful bidders (e.g., Advanced Radio Telecom, Teligent, and WinStar) since have declared bankruptcy and the licenses have been acquired by others.

Personal Communications Services (PCS), the U.S. term for the Personal Communications Network (PCN) concept originally developed in the United Kingdom, is a service concept that is technology dependent and operates on a set of frequencies set aside specifically for that purpose. PCS ultimately intends to provide a full range of enhanced services through a single device and utilizing one telephone number, which will work anywhere and anytime, for life. Note: The term PCS also is used in the United States for a digital cellular telephony alternative using analog AMPS bands in the range 800–900 MHz. Actually, at this point, PCS means just about anything. In this section, I explore PCS in the context of the FCC licenses, with a focus on WLL.

While the spectrum allocation varies by country or region, the designated U.S. frequencies include narrowband, broadband, and unlicensed PCS. Actually, much of the PCS spectrum was already spoken for by high–band microwave systems, and clearing that spectrum is the responsibility of the PCS licensees. For the first time in U.S. history, that spectrum was auctioned (and reauctioned in some cases, after default on down payments—a long story) by geographic area. The 1995 auctions of two frequency blocks brought in $ 7.7 billion, and the 1996 auctions accounted for another $ 10.22 billion.

· Narrowband PCS has been allocated spectrum in the ranges of 900–901 MHz, 930–931MHz, and 940–941MHz. That AMPS spectrum is used to extend the capabilities of pagers and cell phones to include acknowledgment paging, two-way messaging, and digital voice.

· Unlicensed PCS spectrum serves low–tier applications such as wireless LANs, wireless PBXs, PDAs, and PCS voice and data services within a building or campus environment. Unlicensed PCS spectrum has been set aside in two ranges, with 20 MHz in the range 1.91–1.93 GHz and 10 MHz in the range 2.39–2.40GHz.

There are no mandatory service offerings associated with PCS licenses. Rather, each license holder defines its own services and applications.

In a local loop context, infrared (IR) light transmission systems also are known as Free Space Optics (FSO) and sometimes characterized as wireless fiber. FSO systems are airwave systems that use the infrared light spectrum in the terahertz range to send a focused light beam to a small receiver that often resembles a Direct Broadcast Satellite (DBS) dish. The transmitter/receivers can be mounted on rooftops or even indoors behind windows.

Contemporary IR systems offer substantial bandwidth at relatively low cost. FSO systems currently operate at rates of 1.544 Mbps (T1), 45 Mbps (T3), 155 Mbps (OC-3), and 622 Mbps (OC–12). Some vendors have advertised systems operating at 1 Gbps in support of GbE (Gigabit Ethernet) and 10 Gbps in support of 10GbE, and systems running at up to 160 Gbps have been demonstrated in the labs. Like microwave systems, FSO systems require LOS. In fact, some systems are so sensitive to LOS as to require autotracking mechanisms where they must adjust to the movements of high–rise office buildings due to wind sway, tremors, and other forces of nature. FSO systems also suffer from environmental interference, particularly fog, which acts like a prism to scatter the light beam. Under optimum conditions, distances are limited to about 2–5 km, although most tests have shown optimum performance at distances 500m to 1 km. In areas where there is a lot of fog, links more typically are limited to about 200 m, and some manufacturers offer redundancy via RF systems running in unlicensed bands. Under optimum conditions, error performance is in the range of 10 -8, which compares favorably with UTP and microwave.

Wi-Fi certainly enjoys great popularity, which has increased considerably with the introduction of 802.11g, and 802.11n promises to build on that popularity with increased transmission speeds and ranges due to the incorporation of MIMO technology. In addition to private LAN applications, Wi-Fi works well in public settings. Wi-Fi hotspots are now quite commonplace in public venues such as airports and coffee shops and even on airplanes, usually on a pay–per–minute basis. A number of municipalities (e.g., Philadelphia, Pennsylvania) in the United States have installed free Wi-Fi hotspots in downtown areas to encourage people to frequent those areas and patronize local merchants, and great many more municipalities plan to offer such services in the future. Google, one of the leading ISPs, announced in August 2006 that it invested approximately US $ 1 million to build a Wi-Fi network of 380 radio antennas designed to provide free Internet access to the entire population of its home town, Mountain View, California. The town covers some 11.5 square miles and has a population of approximately 72,000. Commercial ISPs quite naturally are opposed to Internet access as a public utility and are not particularly happy about Google's initiative, but they so far have had only very limited success in battling this trend. Interestingly, groups of residential end users have installed Wi-Fi hotspots in some neighborhoods for their joint use. This approach is reminiscent of some of the old co–op CATV networks.

In any case, a public Wi-Fi hotspot essentially is a WLAN positioned as a highly localized wireless access point. The Wi-Fi network connects to the Internet via a truly broadband local loop technology such as ADSL, WLL, or PON.

Elements of a PON (Figure 9.13) comprise an Optical Line Terminal (OLT), an Optical Network Terminal (ONT), an Optical Network Unit (ONU), a passive splitter, and optical fiber:

· Optical Line Terminal (OLT): The OLT is located in the carrier's CO, or head end, where it serves to terminate the optical local loop at the edge of the network. The OLT can either generate optical signals on its own or pass SONET signals from a collocated SONET cross–connect or other device, broadcasting them downstream through one or more ports. The OLT also receives the upstream signals from the ONTs and ONUs.

· Optical Network Unit (ONU): An ONU is used in a Fiber-To-The-Curb (FTTC) scenario, in which the fiber stops at the curb, with the balance of the local loop provisioned over embedded coax in CATV networks or UTP in conventional telco networks. An ONU also is used in a Fiber-To-The-Neighbor-hood (FTTN) scenario, in which it is positioned at a centralized location in the neighborhood, with the balance of the local loop being provisioned over embedded coax or UTP. While this FTTN scenario maximizes the use of embedded cable plant and therefore minimizes the costs associated with cable plant replacement, it compromises performance to some extent.

There are several variations on the PON theme, each of which specifies Data Link Layer (Layer 2) protocols, transmission speeds, loop lengths, split ratios, multiplexing techniques, maximum loss levels, and distance limitations. Those standards are listed below and compared in Table 9.7:

· ATM–based Passive Optical Network (APON) is the term applied to the original specifications set by FSAN and ratified by the ITU–T as G.983.1 (1998). As the underlying bearer protocol, APON specifies Asynchronous Transfer Mode (ATM), which was favored by the ILECs for DSL and for their internal backbone networks. In contemporary terminology, APON generally is known as Broadband Passive Optical Network (BPON), which is described in G.983.3 (2001). APON runs in asymmetric mode at 622 Mbps downstream and 155 Mbps upstream or in symmetric mode at 155 Mbps. BPON supports as many as 32 splits over a distance of as much as 20 km (12 miles). The BPON supports voice, data, and video in ATM format.

· Gigabit Passive Optical Network (GPON), formally known as ITU–T G.984 (2004), describes both asymmetric and symmetric configurations and currently supports transmission rates as high as 2.488 Gbps. GPON supports as many as 32 or 64 splits, with future expectations of 128 splits, and a maximum logical reach of approximately 60 km (37 miles). At full speed of 2.488 Gbps with the maximum of 64 splits, each user has access to sustained bandwidth of more than 35 Mbps, which is far beyond that offered by other access technologies. GPON supports voice, data, and video in ATM format. GPON also supports voice in native PCM/TDM format and data in Ethernet format.

The ITU–T standards for Wavelength Division Multiplexing (WDM) certainly apply to BPON and GPON, both of which specify one wavelength for downstream voice and data and another for video. BPON supports 16 wavelengths with 200-GHz spacing and 32 wavelengths with 100-GHz spacing between channels. Upstream transmissions do not make use of WDM, as there is not upstream video requirement. EPON does not make use of WDM at this time.

Security is somewhat of an issue in a PON network, as the OLT is shared among a large number of users. So, PON makes several provisions for enhanced security. First, there are provisions for any number of higher layer encryption mechanisms to act on the downstream transmission. GPON, for example, currently specifies Advanced Encryption Standard (AES) for downstream transmissions. Encryption occurs at the OLT, and decryption occurs at the ONT or ONU. Encryption of the upstream transmissions is considered unnecessary as the shared OLT is at the head end, which is under the control of the service provider, which is a trusted party, and other users do not have access to the upstream channel. BPON provides for enhanced security through a technique known as churning. Downstream transmissions are encrypted through the use of a byte–oriented churn key exchanged between the OLT and the ONU. The ONU or ONT generates the key and sends it to the OLT, which uses it to encrypt downstream transmissions. The key is changed at least once a second—hence the term churning. Again, no security is deemed necessary on the upstream transmissions.

Upstream signals are supported by a TDMA mechanism, with the transmitters in the ONUs operating in burst mode. In this scheme, each subscriber takes turns transmitting upstream, with the length of the transmission window determined by the head end based on the upstream bandwidth subscription. BPON supports only an asymmetric transmission mode, EPON supports only symmetric, and GPON supports both.

· Greenfield deployments are those in new subdivisions, newly developed areas, or other areas where there is no existing telecommunications wireline infrastructure. The nature of this opportunity is such that, while trenches must be dug, conduits placed, poles planted, and so on, the deployment is engineered specifically for fiber optics. A greenfield deployment most likely would be in the form of FTTP, which takes fiber directly to the premises.

· Brownfield deployments take place where there is existing wireline infrastructure in the form of twisted pair, which is removed to create space for PON. A brownfield deployment typically takes the form of FTTN or FTTC, in which case the distribution cables and Digital Loop Carrier (DLC) systems are removed and replaced with optical fiber and PON ONUs. The link from the ONU to the premises is in the form of ADSL or perhaps VDSL over the embedded twisted pair, which yields a hybrid fiber/copper configuration.

· Overlay deployments are those in which the new infrastructure parallels that of the existing infrastructure. This approach allows the service provider to construct the new system and provide service to PON subscribers as required, while continuing to serve others subscribing to more basic services from the old cable plant. This phased approach generally is preferred for FTTP deployments where there is existing infrastructure [58, 59].

The vast majority of the interest in PON is on the part of the ILECs, particularly in the United States, where the CATV providers have captured the greatest share of the broadband access market with cable modem technology. As PON will once again provide the ILECs with a competitive advantage, they are deploying it aggressively in major markets, with the current emphasis on BPON, although GPON will likely supplant it once standards–based equipment becomes widely available. Render, Vanderslice & Associates estimated in a report dated September 2005 that there were FTTH deployments in 652 communities in 43 states, with a total of 2.7 million potential customers passed and 323,000 subscribers [60]. While I find the number of PON subscribers a bit difficult to believe in the context of that time frame, I certainly do not doubt that PON deployment is moving forward quickly.

The Verizon service includes up to nine e-mail accounts, Yahoo! or MSN Premium service, and 10 MB of personal Web space. The inside wire to the wired or wireless router must be upgraded to Cat 5 UTP. If a wireless router is used, 802.11b is acceptable for the low–speed service offering but 802.11g is required for the two higher speed offerings. The customer is responsible for providing an electrical outlet close to the location where the ONT is to be mounted in order to provide power to the battery backup unit. The core offering of the English language FiOS Premier TV is available at $ 39.95 per month, while the Spanish language La Conexion is available for $ 27.95, assuming that the subscriber also has Verizon voice service. Prices are slightly higher for standalone TV service. Various other packages, add–ons, on-demand content, and Pay–Per–View (PPV) services are additional. As many as 350 TV channels are available in total. Pricing is sensitive to bundling, with the lowest prices offered to those subscribing to the full triple play of voice, Internet access, and TV service. While Verizon FiOS is not available everywhere in Verizon territory, the company spent over US $ 1 billion in 2005 on FiOS–related capital expenditures. At an estimated cost of $ 1250–$ 1350 per premises for FTTP, $ 1 billion can go pretty quickly [61].

AT&T (previously SBC) largely is taking a dual approach in its US $ 4 billion capital initiative known as Project Lightspeed. In most new subdivisions and other greenfield scenarios, the company will employ FTTP. In brownfield scenarios, the company favors a hybrid fiber/copper approach, which it calls Fiber-to-the-Node (FTTN), with each node delivering 20–25 Mbps to as many as 300–500 homes. AT&T will use copper to provide ADSL and VDSL services from the node to the customer premises. As the UTP portion of the loop is shortened considerably in an FTTN scenario, many of the associated performance issues are much reduced. BellSouth has taken a relatively passive Fiber–to–the–Curb (FTTC) approach which is quite complementary to the AT&T approach, and that fact quite likely had some impact on the merger discussions between the two companies [62, 63]. FTTN costs are estimated to be in the range of $ 250 for overlays, which compares favorably with the cost of FTTP.

Elsewhere, EPON is being aggressively deployed in Japan, where it is known as GigE PON (Gigabit Ethernet PON), and is the overwhelming choice throughout Asia. PON spending and penetration levels in Japan currently far exceed those in the United States, while Europe currently lags well behind.

Joseph H. Adams, Harper's Electricity Book for Boys, Harper & Brothers Publishers, 1907

Broadband over Power Line (BPL) has given fresh spark to an old technology known as PowerLine Carrier (PLC), which Bell Telephone Laboratories invented in 1928 as a means of supporting voice and data communications over electric power distribution cabling. Although it proved too unreliable for commercial use at the time, electric utility companies have continued to use PLC for telemetry applications and controlling equipment at remote substations. Telephone companies have used PLC for many years in rural areas to provide Plain Old Telephone Service (POTS) to extremely remote customers who had electric service but for whom it would be too costly to provide telephone service over dedicated twisted–pair local loops. BPL is a much advanced broadband version of PLC that represents an alternative to more visible technologies being deployed by the LECs and CATV providers.

Access BPL is a form of PLC that uses certain elements of the existing electrical power distribution grid as a broadband local loop. The typical power grid comprises generators, High–Voltage (HV) lines, substations, Medium–Voltage (MV) lines, transformers, and Low–Voltage (LV) lines. The HV transmission lines are unsuitable for BPL as there is too much hum and buzz (i.e., noise) at 155,000–765,000 V, and the power jumps all over the frequency spectrum in a completely unpredictable manner. MV and LV lines are quite usable, however.

The National Telecommunication and Information Administration (NTIA) of the U.S. Department of Commerce describes three BPL system types:

· System 2 differs from System 1 only in that it bypasses the LV drop altogether through the use of Wi-Fi technology.

· System 3 employs Direct-Sequence Spread Spectrum (DSSS) in addition to Carrier Sense Multiple Access with Collision Detection (CSMA/CD) for data transmission over the MV distribution lines. DSSS is a packet radio technique in which the signal is spread across a wide carrier frequency in a redundant fashion. Thereby, multiple transmissions can share the same wideband, with the transmissions separated from each other through the use of a unique code sequence prepended to each data packet. Multiple systems may be installed on adjacent lines in a two–or three–phase MV distribution line. CSMA/CD is the same MAC used in 802.3 (Ethernet) wired LANs [64].

Access BPL configurations can be either asymmetric or symmetric in nature, with the aggregate amount of bandwidth sensitive to the specific nature of the power grid, for example, one-, two-, or three–phase MV distribution lines. As the BPL local loop is shared, the aggregate bandwidth is shared, and the bandwidth available to each subscriber is highly dependent on system loading and the level of activity at any given moment, much as is the case with CATV cable modem networks and WiMAX networks. Note that as both access BPL and in–house BPL operate over power lines, they cease to work when the lights go out, and there is no point in battery backup. It remains important to have an Uninterruptible Power Supply (UPS) with battery backup to allow graceful shutdown of computer equipment, of course.

Keeping in mind that access BPL is an emerging technology, yet to be standardized, early tests and trials in the United States have been quite positive, and there are a small number of commercial service offerings. In some cases, access speeds are in the range of 200–300 kbps, fully symmetric. Current Communications, which operates in the area of Cincinnati, Ohio, offers fully symmetric access speeds of 1 and 3Mbps.

The IEEE currently is developing P1675, "Standard for Broadband over Power Line Hardware," which is intended to provide electric utilities with a comprehensive standard for the injectors and extractors used on the MV distribution lines. The standard was targeted for completion in 2006 but that year came and went.

With the exception of the interstate power grid, power utilities tend to be regulated at the state and local levels, at least in the United States. Many power utilities are not only regulated by but are owned by municipalities, which creates a considerable conflict of interest. In recognition of this conflict and the inherent unfairness of a municipality's competing with a private sector telecommunications company, at least 14 states have passed laws preventing municipally owned utilities from offering telecom services, and others are considering similar measures [65–68].

Synchronous Optical Network (SONET) is a set of North American standards for broadband communications over SingleMode Fiber (SMF) optic transmission systems, thereby enabling manufacturers to build equipment that supports full inter-connectivity and interoperability. Synchronous Digital Hierarchy (SDH) is the internationalized version of SONET, as specified by the CCITT (now ITU–T). As SONET and SDH differ primarily with respect to low–level line rates and some terminology, I lump them together. SONET/SDH uses a transfer mode that defines switching and multiplexing aspects of a transmission protocol, supporting both asynchronous and synchronous traffic in any form on bit–transparent TDM channels. Intended primarily for the carrier networks, SONET/SDH also can be deployed to the user premises, although such implementations are reserved for sites where there are significant bandwidth requirements. The Network-to-Network Interface (NNI), also known as Network Node Interface, specification allows the blending of national and regional networks into a cohesive global network. The User Network Interface (UNI) provides a standard basis for connection from the user premises to SONET/SDH.

Initial SONET standardization efforts began in 1984, when MCI (now part of Verizon) proposed the development of connectivity standards, or midspan fiber meets, to the Interexchange Carrier Compatibility Forum (ICCF). The ICCF then requested that the Exchange Carriers Standards Association (ECSA) develop those standards [69]. In 1985, Bellcore proposed SONET to the ANSI T1X1 committee, and the process continued until final approval for a much–revised SONET was gained in June 1988. The resulting ANSI specification for SONET Phase 1 was released in T1.105–1988 and T1.106–1988, specifying a basic transmission level of 51.84 Mbps, which carries a signal originating as the electrically based T3 of 45 Mbps, plus additional overhead for optical processing and network management.

The CCITT (now ITU–T) began the initial efforts to internationalize SONET as SDH in 1986. In 1988, the CCITT accepted the SONET standards, with modifications that were mostly at the lower multiplexing levels. These differences largely are due to the requirement to accommodate the complexities of internetworking the disparate national and regional networks. The ITU–T Recommendations referenced are G.707, G.708, and G.709.

Standards work continues on SONET/SDH, with the involvement of standards bodies and associations including ANSI, EIA, ECSA, IEEE, ITU–T, and Telcordia Technologies:

· American National Standards Institute (ANSI): Founded in 1918, ANSI coordinates and harmonizes private sector standards development in the United States. ANSI also serves as the U.S. representative to the International Organization for Standardization (ISO), the originator of the Open Systems Interconnection (OSI) Reference Model.

· Electronic Industries Alliance (EIA; previously Electronic Industries Association): Founded in 1924 as the Radio Manufacturers Association, the EIA is a trade organization representing the interests of U.S. electronics manufacturers. The EIA assists in the development of physical layer interfaces, including optoelectric interfaces and test procedures for SONET. The EIA is best known for its physical layer specification EIA-232, nee RS-232.

· Exchange Carriers Standards Association (ECSA): Formed in 1984, ECSA represents the interests of the U.S. IntereXchange Carriers (IXCs). The ECSA T1 committee addresses issues of functionality and characteristics of intercon-nection and interoperability. The T1X1 committee addresses issues of digital hierarchy and synchronization.

· International Organization for Standardization (ISO): An organization comprising the national standards organizations (e.g., ANSI) of the various nations and regions, ISO heavily influences international standards set by the ITU–T. In the context of telecommunications, ISO is perhaps best known for its involvement in the Open Systems Interconnection (OSI) Reference Model.

· International Telecommunication Union—Radiocommunications Sector (ITU-R): The ITU–R is the sector that deals with issues of radio standards. The ITU–R parallels the ITU–T, which is responsible for the wireline domain.

· International Telecommunication Union-Telecommunications Standardization Sector (ITU-T): Previously the CCITT, the ITU–T is an agency of the United Nations. The ITU–T develops international standards in order to promote a world order of interconnectivity and interoperability in the wireline domain. ITU–T sometimes also is abbreviated as ITU–TS or ITU–TSS. The ITU–T predecessor organizations date to 1865.

· Telcordia Technologies: Previously Bellcore (Bell Communications Research), Telcordia was formed in 1984 under the terms of the Modified Final Judgement (MFJ), which forced AT&T to divest the Bell Operating Companies (BOCs). Bellcore was the research and development arm of the RBOCs, its client/ owners. Bellcore originally focused on standards development, test procedures, and Operations Support System (OSS) development, rather than the physical sciences. Bellcore was privatized and acquired by SAIC in 1998, as the interests of the RBOCs were no longer common in a deregulated, competitive environment. The name was changed to Telcordia Technologies in April 1999, with the stated focus of emerging technologies. Telcordia is now a private, stand–alone organization involved in the development of Operations Support Systems (OSSs) and network management software as well as consulting, testing services, and research services.

SONET defines the Synchronous Transport Signal level N (STS-N) as the electrical signal, which remains necessary until all–optical switching is developed. When converted to an optical signal for transport over a standard fiber-optic medium, the term Optical Carrier N (OC-N) is applied. As noted in Table 9.9, the basic building block of the digital hierarchy is OC–1 at 51.84 Mbps and currently tops out at the OC–768 rate of 40 Gbps. Notably, the various STS levels are considerate of the existing digital signal hierarchy, thereby achieving backward compatibility with legacy systems. In other words and by way of example, a T3 frame maps comfortably into an STS–1 signal that becomes an OC–1 frame. Similarly, multiple T1 frames can be aggregated to map into and form an STS–1 signal, which then becomes an OC-1 frame.

· Optical Carrier (OC) is the definition of the SONET optical signal. The fully defined OC levels begin at OC–1 (51.84 Mbps) and culminate in OC–768 (39.813 Gbps). All SONET/SDH levels are expressed in multiples of OC–1.

· Synchronous Transport Signal (STS) is the electrical equivalent of the SONET optical signal; it is known as Synchronous Transport Module (STM) in SDH. The signal begins in electrical format and converts to optical format for transmission over the SONET optical fiber facilities. Each STS–1 frame is transmitted in 125 μs, yielding raw bandwidth of 51.84 Mbps. The STS frame includes five elements:

· Synchronous Payload Envelope (SPE) carries the user payload data. It is analogous to the payload envelope of an X.25 packet. The SPE consists of 783 octets (87 columns and 9 rows of data octets).

· Transport Overhead (TO) consists of Section Overhead and Line Overhead:

· Section OverHead (SOH) of nine octets is dedicated to the transport of status, messages, and alarm indications for the maintenance of SONET links between ADMs.

· Line OverHead (LOH) of 18 bytes controls the reliable transport of payload data between any two network elements.

· Path Overhead (PO), contained within the SPE, comprises nine octets for the relay of OAM&P information in support of end–to–end network management.

· Payload is the actual data content of the SONET frame and rides within the SPE. Total usable payload at the OC–1 level consists of up to 49.54 Mbps, into which a T3 frame fits quite nicely. The balance of the 51.84 Mbps is consumed by Transport Overhead and Path Overhead.

Multiplexing is on the basis of direct TDM. Either full SONET speeds or lesser asynchronous and synchronous data streams can be multiplexed into the STS–N payload, which then converts into an OC–N payload. In other words, an appropriate combination of DS–0, DS–1, DS–2, and DS–3 signals can be multiplexed directly into an electrical STS–1 payload, which then converts to an optical OC–1 payload. The multiplexing process involves byte interleaving, much as described for traditional TDMs.

· Path-switched rings employ two active fibers. All traffic moves in both directions, thereby providing protection from network failure. A path–switched approach also improves error performance because the receiving stations examine both data streams, selecting the better signal. Should the primary ring fail, the secondary ring is already active in the reverse direction.

· Line-switched rings involve either two or four fibers. The single-ring implementation comprises two fibers, one of which is active and the other of which is held in reserve. Traffic moves in one direction across the active fiber. In the event of a network failure, the backup ring is activated to enable transmission in the reverse direction as well. A four–fiber implementation, which is preferred and typical in carrier–class networks, supports transmission in one direction over one fiber in each of the two rings. By way of example, fibers 1 in rings 1 and 2 might transmit in a clockwise direction, while fibers 2 in rings 1 and 2 transmit in a counterclockwise direction. The second fiber in each ring acts to support transmission in the reverse direction in the event of a failure in the primary ring. Line–switched rings smaller than 1200 km in route distance offer standard restoral intervals of 50 ms or less [73]. Larger rings, such as long–haul transport rings in IXC networks, involve longer restoral intervals due to the increased time of signal propagation.

As illustrated in Figure 9.16, a typical SONET network involves a number of rings operating at multiple speeds. The backbone, for example, might run at the OC–192 nominal rate of 10 Gbps, with subtended metro rings running at the OC–48 rate of 2.5 Gbps and further subtended local rings running at the OC–3 rate of 155Mbps.

SONET is capable of carrying information at Gbps speeds with excellent performance characteristics, including error performance and network management. The pipe also can carry any variety of asynchronous and synchronous information (e.g., voice, data, video, and image) and present it in a number of frame sizes. While the STS–1 frame is the basic building block, multiple STS–1 frames can be linked together in a process known as concatenation. Concatenated STS–Nc (N = Number; c = concatenated) signals are multiplexed, switched, and transported over the network as a single entity. This approach, which currently is defined for STS–3c and STS–12c, offers clear advantages where larger increments of bandwidth are required, because the overall level of overhead is reduced, thereby increasing the payload size. Example applications for concatenation include imaging and HDTV (High Definition TV). The SONET pipe consists of a path, virtual Tributaries, and tributary units:

· Path: A path is an end–to–end communications path (i.e., route or circuit) that carries traffic from one end point to another.

· Tributary Unit: A Tributary Unit (TU) is a VT along with a pointer that enables switching and cross-connecting.

The SPE, which contains the payload data, actually floats within the SONET frame. While SONET is a synchronized transmission system, with all devices relying on a common clocking signal, variations in clocking can occur. Those clocking variations can result from differences in local clocks, thermal expansion and contraction in individual fiber–optic cables, and other phenomena. Rather than buffering individual frames to effect synchronization, floating mode operation enables the network to adjust to frame float with the SPE identified by the pointer. The floating mode reduces cost and delay that would be caused by the use of buffers to synchronize each frame and SPE exactly [75].

The SONET overhead structure mirrors that of the existing digital carrier network for purposes of nonintrusive, end–to–end network management. Overhead layers include transport overhead (TO), which is further divided into section overhead, line overhead, and path overhead.

· 1 byte STS–1 signal ID

· 2 bytes Bit Interleaved Parity (BIP) for error monitoring

· 1 byte Orderwire (connection request)

· 3 bytes Data Communication Channel (DCC)

Line Overhead (LOH) of 18 bytes controls the reliable transport of payload data between major network elements. A Digital Cross–Connect System (DXC) performs Line Layer functions, including error control, switching and multiplexing, orderwire, express orderwire (priority connection request), automatic protection switching to back–up circuits, insertion of payload pointers, and synchronization control. The 18 bytes of LOH comprise

· 3 bytes STS-1 pointer

· 1 byte Bit Interleaved Parity (BIP) for error monitoring

· 2 bytes Automatic Protection Switching (APS)

· 9 bytes Data Communication Channel (DCC)

· 1 byte Orderwire

· 2 bytes reserved for future use

Path Overhead (POH) of 9 bytes comprises all aspects of end–to–end performance monitoring and statistical reporting. Path management is an essential responsibility of the Add/Drop Multiplexers (ADMs). Functions performed at the Path Layer include end–to–end performance monitoring, statistical reporting, STS mapping, and DS–to–OC mapping. The 9 bytes of POH comprise

· 1 byte trace

· 1 byte Bit Interleaved Parity (BIP) for error monitoring

· 1 byte Payload ID

· 1 byte maintenance status

· 1 byte user ID

· 1 byte frame alignment

· 3 bytes reserved for future use

The hardware aspects of SONET are difficult to describe in discrete terms because many of the functional elements overlap. Just as manufacturers of traditional voice and data equipment often build multiple functional elements under the skin of a single box, so do SONET/SDH equipment manufacturers. Given that caveat, the following is a description of the discrete functional elements of hardware devices:

· Concentrators perform the equivalent functions as traditional electrical concentrators and hubs. SONET concentrators combine multiple OC–3 and OC–12 interfaces into higher OC–N levels of transmission.

· Add/Drop Multiplexers do not have exact equivalents in the electrical (Digital Signal, or DS) world, although they perform roughly the same functions as traditional T–carrier TDMs. Generally found in the Central Office Exchange (COE), they provide the capability to insert or drop individual DS–1, DS–2, or DS–3 channels into a SONET transmission pipe. ADMs accomplish the process electrically, with the OC–N channel being converted prior to the process and reconverted subsequently. ADMs offer great advantage over traditional DS–N MUXs. The T–carrier approach, for example, requires that a DS–3 frame be demultiplexed into its DS–2 and then DS–1 frames, which must be broken down into 24 DS–0 channels in order to extract and route an individual channel. Once that is accomplished, the process must be reversed to reconstitute the DS–3, minus the extracted DS–0, and send it on its way. ADMs perform the additional functions of dynamic bandwidth allocation, providing operation and protection channels, optical hubbing, and ring protection.

SONET offers a number of advantages, in addition to the inherent advantages of fiber–optic transmission systems, in general. Certainly, the fact that SONET is highly standardized offers the benefits of interconnectivity and interoperability between equipment of different manufacturers. That standardization translates into freedom of vendor choice and yields lower costs through competition. Additionally, SONET/SDH is extendable to the premises on a fully interoperable basis. The increasing availability of SONET local loops provides end–to–end advantages of enhanced bandwidth, error performance, dynamic bandwidth allocation, and network management.

In bandwidth–intensive applications, the high absolute cost of SONET can be offset by virtue of its extraordinarily high capacity. Whether deployed in a carrier network or extended to the user premises, SONET supports the aggregation of all forms of traffic, including voice, data, video, image, facsimile, and multimedia. As a result, a SONET infrastructure can obviate the need for multiple transmission facilities in support of individual services. The simplicity of multiplexing and demultiplexing via ADMs serves to reduce costs, delay, and error. Clearly, SONET/SDH offers the considerable advantage of network resiliency through its inherent redundancy and self-healing capabilities.

Finally, SONET offers tremendous security, as with fiber-optic transmission systems, in general. Fiber is difficult, if not impossible, to physically tap without detection. Perhaps just as important is the fact that it is difficult to identify the one channel for detection out of the thousands of information channels supported in a SONET mode. As Francis Bacon said in 1625, "There is no secrecy comparable to celerity" (Of Delay, Essays)—celerity, or speed, is a hallmark of SONET.

SONET also has its disadvantages. As is true of many things, its greatest strengths are also its greatest weaknesses. SONET's redundancy, resiliency, and strong network management all come at a price—hence, SONET's legendarily high cost. Further, SONET's network management strengths translate into overhead intensity, which means that a good deal of raw bandwidth is consumed for management, rather than payload, purposes. SONET's redundancy also means that one–half to three–quarters of the available bandwidth is wasted at any given moment, assuming that the primary ring is operating properly. Finally, and perhaps most importantly, much of SONET's network management overhead and its careful timing mechanisms are considered to be wasted on packet data traffic, which neither expects nor appreciates its elegance.

SONET primarily is deployed in backbone carrier networks, where its many attributes can be put to full use. Particularly in a convergence scenario, the carriers have the potential to realize considerable cost savings by using a single–fiber infrastructure in support of bandwidth–intensive video and image streams, in addition to voice, facsimile, and voice–grade data traffic. The value of SONET as a backbone technology is ever more evident with the increasing penetration of broadband local loop technologies such as ADSL, cable modems, PON, and WiMAX. In a hybrid network configuration such as FTTN or FTTC, the fiber portion of the loop from the CO to the DLC may be SONET in nature.

It is interesting to track the developments in SONET speeds over the past decade or so using a few select examples. In 1994, Sprint announced the completion of the Silicon Valley Test Track, linking seven companies and several learning institutions to test ATM switching technologies. Plans called for the initial OC–1 ring to be upgraded to OC–3 in early 1995 and to OC–12 later in the year [76]. Shortly thereafter, Sprint deployed the first international SONET ring, linking New York City with three cities in Canada. The capacity of that OC–48 (2.5 Gbps) SONET ring increased to 10 Gbps through the use of Wave Division Multiplexing (WDM) equipment in order to derive four OC–48 channels [77]. In May 1996 AT&T announced plans to beef up its network with WDM equipment to realize 20 Gbps over OC–48 SONET fiber—an eightfold increase [78].

An interesting application of SONET technology is that of the 1996 Summer Olympics in Atlanta, Georgia. BellSouth, the incumbent LEC, deployed 40 OC–48 SONET rings to transport CD–quality voice, broadcast–quality video, and data simultaneously to nine cities in the Greater Atlanta area where the 26 various competitions took place. All information flowed through the SONET pipes, including competition results, accreditation, data, voice, and video. The uncompressed digital video signals traveled to the International Broadcast Center, where the signals were sold to other broadcasters to produce 40 worldwide video satellite feeds for approximately 3000 hours of Olympic coverage [79].

End–user organizations can gain the advantage of SONET local loops with increasing ease through a number of service providers. Particularly where data traffic requirements are significant, SONET offers the advantage of direct and seamless high–speed access to the SONET backbones. For high–speed T3 (45 Mbps) access to Frame Relay networks, SONET local loops have become increasingly advantageous. For ATM at speeds of 155 Mbps and 622 Mbps, SONET access is virtually a requirement, as T1 and even T3 do not address these levels of bandwidth.

SONET also finds application in campus environments, including not only institutions of higher learning but also business campuses. In such an environment, significant volumes of data and image information often are transmitted between buildings, perhaps between mainframes, Gigabit Ethernet switches, or routers. In such applications, optical fiber makes sense (in fact, it may be the only viable solution), and SONET often makes even more sense. Mission–critical environments, such as airports, find the redundancy and resulting resiliency of a SONET infrastructure particularly attractive.

Something of a debate continues to swirl around the subject of SONET versus Wavelength Division Multiplexing (WDM) and, particularly, Dense Wavelength Division Multiplexing (DWDM). The debate centers on issues of cost and complexity; SONET is relatively expensive, overhead intensive, and complex.

Finally, the nature of the multiplexers differs greatly in WDM and SONET. SONET makes use of very sophisticated and expensive ADMs, which can deal with SONET frames at the STS–1 level, VTs within VTGs, or even individual VCs. Some WDM systems, on the other hand, make use of Optical Add/Drop Multiplexers (OADMs), which multiplex at the lambda level, as illustrated in Figure 9.19. The OADMs also effectively perform a process of optical switching, or photonic switching, at the lambda level, thereby obviating the requirement for complex and expen-sive SONET DXCs.

Some suggest that DWDM, rather than SONET, is the future of optical transmission. Others, including myself, feel that this is not so. Rather, SONET and DWDM will coexist quite nicely into the future. Those promoting DWDM at the expense of SONET assume that a user or carrier can make effective use of a full lambda. Certainly, a large user organization may be able to make effective use of an OC–1 (52Mbps), OC-3 (155Mbps), or even OC-12 (622Mbps) lightstream for access or transport purposes, perhaps in support of integrated voice, data, and video over ATM. But carriers cannot cost effectively dedicate a full lightstream for that purpose when they can run it at OC–192 rates of 10 Mbps in support of a large number of end users and a great variety of native data types. Because SONET is TDM based, it can handle any type of data; in fact and as previously discussed, SONET can support multiple data types over the same facility. Rather, the long–haul carriers typically will use SONET at the access level. For long–haul transport, they will run SONET transparently inside each DWDM wavelength, thereby realizing the advantages of each. The SONET frame format will be supported inside DWDM, EDFAs will be used in conjunction with Raman amplification, OADMs will switch signals at the optical level, and SONET devices will switch signals at the STS–1 level. An additional argument in support of this approach centers on the robust nature of SONET Operations, Administration, and Management (OA&M) features, on which the carriers rely greatly to manage and maintain their networks. Further, WDM is purely a point–to–point technology, with no inherent provisions for recovery in the event of a network failure; the self–healing feature of SONET is of great value in optical networking.

9.3 IEEE 802.17, RESILIENT PACKET RING

The IEEE 802.17 Resilient Packet Ring Working Group published both the base standard and the 802.17a amendment in 2004 and is continuing work on the 802.17b amendment for advanced bridging as well as a maintenance specification and a set of interpretations. Resilient Packet Ring (RPR) is a Medium Access Control (MAC) layer protocol that uses Ethernet switching and a dual counterrotating ring topology to optimize the transport of Ethernet/IP packet data traffic over optical fiber rings while maintaining the resiliency of SONET/SDH, but at a much reduced level of overhead. RPR is Layer 1 independent; therefore it can be implemented over existing SONET/SDH physical rings using Layer 1 only or it can run on a stand-alone basis.

RPR calls for dual counterrotating ringlets that interconnect nodes where data traffic is intended to drop. The stations connecting through the nodes generally are pure Ethernet, unless TDM circuit emulation or some other option has been added. RPR also uses statistical multiplexing, which allows bandwidth to be oversubscribed, while establishing Committed Information Rate (CIR) and peak–rate thresholds per application. The nodes negotiate bandwidth requirements among themselves based on fairness algorithms and in consideration of a classification scheme that recognizes and provides higher priority access to traffic sensitive to latency and jitter. At the same time, best effort traffic, such as Internet traffic, is ensured equal access and a fair share of the remaining bandwidth. RPR addresses QoS issues through the definition of three Class-of-Service (CoS) levels:

· Class A traffic is intolerant of latency and jitter. RPR addresses class A traffic through a high Committed Information Rate (CIR) that ensures the availability of an average suitable amount of bandwidth for high priority traffic. Class A traffic includes real–time voice and video.

· Class B is more tolerant of latency and jitter, yet still has QoS requirements in that regard. RPR addresses class B traffic either through a lower CIR that ensures the availability of an average amount of bandwidth suitable for medium-priority applications that have less stringent QoS requirements or through an Excess Information Rate (EIR) option. In the event of network congestion, class B traffic is subject to fairness–based flow control. Class B is intended for business-class data traffic such as transaction processing.

The RPR protection scheme uses Physical Layer (Layer 1) alarm information and Data Link Layer (Layer 2) communications to detect node and link failures. Once a failure is detected, the RPR switching mechanism can restore the network in 50 ms or less, which is the SONET/SDH benchmark. There are two restoral mechanisms:

· Wrap: The wrap option calls for data to travel around the ring until it reaches the node nearest the break, as illustrated in Figure 9.20. That node turns the traffic around and sends it in the reverse direction over the counterrotating ring.

· Steer: Using the steer option, the originating station intelligently places the traffic on the ring that retains continuity.

The RPR dual counterrotating ringlets design calls for packet data traffic to travel over both fibers (one in each direction) simultaneously, which is much more efficient than the SONET approach. RPR employs statistical packet multiplexing to optimize the ring for packet data transport. This is unlike SONET/SDH, which employs TDM and, therefore, is optimized for PCM–based voice traffic. RPR further increases efficiency through a spatial reuse mechanism. Older LAN ring protocols such as Token Ring and FDDI require a data packet to traverse the entire ring, even if the destination node is only a single hop away from the originating node. RPR allows data inserted into the ring to be sent directly to the destination node, where it drops off and makes the bandwidth once again available to other stations in the forward direction.

A Metropolitan Area Network (MAN) infrastructure technology, RPR is largely oriented toward business–class data applications, with emphasis on Ethernet. Notably, however, RPR traffic classes provide sufficient Grade-of-Service (GoS) distinctions to make it viable in appropriate triple–play implementations [86–89].

Chapter 10: Broadband Network Services

Edward R. Murrow (1908-1965), U.S. broadcaster and journalist

The concept of broadband networking has its roots in the early 1970s when the CCITT (now ITU-T) first defined Broadband ISDN in anticipation of the development of highly bandwidth-intensive applications and the demand for them. Since that time, the demand for computer internetworking grew at rates that astounded even the most astute computer and data network pundits. Imaging systems developed and increasingly required networking. The potential for videoconferencing applications became apparent. The development and commercialization of the Internet and the Web truly has been mind-boggling. Finally, entertainment networking now captures the attention of telcos and others as they compete to develop networks that will carry television and movies as well as voice, data, video, image, facsimile, and all other forms of data over a single network infrastructure. These applications—and more—are the drivers of technology.

(Source: Frame Relay Forum.)

Figure 10.1: Fast packet services networking

This chapter concludes with a brief exploration of B-ISDN and Advanced Intelligent Networks (AINs). AINs reflect the marriage of computer, database, and network technologies, offering tremendous potential in the creation and delivery of services.

Gunpowder, since its invention, has been used for signaling, and still is employed in rockets, flares and railroad-track torpedoes. Cannons, since their invention, have been fired as warnings or to announce events. … The greatest distance covered by cannon signals was 428 miles from Buffalo to New York City when the Erie Canal was opened, October 26, 1825. Relays of guns got the message through in an hour and twenty minutes.

Telephone Almanac for 1956, Bell Telephone System

Relays of guns may have been effective in 1825 but were obsoleted by the telegraph beginning in 1844. Frame Relay experienced unprecedented growth for at least a decade from the time of its commercial introduction by Wiltel (since acquired by LDDS Worldcom, which was acquired by MCI, which is now part of Verizon) in 1992, before suffering from competition against Virtual Private Networks (VPNs) based on Internet Protocol (IP). Clearly, users in the United States have had the greatest appetite for Frame Relay, although it also enjoyed great popularity in the European Union (EU) and remains widely available at an international level.

Very much analogous to a streamlined and supercharged form of X.25 packet switching, Frame Relay can set up and tear down calls with control packets although no carrier ever activated these features for Switched Virtual Circuits (SVCs) in a public network. Frame Relay also forwards packets of data in the form of frames. Similar to X.25, Frame Relay statistically multiplexes frames of data over a shared network of virtual circuits for maximum network efficiency. The interface is in a Frame Relay Access Device (FRAD) that can be implemented on the customer premises and which is analogous to an X.25 Packet Assembler/Disassembler (PAD). Like X.25, Frame Relay is intended for bursty data traffic, although it works quite well with fixed-bit-rate applications. While both X.25 and Frame Relay can support voice, video, and audio, the inherently unpredictable levels of packet delay and packet loss over such a highly shared network can yield results that are less than totally satisfactory.

Frame Relay standards specify a frame payload that is variable in length up to 4096 octets, which supports a payload as large as that of 4-Mbps Token Ring, which generally is considered obsolete. Subsequently, the Frame Relay Forum (FRF) developed an Implementation Agreement (IA) that sets the minimum size at 1600 octets for purposes of interconnectivity and interoperability of FR networks and the switches, routers, and other devices that form the heart of the networks. This frame size easily supports the largest standard 802.3 Ethernet frame of 1518 octets. As Ethernet clearly is the dominant LAN standard and as FR was designed for LAN-to-LAN internetworking, this frame size generally is adequate. Regardless of the minimum frame size, the frames vary in length, just as do LAN frames. This variability results in some level of unpredictability, which contributes to congestion.

A wide range of manufacturers and carriers, both domestic and international, support Frame Relay. Standards bodies include the American National Standards Institute (ANSI), European Telecommunications Standards Institute (ETSI), and ITU-T. Table 10.2 notes select relevant standards.

The Frame Relay Forum (now merged with the ATM Forum and MPLS Forum to form the MFA Forum), a voluntary group of manufacturers and other interested parties, developed and promoted Implementation Agreements (IAs), which address manufacturer interoperability issues. IAs include the following:

· FRF.1.2: User-to-Network Interface (UNI) Implementation Agreement (April 2000)

· FRF.2.1: Frame Relay Network-to-Network Interface (NNI) Implementation Agreement (July 1995)

· FRF.3.2: Multiprotocol Encapsulation Implementation Agreement (MEI) (April 2000)

· FRF.5: Frame Relay/ATM PVC Network Interworking Implementation (December 1994)

· FRF.6.1: Frame Relay Service Customer Network Management Implementation Agreement (MIB) (September 2002)

· FRF.7: Frame Relay PVC Multicast Service and Protocol Description (October 1994)

· FRF.8.2: Frame Relay/ATM PVC Service Interworking Implementation Agreement (February 2004)

· FRF.9: Data Compression Over Frame Relay Implementation Agreement (January 1996)

· FRF.10.1: Frame Relay Network-to-Network SVC Implementation Agreement (September 1996)

· FRF.11.1: Voice over Frame Relay Implementation Agreement (May 1997, Annex J added March 1999)

· FRF.12: Frame Relay Fragmentation Implementation Agreement (December 1997)

· FRF.13: Service Level Definitions Implementation Agreement (August 1998)

· FRF.14: Physical Layer Interface Implementation Agreement (December 1998)

· FRF.15: End-to-End Multilink Frame Relay Implementation Agreement (August 1999)

· FRF.16.1: Multilink Frame Relay UNI/NNI Implementation Agreement (May 2002)

· FRF.17: Frame Relay Privacy Implementation Agreement (January 2000)

· FRF.18: Network-to-Network FR/ATM SVC Service Interworking Implementation Agreement (April 2000)

· FRF.19: Frame Relay Operations, Administration and Maintenance Implementation Agreement (March 2001)

· FRF.20: Frame Relay IP Header Compression Implementation Agreement (June 2001)

Customer DCE is the form of a Frame Relay Access Device (FRAD). The FRAD can be a stand-alone device that performs the sole function of assembling and disassembling frames, although it generally is embedded under the skin of another device, such as a router or Channel Service Unit (CSU). From the FRAD, access is gained to the link and, subsequently, to the Frame Relay network node at the edge of the carrier network. At the node resides the Frame Relay Network Device (FRND). The User Network Interface (UNI), as defined by ANSI and ITU-T, defines the nature of this access interface. Many service providers also offer access to Frame Relay on a dial-up modem basis, in which case the FRAD resides at the edge of the carrier network.

The Frame Relay network (Figure 10.2) consists of specified network interfaces in the form of the User Network Interface (UNI) and Network-to-Network Interface (NNI). The specifics of the internal carrier network are based on ISDN, generally making use of Permanent Virtual Circuits (PVCs) between ATM switches. Some carriers also support Switched Virtual Circuits (SVCs), although it is extremely unusual. Frame Relay networks can be public, private, or hybrid.

· Network-to-Network Interface (NNI) is defined as the interface between FRNDs in different Frame Relay networks and is based on multinetwork PVCs. Most network-to-network connections are provided over digital SONET/SDH optical fiber trunks and through ATM switches. This current method of internetwork connection has implications relative to Implicit Congestion Notification (ICN) and network management. An NNI is required, for example, if a user organization wishes to connect from an intraLATA Frame Relay network provided by a Regional Bell Operating Company (RBOC) to an interLATA network pro-vided by a long-haul carrier. The NNI is provided at additional cost.

· Integrated Services Digital Network (ISDN) is the basis of the internal Frame Relay network, as the LAP-F protocol that governs the user links is an adaptation of the ISDN LAP-D protocol. The links generally are not ISDN, however, for reasons that include lack of availability, lack of capacity, and additional cost.

· Permanent Virtual Circuits (PVCs) define fixed paths through the network for each source-destination pair. The network administrators permanently fix the path definitions in network routing tables, until such time as they permanently change them. As bandwidth over the physical path is used only when required, the circuit is virtual in nature. As Frame Relay is a shared packet network, a great many source-destination pairs may share each PVC. Once defined and regardless of the network traffic load, the PVC is always used to serve a given source-destination pair. In order to protect against a catastrophic failure affecting a PVC element (e.g., link or switch), a backup PVC can be provisioned, at additional cost, over a separate physical path. Frame Relay networks generally are based on PVCs, as depicted in Figure 10.2. PVCs typically are fully symmetric, with equal bandwidth in each direction. They also may be asymmetric, with more bandwidth in one direction than in the other.

· Mesh networking is easily accomplished with Frame Relay, and on a scalable basis. In other words, the cost of Frame Relay implementation is relatively proportionate to the task at hand. As more sites are added to the network, more FRADs, access links, and FRNDs can be added. As the bandwidth requirement increases at any given site, the capacity of the access link can be increased. The costs grow relatively gracefully as the network expands to provide access to additional sites and as bandwidth increases on a site-by-site basis. While full mesh networking is unusual, it can be accomplished at significantly lower cost than with a private, leased-line network, which requires a separate access link at every site for every other site. Frame Relay networks can be classified as VPNs, as they are virtually equivalent to a private network, at least in many respects.

· Network processing is not performed in Frame Relay, at least not to the extent experienced in X.25 networks. A Frame Relay network assumes that the link is error free. As the link is digital (ideally fiber optic) and based on ISDN, there is no compelling requirement to check for errors, although every switch does so. Once the frame is identified as valid and the address is recognized, the frame is relayed through the core of the network to the next node and, ultimately, to the end user. If a frame is errored or somehow mutilated, Frame Relay simply discards it, making no attempt to inform the endpoints. Frame Relay does not support sequence numbers and sends no acknowledgments to endpoints. This is the equivalent of removing OSI Layer 3 functions from the X.25 model. When compared with X.25 and traditional packet switching, this reduction in network processing yields lower cost and reduced packet latency, somewhat counterbalancing the effect of the variable-size frame to reduce issues of overall network congestion.

Frame Relay, as a service offering, depends on certain hardware and firmware in order to accomplish the interface. That equipment includes the FRAD, the FRND, and the FR switch:

· Frame Relay Network Devices (FRNDs) are Frame Relay switches at the network side of the connection. The FRND (pronounced friend) is provided by a friendly service provider, which seems like an amicable arrangement. Frame Relay switches are nodal processors capable of switching frames at very high speed. They contain buffers for flow control purposes. Within limits, the buffers can absorb incoming frames until the switch can act on them as well as absorb outgoing frames until the forward link becomes available. The switches contain very high speed switching matrixes and internal buses. They have sufficient intelligence, in the form of routing tables, to read the control information embedded in the frame header and route the frames correctly over the PVC identified previously in the call set-up process. The nodal processors also have intelligence sufficient to check for errors in the frame. Additionally, the nodal processors may have the intelligence to prioritize certain types of delay-sensitive traffic, such as SDLC and real-time uncompressed voice and video. Since the switches perform no error correction or protocol conversion functions, they act on the frames very quickly, thereby minimizing latency. Frame Relay nodal processors can be managed from a centralized Network Operations Center (NOC), thereby enhancing the scalability of the network [6]. The switch also may contain a voice compression module. While Frame Relay is intended for LAN internetworking, it can accommodate voice with varying degrees of quality.

· Flag Field: One-octet, fixed binary sequence (01111110) is employed to identify and separate frames. Flag fields appear at the beginning and end of the frame. The flags technically are not part of the frame. Rather, they are frame delimiters that precede and follow all frames. The specific bit sequence is recognized by all network devices as a delimiter and nothing more. Essentially, they are much like idle bits or keep-alive bits. When a network device sees any other bit pattern, it recognizes those bits as being something of significance, rather than a flag field. There may be only a single flag between frames sent in rapid succession or there may be many flags between frames sent at a slow pace.

· Address Field: Two octets are used as a default length, although some networks use four octets. Address fields include the necessary control information in the data link connection identifier. The address field also contains a command/ response field, address field extension, forward and backward explicit congestion notification fields, and discard eligibility data.

· Data Link Connection Identifier (DLCI, pronounced delsey): Ten bits that identify the data link, the virtual circuit (i.e., PVC or SVC), and its service parameters to the network. Those service parameters include frame size, Committed Information Rate (CIR), Committed Burst Size (Bc), Burst Excess Size (Be), and Committed Rate Measurement Interval (Tc). I discuss the significance of these service parameters later in this chapter.

· Command/Response (C/R): One bit reserved for use of the FRADs, rather than the Frame Relay network. C/R is defined to facilitate the transport of polled protocols such as Systems Network Architecture (SNA), which require a C/R for signaling and control purposes.

· Address Field Extension (EA): Two bits that signal the extension of the addressing structure beyond the two-octet default. The use of EA must be negotiated with the carrier when the service is established.

· Forward Explicit Congestion Notification (FECN, pronounced feckon): One-bit field available to the network to advise upstream devices that the frame has experienced congestion. The FRAD clearly recognizes when the frame carrying the FECN survives. It also is advised that subsequent frames may not be so fortunate. Should subsequent frames be discarded or corrupted in transmission, the receiving device is advised that recovery may be required in the form of requests for retransmission. If the upstream device controls the rate of upstream data transfer, it has the opportunity to throttle back.

· Discard Eligibility (DE): One-bit field indicating the eligibility of the frame for discard under conditions of network congestion. Theoretically, the user equipment sets the DE in consideration of the acceptability of the application to packet loss. Should the user equipment not set the DE, the network switches may do so on a random basis, with results that may be less than totally pleasing. SDLC, voice, and video traffic, for example, do not tolerate loss and so typically are engineered to avoid DE status.

· Information Field: Contains user information, either in the form of payload data or internetwork control information passed between devices such as routers. Although the information field may be 4096 octets in length, ANSI recommendations dictate a maximum size of 1600 octets. This payload size is more consistent with the frame size (64-1518 octets) of 802.3 Ethernet LANs, which generate the vast majority of LAN traffic. This frame size of 1600 octets also is addressed in Implementation Agreements from the Frame Relay Forum, which were developed to ensure the interconnectivity and interoperability of Frame Relay networks.

· Frame Check Sequence: Two-octet Cyclic Redundancy Check (CRC) supporting header error detection in frames with information fields up to 4096 octets in length. Note that no error correction is provided. Rather, error correction for payload data is accomplished by running the TCP/IP suite or some other error-correcting protocol within the information field as part of the payload.

The Local Management Interface (LMI) protocol provides operational support for the UNI. Originally defined by the Frame Relay Forum in 1990, it subsequently was adopted by ANSI and the ITU-T. The LMI is a polling protocol between the FRAD and the network, which periodically verifies the existence and availability of the PVC as well as the integrity of the UNI link.

Frame Relay, as a highly shared packet network, is extremely efficient. Since all the links and all the switches that comprise the network are shared among large numbers of users and user organizations, the network is subject to variable and somewhat unpredictable levels of congestion. Indeed, it is designed for congestion as a natural occurrence and in support of bursty LAN-to-LAN internetworking applications, which are tolerant of delay and loss and have the time and ability to adjust and recover. This fact serves to reduce the overall cost of the network service, although it compromises the level of performance. This is a classic network optimization scenario in which an appropriate balance is struck between cost and performance. Congestion management is addressed through the following parameters, specified in an addendum to ANSI T1.606 [6–8]:

· Committed Information Rate (CIR): The data rate that the network guaran-tees to handle across the Virtual Circuit (VC), which can be either a PVC or an SVC, under normal conditions over a period of time. While the standards do not specify the time period, the CIR typically is an average data rate over a period of a few seconds. The CIR is based on mutual agreement between the carrier and the customer and should be based on the average maximum amount of traffic required to be supported reliably over the VC during the busiest hour of the day. The VC is defined as a matched source-destination pair, which can be either symmetric or asymmetric in terms of bandwidth. In the event that the CIR is exceeded, the network reserves the option to mark excess frames as Discard Eligible (DE) if the DCE has not already done so. The marking function typically takes place in the entry node at the edge of the network, after which discarding is done by any switch that experiences congestion. In order to obviate any issues of unnecessary congestion in the network core, the first edge device may discard frames rather than just mark them. There is no model for the relationship between the CIR and Access Rate. Some carriers permit zero CIR, meaning that the CIR is set at 0 percent. At zero CIR, all Offered Load is handled on a best effort basis, with absolutely no commitments or guarantees. Some carriers permit, or even require, the CIR to be set at 100 percent, meaning that all offered load will be guaranteed, with the maximum offered load determined by the Access Rate and Port Speed. Some carriers require that the CIR for a VC be set at no less than 50 percent of the Access Rate and Port Speed, with that relationship capped at T1/E-1 speeds—at E-3 levels of 34 Mbps and certainly at T3 levels of 45 Mbps, the bursts are too wildly variable to be managed effectively. Significantly, multiple VCs can be supported over a single access line.

· Offered Load: The data rate offered to the network for delivery measured in bps. The aggregate offered load can be less than the Access Rate supported by the access link and/or the port speed of the FRND but can never exceed whichever is less. As is always the case, the lowest common denominator defines the maximum level of performance.

· Committed Burst Size (B c): Maximum amount of data that the carrier agrees to handle without discard under normal conditions. The Bc and line rate affect the calculation of the CIR for a VC.

· Excess Burst Size (B e): Maximum amount of data that the network will accept in a block from a user without discard if bandwidth is available and over a specified time (T). In recognition of the bursty nature of LAN-to-LAN communications, the transmitting device may burst above the CIR/ Bc for a brief period of time; the network attempts to accommodate such bursts within limits of burst size and burst interval. The network reserves the option to mark the excess data above Bc as Discard Eligible (DE) should the user CPE not have done so already. The last parts of larger bursts usually are discarded.

Additional congestion management terminology includes the following:

· Discard Eligibility (DE): Indicates the eligibility of the frame for discard, in the event of congestion. Either the FRAD or the FRND may set discard eligibility. Theoretically, at least, the FRAD is programmed to recognize when the CIR is exceeded and to volunteer frames for discard should the network suffer congestion. The FRAD must make DE decisions in consideration of the user-layer application's tolerance for frame loss. As SDLC data, voice, and video, for example, are not tolerant of loss, such data must be confined to the CIR and must not be marked DE. Ethernet LAN data, on the other hand, are tolerant of loss; such data can exceed the CIR for the VC and can be marked DE without compromising the best effort expectations of the user-layer application.

· Explicit Congestion Notification (ECN): The means by which the network advises devices of network congestion. Forward Explicit Congestion Notification (FECN) advises the target device of network congestion so it can adjust its expectations. Backward Explicit Congestion Notification (BECN) advises the transmitting device of network congestion so it can reduce its rate of transmission accordingly. Theoretically, it is the responsibility of the various devices on the originating end of the data communication to adjust in some way. While not all FRADs have the ability to adjust, high-end routers can impose a certain amount of flow control through buffering and withholding acknowledgments to sources. The switches and hubs behind the FRAD have no ability to adjust other than filling their buffers, which typically are highly limited in capacity, if they exist at all. The originating terminal devices and peripherals have no ability to adjust, other than filling their very limited buffers, if any, and then simply stopping. If the devices run TCP/IP, then TCP slows down when it detects a lost frame.

· Implicit Congestion Notification: Inference by user equipment that congestion has occurred. Such inference is triggered by realization of the user device (e.g., FRAD, mainframe, or server) that one or more frames have been lost. Based on control mechanisms at the upper protocol layers of the end devices, the frames are resent to recover from such loss.

Aside from the congestion mechanisms specified in the frame standards, there are two types of congestion control algorithms. Open-loop algorithms permit the acceptance of the frames with no prior knowledge of the likelihood of their successful delivery. Closed-loop algorithms prevent frames from entering the network unless there is an extremely high probability of their being accepted, transported, and delivered without discard. The closed-loop algorithm fairly allocates backbone trunk bandwidth among all the PVCs configured on a particular trunk and in proportion to the CIRs [9]. Closed-loop networks clearly are more sophisticated and offer better performance, although at some additional cost.

In addition to public, carrier-based Frame Relay services, a number of end-user organizations run the Frame Relay protocol over their legacy T/E-carrier, leased-line private networks. This approach takes advantage of both the performance characteristics of the legacy network and the protocol independence of Frame Relay, thereby enabling the sharing of the existing infrastructure for a wide variety of data traffic (e.g., SDLC, LAN-to-LAN, and IP-based traffic). As the leased-line network is private and dedicated, rather than public and shared, issues of congestion management are entirely under the control of the user organization. Therefore, desired performance levels are easily established and controlled. In such an environment, issues of congestion that might affect voice and video in a negative way are eliminated. Virtually all the major manufacturers of T/E-carrier nodal MUXs support Frame Relay through special interface cards.

The very concept of transmitting voice over any sort of data network, much less a packet data network, would have been viewed as completely bizarre not so long ago. Voice over packet data networks includes Voice over Frame Relay (VoFR), Voice over Internet Protocol (VoIP), and Voice over Asynchronous Transfer Mode (VoATM). Frame Relay certainly was not originally designed for voice traffic, and neither was I P, although ATM was designed from the ground up for all types of traffic. For the moment, let us focus on VoFR, as specified in the Frame Relay Forum's FRF11.1, Voice over Frame Relay Implementation Agreement (May 1997).

While Frame Relay is intended as a data networking service, specifically and originally for LAN-to-LAN internetworking, it also can be used in support of isochronous voice and voice-band data (i.e., fax and analog modem) traffic. But you must consider the fact that issues of delay can affect the quality of the voice transmission over time. Even within the constraints of the CIR, the level of bandwidth is measured and ensured only as an average over a period of time, rather than on an instantaneous basis. Because Frame Relay is a packet-switching, rather than a circuit-switching, network service technology, there is no provision for temporary, continuous, and exclusive bandwidth through a Frame Relay network. As a result, voice frames suffer from delay, loss, and error at levels that are variable and unpredictable in nature. The result can be less than aesthetically pleasing. In the event that the received voice stream becomes unintelligible, the listener always can resort to the Huh? protocol, as in "Huh? What did you say?"

Actually, it turns out that there might be another reason, or at least an additional reason. A number of major domestic and international carriers offer, and even promote, VoFR as part of an integrated network solution for the data-heavy enterprise. Those carriers particularly promote VoFR as part of a managed-service offering. Such an offering involves the carrier's managing the service and equipment from end to end, thereby ensuring its ability to strike the proper network design to support business class voice over a packet data network. That design may include separate VoFR PVCs and adequately sized (i.e., oversized) CIRs. Ultimately, however, the only reason to consider VoFR is that it can ride free within the CIR limits of a Frame Relay network intended for LAN-to-LAN communications. Ironically, people hesitant to put voice on Frame Relay eagerly put voice on IP when that VoIP packet may very well be encapsulated in Frame Relay and transmitted over an ATM network.

In consideration of the fact that the voice information stream is not constant and certainly is not constantly changing at a high rate of speed, VoFR can take advantage of various compression techniques to relax the amount of bandwidth required for the voice stream. Specifically, only the changes in speech patterns from one set of samples to another need to transmit across the network. The natural and predictable pauses in human speech, for example, can be noted, eliminated on the transmit side, and reinserted on the receive side through a process known as silence suppression. Similarly, redundant or repetitive speech sounds can be removed on the transmit side and reinserted on the receive side. Once the silence is suppressed and the redundancy noted, the remainder of the set of voice samples can be further compressed and inserted into relatively small frames. The small frame size allows a number of sound bytes to be presented to and delivered through the network on a frequent basis. While this approach is somewhat overhead intensive, it provides for a relatively normal flow of speech information. There exist a number of low-bit-rate compression algorithms, both standard and proprietary, that are implemented by various router manufacturers. The most popular are those in the CELP (Code-Excited Linear Prediction) family. CELP and other compression algorithms support very reasonable business-quality voice (under conditions of low network congestion or where the voice traffic remains within its CIR) at bit rates as low as 8 kbps. Additionally, the Frame Relay Forum adopted Conjugate Structure-Algebraic Code-Excited Linear Prediction (CS-ACELP) as FRF.11. According to the Frame Relay Forum, only 22 percent of normal speech comprises essential components, with the balance comprising either pauses or repetitive speech patterns [10].

Other, nonstandard compression algorithms take the bit rate as low as 4.8 kbps (or lower), but with considerable loss of quality. Current ITU-T standards for voice compression used in various VoFR codecs include those presented in Table 10.3, all of which begin with PCM voice. In addition to the base compression algorithms, various continuity algorithms are employed to intelligently fill the void of missing or errored compressed voice frames by stretching the previous voice frames or blending several frames together.

· Opinion Scale: conversation test (bad to excellent), difficulty scale (yes or no)

· Opinion Scale: listening test (bad to excellent), listening effort scale ("No meaning understood with any reasonable effort" to "complete relaxation possible, no effort required")

· Loudness: preference scale ("Much quieter than preferred" to "Much louder than preferred").

An MOS of 4.0 is considered to be toll quality. While P.800 is an effective means of quantifying such a highly subjective perception, it is expensive and time consuming. In the mid-1990s, therefore, the ITU-T began the process of automating the objective measurement and testing of end-to-end voice quality across both circuit-switched and packet-switched networks. The strongest of those techniques include the following:

· Perceptual Analysis/Measurement System (PAMS)

· Perceptual Speech Quality Measurement (PSQM), standardized as P.861

· Perceptual Evaluation of Speech Quality (PESQ), standardized as P.862

· Single Ended Method for Objective Speech Quality Assessment in Narrow-Band Telephony Applications, standardized as P.563

Regardless of the compression technique employed, VoFR suffers from some additional loss of quality due to issues of echo, delay, delay variability (i.e., jitter), and loss. Through network buffering and voice frame prioritization, these factors can at least be mitigated but never entirely be overcome. Additionally, some FRAD manufacturers support the definition of a separate PVC for delay-sensitive traffic such as voice, while others support multiplexing of voice data frames only over the same PVC as that used in support of data. Some manufacturers support both small frame sizes for voice and large frame sizes for data, while others force all data into the smaller frame size more appropriate for voice; the latter approach forces all data into the less efficient (more overhead intensive) frame size. Notably, VoFR FRADs are proprietary in nature, thereby requiring a single-vendor VoFR network [12–20].

In order to mitigate the inherent difficulties of transmitting delay-sensitive voice traffic over Frame Relay, some manufacturers and carriers offer various priority management techniques. Several manufacturers of multifunction FRADs and routers, for example, support the identification of high-priority traffic in the frame header. Some service providers also offer PVCs of varying levels of delay/priority, usually by mapping the Frame Relay connection to an ATM connection with these properties. Priority levels generally are defined as follows [21]:

· Real-Time Variable Frame Rate: Top priority; suited to delay-sensitive, mission-critical applications such as voice and SNA; premium pricing

· Available/Unspecified Frame Rate: Low-priority designation; suited to Internet access, e-mail, file transfer, monitoring, and other low-priority applications that can tolerate relatively significant levels of delay

The costs of Frame Relay vary widely among equipment manufacturers and carriers. It is clear, however, that Frame Relay is highly cost effective for internetworking large numbers of sites at rates up to 44.736 Mbps (T3). Such a network can be configured as a full mesh, a partial mesh, or a star.

Frame Relay networks originally were priced on a case-by-case basis. In consideration of the widespread popularity of the service, however, in October 1995 the Federal Communications Commission (FCC) classified Frame Relay as a basic service in the United States. This reclassification required that the major carriers file tariffs that included rate structures, which enabled smaller users to comparison shop, which simplified the comparison process and likely reduced their costs. Larger users, on the other hand, lost some bargaining power.

Prices have varied considerably in the United States and worldwide since the FCC decision. In the highly competitive Frame Relay market, mesh or even partial-mesh networking can be accomplished for far less than the cost of a comparable FT1 network. Costs can be mitigated to some extent through the use of compression devices, which permit transmission at lower port speeds and CIRs through the recognition of patterns within data frames intended for multicasting or transmission to multiple addresses.

FR unique characteristics offer some advantages over its predecessor technologies of X.25, Switched 56/64 kbps, and leased lines. Additionally, Frame Relay is widely available both domestically and internationally. Also, the market is highly competitive, which puts it in an advantageous position relative to ATM. However, the recent market shift toward IP has caused not only a slowing in Frame Relay demand but actually a decline in Frame Relay service penetration. The International Data Corporation (IDC) estimated that there were approximately 1.3 million Frame Relay ports in the United States at the end of 2005, but that number was expected to drop to 1.2 million in 2006 and fall to 751,000 in 2008. The IDC reports that not only is Frame Relay decreasing but also carriers are now beginning to eliminate it completely in favor of IP-based VPNs [23].

Advantages of Frame Relay include its excellent support for bandwidth-intensive data and image traffic. Even stream-oriented traffic such as real-time voice and video can be supported with reasonable quality. The absolute speed of Frame Relay, its improved congestion control, and reduced latency certainly are improvements over X.25 networks. Because Frame Relay is protocol insensitive, it can carry virtually any form of data in variable-size frames. Bandwidth on demand, within the limit of the access line, is provided, generally with reasonable costs for high-speed bursts. Additionally, the network is highly redundant, thereby providing improved network resiliency. SVCs are especially resilient, although few carriers support them.

Disadvantages of Frame Relay include its latency, which is inherent in any highly shared data-oriented VPN, including X.25 and IP-based networks. Although Frame Relay's latency level is an improvement over X.25, it does not compare well with ATM. The latency issue can be significant when Frame Relay is used in support of SNA traffic, although FRADs can spoof SDLC. The latency issue generally renders it largely unsuitable for highly intensive voice and video, at least when measured in terms of traditional toll quality, although priority mechanisms go a long way toward resolving that issue.

Frame Relay also offers no inherent mechanisms for frame sequencing and error correction for the user payload, unlike its predecessor X.25, which performed these on a hop-by-hop basis at the network nodes. As Frame Relay assumes that these functions are accomplished at higher layers in the end-user domain, Frame Relay is relieved of these responsibilities. Stripping out these functions offers the advantage of reduced processing at the network nodes and, therefore, reduced latency, congestion, and cost. But the elimination of these functions also compromises the reliability of the data stream. In order to compensate for this shortcoming, data applications typically involve running TCP/IP inside the information field as part of the user payload. As a result, the error control is performed on an end-to-end basis in the CPE domain. Since TCP/IP commonly is used in support of LAN communications, this additional step and the additional overhead associated with TCP/IP is not considered particularly burdensome—after all, Frame Relay is intended for LAN-to-LAN internetworking. Where real-time voice and video applications are supported over Frame Relay, User Datagram Protocal (UDP) replaces TCP in the protocol mix because there is no time to recover from loss or error through retransmission, in any event.

Frame Relay was designed to fill the gap between packet-switched networks (X.25), circuit-switched networks (Switched 56/64 kbps), and dedicated data networks (DDS, T/E-carrier). It is intended for intensive data communications involving block-level communications of data and image information. Frame Relay supports voice and low-speed video, although the quality can be spotty due to intrinsic issues of frame latency, jitter, and loss in a poorly designed network.

Frame Relay applications primarily, therefore, are data or image in nature. LAN internetworking is the driving force behind Frame Relay, although controller-to-host, terminal-to-host, and host-to-host applications abound. The more recent availability of dial-up access also makes the service cost effective for bandwidth-intensive telecommuting application. Internet Service Providers (ISPs) and Internet backbone providers make significant use of Frame Relay, both for user access to the ISP and for backbone network application. In Internet applications (indeed, a great many data applications), the TCP/IP protocol suite runs inside Frame Relay in order to ensure the reliability of the data stream.

An excellent example of the application of Frame Relay is that of the airline reservation networks, all of which are transitioning from X.25. In 1995, Apollo Travel Services converted its reservations network to Frame Relay to connect approximately 15,000 travel agency workstations at 1000 sites. Owned by United Airlines, USAir, and Air Canada, Apollo expected response time to improve to 2s from the previous 4s. The network runs at 56 kbps, as opposed to X.25 speeds of 2400–4800 bps. Using AT&T's InterSpan Frame Relay Service, Apollo is able to establish end-to-end TCP/IP connections from each agency to the Apollo reservations center. The service also supports mesh networking between travel agencies, thereby eliminating the requirement for all traffic to go through the Apollo head end [26].

10.2 SWITCHED MULTIMEGABIT DATA SERVICE

Switched Multimegabit Data Service (SMDS) is an offshoot of the Distributed Queue Dual Bus (DQDB) technology defined by the IEEE 802.6 standard for Metropolitan Area Networks (MANs) as a means of extending the reach of the LAN across a metropolitan area. The original work on the DQDB concept was done at the University of Western Australia where it was known as Queued Packet Synchronous Exchange (QPSX). Subsequently, the original technology was licensed to QPSX Ltd., formed by the University of Western Australia and Telecom Australia [6]. The commercial success, although highly limited, of SMDS is attributable to further development work by Bellcore (now Telcordia Technologies) at the request of the RBOCs. Bell Atlantic (now part of Verizon) introduced the first commercial offering in 1992 at 5 Mbps based on a test implementation at Temple University [27].

Since then, SMDS enjoyed limited, short-lived success in the United States through deployment by most of the RBOCs and GTE (now part of Verizon). As the RBOCs and their subsidiary BOCs were primarily Local Exchange Carriers (LECs) generally limited to providing service within the confines of the LATA at the time, they exhibited particular interest in network technologies that were appropriate for providing service in confined geographic areas. The geographical limitation of SMDS, therefore, did not pose a practical limitation for the LECs.

SMDS was originally described by Bellcore (now Telcordia Technologies) as a high-speed, connectionless, public, packet-switching service that extends LAN-like performance beyond the subscriber's premises, across a metropolitan or wide area. SMDS is a MAN network service based on cell-switching technology. Generally delivered over a SONET/SDH ring, SMDS has a maximum effective serving radius of approximately 30 miles (50 km). SMDS is a connectionless service that accepts user data in the form of a Protocol Data Unit (PDU) up to 9188 octets in length over access lines of up to 45 Mbps (T3). The user data are segmented into 53-octet cells, 48 octets of which are payload and 5 octets of which are overhead. The segmentation process can occur either at the network node or in the user equipment.

SMDS is designed to support LAN-to-LAN traffic under the IEEE 802.6 standards, although other data applications are supported effectively. SMDS offers excellent performance characteristics, including a guaranteed rate of access, transport, and delivery. SMDS also provides a smooth migration path to ATM. In fact, the 53-octet cell format was chosen specifically for this reason.

SMDS also enjoyed moderate success in Western Europe, where the nations tend to be small in geographic terms and where the population density of large businesses is high in the major metropolitan areas. The European version of SMDS was known as CBDS (Connectionless Broadband Data Service). CBDS is available in Australia (Telecom Australia), Germany (DBP), and England (BT).

The history of the SMDS Interest Group (SIG) underscores the overall failure of SMDS. The SIG was a consortium of vendors and consultants that worked to advance SMDS as a standards-based, interoperable solution for high-performance data connectivity. On June 17, 1997, the board of trustees announced that its mission of advancing the cause of SMDS was fulfilled and the group was disbanded, turning all its responsibilities over to unnamed regional organizations. Essentially, the SIG lost the battle, declared victory, and everyone went home a winner. At this point, SMDS is essentially a historical footnote that promoted the cause of high-performance data communications and served as a technological precursor to ATM.

10.3 ASYNCHRONOUS TRANSFER MODE

It seems the only people using the machines were a small number of prostitutes and gamblers who didn't want to deal with tellers face to face.

Luther George Simjian (1905–1997), inventor of the Bankmatic Automatic Teller Machine, discussing the slow pace of its early adoption

The automatic teller machine (usually called ATM), also known by at least a dozen other names including cash machine, has little to do with Asynchronous Transfer Mode (ATM), except for the fact that the two technologies share a common acronym. Actually, there are some ATMs that interconnect over networks running the ATM protocol, but the real reason that I selected this quote is that I couldn't find any interesting—much less entertaining—quotes about the ATM protocol. This quote also illustrates the dangers of using an Internet search engine when researching a topic for a technical book. In the event that you are interested in the history of the ATM, aka hole-in-the-wall machine, Simjian's Bankmatic cash-dispensing machine was installed in a City Bank of New York branch in New York City in 1939. It was removed after only six months, due to lack of acceptance. The modern ATM was invented in 1963 or 1968, depending on which website you believe, by a trio of Docutel engineers, Don Wentzel, Tom Barnes, and George Chastain. The patent for that device was issued in 1973 and the first machine was installed in a New York Chemical Bank [28–30].

The first ATM network in the United States was a testbed gigabit network known as the National Research and Education Network (NREN). Sponsored by the U.S. Advanced Research Project Agency (ARPA) and the National Science Foundation (NSF), the project began in 1990. In Europe, a consortium of carriers, end users, and universities sponsored the Research for Advanced Communications in Europe (RACE) project 1022, which was initialized in 1987 to demonstrate the feasibility of ATM. The result of the RACE initiative was the R1022 ATM Technology Testbed (RATT). RACE project 2061, also known as EXPLOIT, is a more recent RACE project intended to prove the viability of Integrated Broadband Communications (IBC) in the European Community (EC). Advanced Communications Technologies and Services (ACTS) was developed as the successor program to RACE and continues that work on ATM networking and some 200 other projects.

Beginning in the early 1990s and for nearly a decade, ATM was highly touted as the ultimate network-switching solution, given its high speed, its ability to serve all information types, and its ability to guarantee each information type an appropriate Quality of Service (QoS). ATM was ultimately expected by many to replace all of the existing circuit-and packet-and frame-switching technologies currently in place, but that now appears highly unlikely. ATM remains widely used as a backbone switching technology, where it was heavily favored by the incumbent LECs and IXCs until just recently, when Internet Protocol (IP) and Multiprotocol Label Switching (MPLS) based networking found support. ATM remains favored in Asymmetric Digital Subscriber Line (ADSL) applications and as the Data Link Layer protocol for Broadband Passive Optical Network (BPON) and Gigabit PON (GPON) broadband local loop.

Asynchronous transfer mode is a fast-packet, connection-oriented, cell-switching technology for broadband signals. ATM is designed, from concept up, to accommodate any form of information—voice, facsimile, data, video, image, and multimedia—whether compressed or uncompressed, at broadband speeds. Further, all such data can be supported with a very small set of network protocols, regardless of whether the network is local, metropolitan, or wide area in nature. In terms of user access rates, ATM generally operates at minimum access speeds of DS-1 and DS-3. The OC-1 (51.84 Mbps) interface was defined but not commonly implemented, and OC-3 (155 Mbps) is not unusual. The backbone transmission rates are DS-3 and OC-1 at a minimum and generally OC-3 or better.

In any case, the data are presented to and accepted by the network on an asynchronous basis, and the ATM switches transfer the data from switch to switch in a hop-by-hop mode, hence the term asynchronous transfer mode. The optical fiber transmission facilities, of course, are tightly synchronized. The ATM switch and all other network elements are synchronized with the pipe as well.

The user data are sent to the network over a digital facility. At the workstation, router, or ATM switch, data are organized into 48-octet cells. Each cell is prepended with a header of five octets to create a cell of 53 octets. The cells are multiplexed and contend for access to a broadband facility that ideally is SONET in nature. In combination with SONET/SDH, ATM was intended to be the ultimate and complete integrated network solution, supporting a theoretically infinite range of services: B-ISDN.

The small cell size reaps several advantages. First, it can accommodate any form of data—digital voice, facsimile, data, video, and so on. Second, the fixed length of the cell offers the network switches the advantage of predictability, as compared to a variable-length frame. Third, the fixed cell size facilitates the implementation of switching functions in hardware (i.e., silicon), which enables processes to be accomplished at significantly greater speed than does software, especially if processing variable-size frames. These last two considerations yield decreased delay, as data move through the switching systems and across the transmission links in frequent little blasts. Long, and especially variably long, frames occupy the attention of the network for relatively long periods of time, causing delay as other data wait to be processed.

ATM is the first network technology to offer truly guaranteed bandwidth on demand, as the bandwidth can vary during the course of the call [7]. True enough, other services offer bandwidth that can vary with each call, but none can offer the ability to adjust the amount of bandwidth required to support a call once the call is established and to guarantee that it will be available when required. A high-quality videoconference, for example, might require 1.544-Mbps (T1) capacity as a rule. Yet, with the sophistication of contemporary compression techniques, that call might require much less bandwidth much of the time. Once that call is set up, a full T1 is dedicated to it, regardless of the actual bandwidth requirement moment by moment. ATM is not so rigid; it can adapt dynamically to the bandwidth actually required.

ATM networks provide for error detection of the header only and not the payload. ATM networks make no provision for error correction. The main concern is to deliver a cell to only the addressee. If the address is corrupted, the cell is discarded, and the endpoint is responsible for determining that fact and recovering from the loss through a request for retransmission. The advantages of this simplified approach to error control are increased speed of switching, reduced latency, and lowered cost, as the ATM switches require less memory and processing power.

· I.113: B-ISDN Vocabulary

· I.121: Broadband Aspects of ISDN

· I.150: B-ISDN ATM Functional Characteristics

· I.211: B-ISDN Service Aspects

· I.321: B-ISDN Protocol Reference Model

· I.327: B-ISDN Functional Architecture Aspects

· I.361: B-ISDN ATM Layer Specification

· I.362: B-ISDN ATM Adaptation Layer Functional Description

· I.363: B-ISDN ATM Adaptation Layer Specification

· I.413: B-ISDN User-Network Interface

· I.432: B-ISDN User-Network Interface-Physical Layer Specification

· I.555: Frame Relay and ATM Internetworking

· I.610: B-ISDN Operations and Maintenance Principles and Functions

Significant ATM Forum implementation documents include the following:

· ATM User-Network Interface (UNI) Specification for PVCs

· ATM Broadband InterCarrier Interface (BICI) Specification

· ATM Data Exchange Interface (DXI) Specification

Internet Engineering TaskForce (IETF) RFCs (Requests for Comment) include the following:

· RFC 1821: Integration of Real-Time Services in an IP-ATM Network Architecture

· RFC 2225: Classical IP and ARP (Address Resolution Protocol) over ATM

· RFC 2684: Multiprotocol Encapsulation over ATM Adaptation Layer 5

Access to an ATM backbone network occurs at rates of DS-1 or greater. Whether access is from an end-user CPE environment, a Frame Relay or X.25 network switch, or an IP-based router, the access rate is DS-1, DS-3, or a higher rate such as OC-3 (155 Mbps) or OC-12 (622 Mbps). Access is provided through a User Network Interface (UNI) specified by the MFA Forum:

· User Network Interface (UNI) specifically refers to a UNI between a user device and an ATM network. A private UNI is employed in a private ATM network (e.g., LAN domain) and for access to a public network. A public UNI is used between switches in a public ATM network. As the UNI is cell based, end-user CPE converts the user Protocol Data Units (PDUs) into ATM PDUs and cells.

· Data Exchange Interface (DXI) is a private UNI for end-user access to an ATM network from DTE/DCE such as a bridge, router, or ATM DSU The DXI concept enables the sharing of protocol responsibility between the user and the network provider. At the physical layer, the DXI permits connection via V.35, EIA 449/530, or EIA 612/613 HSSI connection. A variation of the HDLC pro-tocol is used at the data link layer. The user information is encapsulated within a DXI (HDLC) frame and converted to the appropriate, class-specific ATM protocol at the DCE. The DCE assumes responsibility for functions through Segmentation and Reassembly (SAR), presenting the data to the ATM network switch in ATM cells. Several DXI modes exist that correspond to AAL 3/4 and AAL 5.

· The Network-to-Network Interface (NNI), also known as B-ISDN InterCarrier Interface (B-ICI), is a public UNI for interconnection of public networks via PVCs. NNIs exist for interconnection of ATM networks and for interconnection of ATM and Frame Relay networks. Interoperability issues that need to be addressed include protocol conversion, mapping between virtual circuits, alignment of traffic management parameters, and mapping of local network management information.

· Frame UNI (FUNI) is a derivative of the DXI standard to extend ATM access to smaller sites at rates from 56 kbps to 1.544 Mbps (T1). Low-speed data enter a router, which forwards the data to the ATM switch as frames similar to Frame Relay frames. Those frames then are converted to cells.

Customer Premises Equipment(CPE) comprises Data Terminal Equipment (DTE), Data Communications Equipment (DCE), and voice and video switches, servers, and terminal equipment.

· DTE includes mainframe, midrange, and PC-server host computers connected through the UNI.

· DCE comprises ATM-equipped bridges, switches, routers, and gateways connected through the DXI UNI. A number of vendors offer ATM premises switches. Such switches are used for LAN interconnection, employing a highly redundant cell-switching fabric capable of switching speeds that can reach as high as OC-12 (622 Mbps).

· CPE includes ATM-based PBXs and video servers connected via the UNI or the DXI. These devices were always somewhat unusual and now have been superseded by IP-based devices.

Broadband Switching Systems (BSSs)are carrier exchange switches capable of broadband switching and transport. They employ highly redundant cell-switching fabrics that currently can operate in a range up to 1 Tbps in terms of aggregate bus speed. Maximum port speeds generally are OC-12 (622 Mbps). ATM switches also are highly intelligent, providing buffering, routing, and flow control as well as segmentation and reassembly. The switches are highly redundant and fault tolerant. They fall into two categories, core and edge switches:

· Core switches also are known as backbone switches. They generally are housed in a wire center, along with a traditional CO circuit switch or tandem. Core switches involve cell-switching fabrics that generally operate at aggregate rates of up to 1 Tbps.

Transmission facilities at the Network-to-Network (NNI) level and between network nodes most commonly are SDH/SONET fiber optic in nature, although other media are possible. Local loop facilities (UNI and DXI) can be any medium capable of supporting transmission speeds of DS-1 (T1 at 1.544 and E1 at 2.048 Mbps) or better, although fiber always is preferable. For very short distances in a premises-based ATM environment, Category 5e (Cat 5e) UTP supports ATM speeds up to 155 Mbps. Cat 6 cable supports data rates up to 622 Mbps and beyond.

Inverse Multiplexing over ATM (IMA) commonly is used for high-speed access to a public ATM network when speeds above T1/E-1 are required but T3/E-3 or fractional services are not available. IMA also is used when a T3 or FT3 is not cost effective, with the crossover point generally in the range of 12 Mbps or 8 T1s. With that crossover point in mind, IMA permits as many as 8 T1s to link together to support aggregate bandwidth of up to 12.288 Mbps (1.536 Mbps × 8). The ATM cell stream is spread across the separate T1s in a round-robin fashion and is resynchronized and reconstituted at the edge switch. While the preferred communications link is in the form of SONET optical fiber, it is not always available. Significantly, IMA actually provides some benefit at this bandwidth level in the form of redundancy. As an integrated network service, ATM can support all information types but typically requires an optical fiber link, which is susceptible to catastrophic failure. Multiple T1 links provide some level of redundancy in an IMA scenario, as all are unlikely to fail at the same time, unless one very powerful backhoe manages to rip an entire copper cable out of the ground. And, assuming that the T1 circuits travel different physical routes, even backhoes pose manageable threats [34].

(Source: TA-NWT-00113. Copyright 1993 by Bellcore.)

Figure 10.8: ATM cell structure

While this level of overhead, commonly known as the cell tax, might seem wasteful of bandwidth, the advantages far outweighed the drawbacks—at least initially—a classic trade-off between effectiveness and efficiency. The small cell size offers the advantage of supporting any type of data, including voice, fax, text, image, video, and multimedia—whether compressed or uncompressed. The fixed cell size offers the advantage of predictability, very unlike the variable-length frames of X.25, FR, and Ethernet. This level of predictability yields much improved access control and congestion control. Bandwidth is cheap over optical pipes, especially given the development of DWDM (Dense Wavelength Division Multiplexing) and its variants. ATM switches are geared to process ATM cells in hardware at very high speeds. Finally, ATM switches currently operate at an aggregate internal bus speed in the range of 1 Tbps or more, which makes bandwidth cheap in the switch as well. In any event, the 48-octet payload was set as a standard for the ATM cell.

The cell header provides limited Data Link Layer (Layer 2) functionality, managing the allocation of the resources of the underlying Physical Layer (Layer 1) of the transmission facility, which ideally is SONET/SDH in nature. The ATM cell switches also perform layer 1 functions such as clocking, bit encoding, and physical-medium connection. The header also is used for channel identification, thereby ensuring that all cells travel the same physical path and, therefore, arrive in sequence [36]. The header is structured as follows:

· Virtual Path Identifier (VPI): Eight bits identifying the Virtual Path (VP). The path is determined at the input port and is fixed for each call but is shared among multiple calls. The path is from the switch input port, through the switching matrix, to the output port, and then across a link between any two consecutive ATM entities. The VPI and Virtual Channel Identifier (VCI) jointly can be considered as a label for the allocation of resources across an end-to-end path, realizing that the label may require translation (change in value) from link to link; this translation process would take place at the ATM switching nodes or cross-connect points. As ATM switches, like all switches, work at Layers 1 (Physical Layer) and 2 (Data Link Layer) of the OSI Reference Model, they work on a link-by-link basis. That is to say that the cell address has local significance only.

· Virtual Channel Identifier (VCI): Sixteen bits identifying the Virtual Channel (VC), which is established each time a call is set up in the ATM network. A VC is a unidirectional channel for transporting cells between two consecutive ATM entities (e.g., switches) across a link.

· Payload-Type Indicator (PTI): Three bits distinguishing between cells carrying user information and cells carrying service information.

· Cell Loss Priority (CLP): One bit identifying the priority level of the cell to determine the eligibility of that cell for discard in the event of network congestion. Clearly, some applications, such as LAN-to-LAN traffic, are tolerant of loss. Other applications, such as real-time voice, are highly intolerant of loss.

· Header Error Control (HEC): Eight bits providing error checking of the header but not the payload. There is no provision for error correction.

The ATM reference model is multidimensional, with three planes and four layers, as illustrated in Figure 10.9. The lower two layers of this reference model loosely compare to the Physical Layer of the OSI Reference Model. As in the OSI model, each layer of the ATM model functions independently to perform its designated functions, all layers are tightly linked, and the functions are highly coordinated. The layers of the ATM reference model are the Physical Layer, ATM Layer, ATM Adaptation Layer, and higher layers and functions. The planes include the Control Plane, User Plane, and Management Plane.

(Source: Network VAR.)

Figure 10.9: ATM protocol reference model

· Physical Medium (PM) sublayer specifies the physical and electro-optical interfaces with the transmission media on both the transmit and receive sides. Timing functions are provided at this level. SDH/SONET optical fiber is the preferred physical medium.

· Transmission Convergence (TC) sublayer handles frame generation, frame adaption, cell delineation, header error control, and cell rate decoupling. The frame generation function takes the frame of data presented by the transmitting device across the PM sublayer for presentation to the ATM Layer and subsequent segmentation into cells. On the receive side, the TC sublayer receives data in cells and decouples it to reconstitute the frame of data, checking all the while for header errors before presenting the data to the PM sublayer, which passes the data to the end-user device.

ATM Layer (ATM) functions include multiplexing of cells, selection of appropriate VPIs and VCIs, generation of headers, and flow control. At this layer, all multiplexing, switching, and routing take place for presentation to the appropriate Virtual Paths (VPs) and Virtual Channels (VCs) of the SONET fiber optic transport system, which interfaces through the Physical Layer. Virtual Channels exist within Virtual Tributaries (VTs). For example, VT1.5 (i.e., T1 frame) might carry 24 channels, each of which carries a single voice communication in a time slotor a data communication in multiple time slots. That individual voice or data trans mission comprises a VC. As illustrated in Figure 10.10, a VC travels over a VP, which is neither fixed nor dedicated.

· Convergence Sublayer (CS) functions are determined by the specifics of the service supported by that particular AAL. Service classes are designated as A, B, C, and D.

· Segmentation and Reassembly (SAR) sublayer functions segment the user data into payloads for insertion into cells on the transmit side. On the receive side, the SAR extracts the payload from the cells and reassembles the data into the information stream as originally transmitted. In other words, the process of segmentation takes place at the ingress edge of the ATM domain. Across the entire ATM network, from switch to switch and from edge to edge, data flows in a cell stream. The cells decouple and the data reassemble (reconstitute to original form) at the egress edge of the ATM domain.

· AAL Types are supported by the functions of the Convergence Sublayer (CS). There exist defined AAL Types 1, 2, 3/4, and 5, each of which supports a specific class of traffic (see Table 10.4). AAL information is nested within the payload of user information cells.

· AAL Type 1 supports Class A traffic, which is connection-oriented Constant Bit Rate (CBR) traffic timed between source and sink. Such traffic is stream oriented and intolerant of delay. Isochronous traffic such as digitized, uncompressed voice is supported via Class 1 AAL, which essentially permits the emulation of a T/E-carrier circuit. All such traffic is carefully timed and must depend on a guaranteed rate of network access, transport, and delivery. Such traffic is marked as high priority in the cell header, as transmission delays could considerably impact presentation quality. Class A traffic is transmitted over a Virtual Path (VP) and in a Virtual Channel (VC) appropriate for such high-priority traffic.

· AAL Type 2 supports Class B traffic, which is connection-oriented, real-time Variable Bit Rate (rt-VBR), isochronous traffic timed between source and sink. Compressed audio and video are Class B. Class B traffic, for example, includes compressed voice using the relatively simple Digital Speech Interpolation (DSI) technique for silence suppression. Compressed video using the Moving Pictures Experts Group (MPEG) compression algorithms also are Class B. Class B traffic is marked as high priority in the cell header and transmitted over an appropriate VP and VC.

· AAL Type 5 supports Class C traffic in message mode only. Such traffic is Variable Bit Rate (VBR) traffic with no timing relationship between source and sink and consisting of only 1 IDU, as illustrated in Figure 10.11. AAL Type 5 also is known as Simple and Efficient AAL (SEAL), as some of the overhead has been stripped out of the Convergence Layer. AAL Type 5 initially was intended solely for use in signaling and control (e.g., NNI applications) and network management (e.g., Local Management Interface, or LMI). The IDUs can vary in length, up to 65,535 octets. A 32-bit CRC is appended to the IDU at the Convergence Layer as part of the trailer. AAL 5 also supports Class X traffic, which is either Unspecified Bit Rate (UBR) or Available Bit Rate (ABR). Such traffic is VBR and either connection oriented or connectionless [37]. AAL 5 is used in support of a wide variety of data traffic, including LAN Emulation (LANE) and IP.

ATM-defined service categories, which relate back to AAL types and are reflected in Table 10.4, include CBR, rt-VBR, nrt-VBR, UBR, and ABR. Guaranteed Frame Rate (GFR) is the most recently defined service:

· Real-Time Variable Bit Rate (rt-VBR) traffic is bursty in nature but depends on timing and control information to ensure the integrity of the data stream. Traffic parameters include Peak Cell Rate (PCR), Cell Delay Variation Tolerance (CDVT), Sustainable Cell Rate (SCR), Maximum Burst Size (MBS), and Burst Tolerance (BT). The QoS parameter is Cell Loss Ratio (CLR). Voice, audio, and video encoded at variable bit rates are examples of rt-VBR traffic.

· Non-Real-Time Variable Bit Rate (nrt-VBR) traffic is bursty, but its non-real time nature is not dependent on loss or delay because there is time to recover. Traffic parameters include Peak Cell Rate (PCR), Cell Delay Variation Tolerance (CDVT), Sustainable Cell Rate (SCR), Maximum Burst Size (MBS), and Burst Tolerance (BT). The QoS parameter is Cell Loss Ratio (CLR). Examples include data traffic such as X.25, Frame Relay, transaction processing, LAN-to-LAN, and non-real-time buffered voice and video.

· Unspecified Bit Rate (UBR) traffic is a best effort type. Traffic parameters include Peak Cell Rate (PCR) and Cell Delay Variation Tolerance (CDVT). No QoS commitment is made. Traditional computer applications, such as file transfer and e-mail, fall into this category.

· Available Bit Rate (ABR) is a best effort category in which the network attempts to pass the maximum number of cells but with no absolute guarantees. Subsequent to the establishment of the connection, the network may change the transfer characteristics through a flow control mechanism that communicates to the originating end-user device. This flow control feedback mechanism is in the form of Resource Management (RM) cells. In other words, during periods of congestion, the network can buffer cells and advise the sender to throttle back on the rate of transmission. ABR supports VBR traffic with flow control, a minimum transmission rate, and specified performance parameters. Traffic parameters include Peak Cell Rate (PCR), Cell Delay Variation Tolerance (CDVT), and Minimum Cell Rate (MCR). No QoS commitment is made. ABR service is not intended to support real-time applications.

The aforementioned references to AAL Service Classes and Service Categories would be incomplete without defining the key QoS parameters. ATM network performance parameters are defined in ITU-T Recommendation I.356. Those parameters are used to measure B-ISDN performance of an end-to-end user-oriented connection in terms of quality of service specific to each service class and category. The ATM Forum extended this standard through the definition of QoS parameters and reference configurations for the User Network Interface (UNI). The relevant ATM Forum (now MFA Forum) document is Traffic Management Specification Version 4.0 (af-tm-0056.000, April 1996). QoS objectives are not strictly defined.

Performance parameters defined by the ITU-T address accuracy, dependability, and speed. Accuracy parameters include Cell Delay Variation (CDV), Cell Error Ratio (CER), Cell Loss Ratio (CLR), Cell Misinsertion Rate (CMR), Cell Transfer Delay (CTD), and Severely Errored Cell Block Ratio (SECBR):

· Cell Delay Variation (CDV) is the variation in an individual cell's Cell Transfer Delay (CTD) and its expected transfer delay. CTD is a form of jitter, which can seriously degrade the quality of voice and video payloads. If cells arrive sooner than expected, the clumping can cause the PCR (Peak Cell Rate) to be exceeded and the excess cells to be discarded. If some cells arrive too late, the result may be gaps in the received information stream. Cell Delay Variation Tolerance (CDVT) is a measurement of the maximum allowable CDV tolerance between two end stations. Peak-to-peak CDV is negotiated between the end station and the network; peak to peak refers to the best case compared with the worst case, that is, the difference between the earliest and the latest arriving cells on a connection.

· Cell Error Ratio (CER) is a dependability parameter expressed as the ratio of the number of errored cells to the total number of transmitted cells sent over a measurement interval. CER is not negotiated.

· Cell Loss Ratio (CLR) is a dependability parameter expressed as the ratio of the number of lost cells to the number of transmitted cells. Cell loss can occur for reasons that include misdirection of cells by a switch, a congestion problem causing a discard in consideration of buffer capacity, a station exceeding its PCR resulting in cell discard, or a cell that exceeds the maximum CTD and arrives too late for consideration and processing. CLR is negotiated between the end stations and the network and applies to the lifetime of the connection. CLR applies to all service categories except UBR.

· Cell Transfer Delay (CTD) is the average time it takes a cell to transverse the network from source to destination across a UNI. CTD is the sum of all delays imposed by coding and decoding, segmentation and reassembly, propagation across transmission media, cell processing at the nodes, queuing of the cell in input and output buffers, and loss and recovery. If a cell arrives too late at the receiving station, it may be considered lost or late and may be disregarded. If the subject cell is a segment of a larger data packet, the entire packet must be discarded and forgotten or retransmitted. Maximum CTD (maxCTD) is negotiated between the end stations and the network.

· Severely Errored Cell Block Ratio (SECBR) refers to a sequence of some number (n) of cells transmitted consecutively (sent in a block) on a given connection, perhaps between OA&M cells. SECBR is a dependability parameter expressed as the number of severely errored cell blocks compared with the total number of cell blocks sent over a period of time, or measurement interval. A severely errored cell block outcome is realized when more than some number of cells in a block is errored, lost, or misinserted. SECBR is not a negotiated parameter.

An ATM traffic contract specifies all characteristics of a connection negotiated between a source endpoint and an ATM network. Traffic parameters are descriptions of the traffic characteristics of a source endpoint; they may be quantitative or qualitative in nature. Traffic parameters include (PCR), Sustainable Cell Rate (SCR), Maximum Burst Size (MBS), Maximum Frame Size (MFS), and Minimum Cell Rate (MCR). A traffic descriptor is the entire set of traffic parameters associated with a source endpoint. The traffic descriptors used during connection establishment include the following:

· Peak Cell Rate (PCR) is the maximum number of cells per second, in a burst, that is, within a short interval, that the network agrees to accept and transfer for a given UNI. Excess cells may be discarded by the ingress switch or marked as eligible for discard. For CBR service, the PCR is the guaranteed Constant Bit Rate for the virtual circuit. Enforcement of the PCR enables the network to allocate sufficient resources to ensure that the QoS parameters (e.g., Cell Loss Ratio and Cell Transfer Delay) are met. PCR can apply to all service categories.

· Sustainable Cell Rate (SCR) is the maximum average rate at which the network agrees to accept cells and support their transfer from end to end for each UNI. In other words, SCR is the average throughput. Enforcement of the SCR enables the network to allocate sufficient resources to ensure that the QoS parameters (e.g., Cell Loss Ratio and Cell Transfer Delay) are met over a period of time. SCR applies to VBR services.

· Maximum Frame Size (MFS) is the maximum size of a PDU, or frame, supported by the network. MFS relates specifically to the Guaranteed Frame Rate (GFR) service category.

· Minimum Cell Rate (MCR) is the minimum number of cells per second that the network agrees to support for a given originating endpoint across a UNI. This ABR service descriptor, expressed in cells per second, is that rate at which the originating endpoint can always transmit during the course of the connection.

These relate to the specifics of the user Protocol Data Unit (PDU), such as an SDLC frame or Frame Relay frame:

· Control Plane functions include all aspects of network signaling and control, such as call control and connection control.

· User Plane functions deal with issues of user-to-user information transfer and associated controls (e.g., flow control and error control mechanisms).

· Management Plane functions involve the management of the ATM switch or hub. The Management Plane is divided into Plane Management and Layer Management. Plane Management acts on the management of the switch as a whole, with no layered approach. Management of and coordination between the various planes is accomplished in Plane Management. Layer Management acts on the management of the resources at each specific layer of the model, for example, Operation, Administration, and Maintenance (OA&M) information.

Note, however, that LANE cannot resolve translational problems between disparate LANs (i.e., Ethernet and Token Ring); rather, a router must accomplish protocol conversion. LANE also supports only Ethernet and Token Ring; FDDI, for example, cannot be supported without the intervention of a router. Also, a router or router function is required to support communications between ELANs, which operate as closed user groups, much like Virtual LANs (VLANs) [31, 38–40].

MultiProtocol Over ATM (MPOA) is a specification (July 1997) from the ATM Forum designed to enhance LANE by enabling interELAN communications without the intervention of a router and the associated packet delay. MPOA provides high-performance, scalable routing functionality over an ATM platform. MPOA expands on LANE, Classical IP over ATM (RFC 1577), and the IETF's Next Hop Resolution Protocol (NHRP) in order to create a standardized notion of a virtual router within an ATM network. Between any two MPOA-capable end devices (e.g., MPOA-enhanced LANE hosts, bridges, or switches), MPOA maps routed and bridged flows of connectionless LAN traffic over cut-through ATM Switched Virtual Channels (SVCs), offloading the packet-by-packet processing steps performed by traditional routers. A route server in the ATM network contains the core intelligence to dynamically track the network topology and performance, thereby providing guaranteed bandwidth, reduced latency, and QoS. Since the routing intelligence is divorced from the packet-forwarding function, MPOA offers a scalable and flexible solution for LAN interconnectivity, corporate intranets, and multimedia applications such as distance learning, desktop videoconferencing, and collaborative work sessions. MPOA supports protocols such as Ethernet, Token Ring, FDDI, and I P. Through the NHRP protocol from the IETF, routing entities within the MPOA network can intercommunicate to determine the most appropriate path for a communication between edge devices on an interELAN basis. MPOA essentially synthesizes bridging and routing of connectionless interLAN traffic over a connection-oriented ATM network [31, 38, 40].

The Framed ATM over SONET/SDH Transport (FAST) specification from the ATM Forum (July 2000) defines the mechanisms and procedures required to support the transport of variable-length datagrams, known as ATM frames, over an ATM infrastructure using SONET/SDH facilities. Interoperability with existing ATM infrastructure is ensured through the use of existing ATM control and management plane functions such as signaling, routing, addressing, and OAM&P. The ATM Forum has specified both FAST UNI and FAST NNI interfaces. The FAST interfaces are functionally equivalent to their cell-based counterparts with the exception of the maximum size of the Service Data Unit (SDU), which is defined as a unit of interface information whose identity is preserved from one end of a layer connection to another.

Two encapsulation formats are defined. Frame encapsulation supports the transmission of variable-length frames with a maximum size of 9216 octets and using the AAL5 Adaptation Layer. This approach involves minimal overhead. Cell encapsulation involves the encapsulation of ATM cells within a FAST frame, supporting an SDU with a maximum size of 65,535 octets, with the specific size sensitive to the AAL used.

FAST is similar to DXI (Data eXchange Interface) and FUNI (Frame User Network Interface) with several major exceptions. DXI and FUNI are designed for access to an ATM network over relatively low speed plesiochronous transmission facilities, while FAST is designed for access and/or interswitch trunking over very high speed SONET/SDH transmission facilities [42].

ATM clearly offered the great advantage of lots of bandwidth, with access rates usually at a minimum of T1/E-1 and backbone interfaces as high as OC-192 (10 Gbps). The advantages of SONET optical fiber, for both access and transport, include bandwidth, error performance, and fault tolerance. ATM is linked closely to SONET/ SDH, which is not to say that ATM is not medium independent. While the backbone ATM networks are designed around a SONET/SDH fiber-optic transmission system, UTP and STP can attach workstations to an ATM LAN switch. The local access facilities generally are copper or microwave at the T1 and T3 levels, but fiber is required at higher speeds. ATM over VSATs is available internationally at rates up to 34/45 Mbps (E-3/T3) to leapfrog national wireline networks that do not support ATM.

Much as I discussed relative to Frame Relay, ATM offers mesh networking of virtual circuits with a single access circuit at each site without the need for complex and expensive leased lines. As a result, network configuration and reconfiguration are much simplified. Finally, ATM networks and associated costs are highly scalable, as is the case with any VPN technology. In other words, the cost of the network is very much in proportion to the scale of the network in terms of attached devices, networked locations, and bandwidth requirements.

The incumbent LECs and IXCs, worldwide, aggressively began deploying ATM in the cores of their data networks in the early 1990s. As they experienced growth in voice traffic, some gradually transitioned voice to those backbones, rather than continuing to invest in circuit-switching technology. However, ATM is a forklift upgrade and a very expensive and complex one at that.

Since roughly 2000, Ethernet and IP have greatly increased in popularity at the expense of both ATM and SONET. Ethernet now offers port speeds up to 40 Gbps, a speed that ATM switches likely will never achieve, as development has stopped. The CO remains an ATM stronghold, however, where ATM over OC-3 SONET interfaces is popular as the new highest common denominator for patching together devices from multiple vendors, all of which implement this interface in the same way.

ATM is intended to support any application that the contemporary mind can conceive, and it does an excellent job. The driving force behind ATM development and deployment was primarily that of data applications, where most of the growth was and where it remains, but ATM supports voice, video, and multimedia exceptionally well also and simultaneously with guaranteed QoS.

As a backbone technology, however, the ATM's cost and inherent complexity have not positioned it well against a combination of IP and MPLS. In fact, it was cost and complexity that prevented ATM from taking over the LAN domain, despite its technical elegance. Ultimately, Ethernet's low cost and simplicity prevailed. However, ATM has found its way into the broadband local loop, specifically ADSL and several versions of Passive Optical Network (PON). Notably, the impetus for ATM, ADSL, and PON all originated in the telco domain, where guaranteed QoS is a way of life.

10.5 BROADBAND ISDN

Potential means that you haven't done anything yet.

Ray Horak, 2006

Deployment of N-ISDN varies widely from country to country and within each nation on a state, province, and metropolitan area basis. N-ISDN has not been well received in the United States, although it was quite popular in certain countries within Western Europe and was widely deployed in parts of Asia. As the ITU-T anticipated quite early on, N-ISDN proved to be underpowered in the context of the contemporary appetite for bandwidth-intensive applications.

B-ISDN is defined by the ITU-T as a service requiring transmission channels capable of supporting rates greater than the primary rate. The primary rate, of course, is defined in the North American Primary Rate Interface (PRI) as 1.544 Mbps and in the European Primary Rate Access (PRA) as 2.048 Mbps. There are three underlying sets of technologies and standards that are absolutely critical to B-ISDN. First, Signaling System 7 (SS7) is viewed as the signaling and control that supports B-ISDN, just as it supports N-ISDN. Second, Asynchronous Transfer Mode (ATM) is the backbone network switching and multiplexing technology. Third, SDH/SONET is the physical backbone network transmission technology.

B-ISDN user access is broadband in nature, relying on SDH/SONET fiber-optic transmission standards. Currently, there are two access interfaces specified. User-Network Interface A (UNI A) operates at OC-3 rates of 155 Mbps, while User-Network Interface B (UNI B) operates at OC-12 rates of 622 Mbps. Network-to-Network Interfaces (NNIs) are required for network access to B-ISDN from FR and N-ISDN networks.

The ITU-T defines two types of B-ISDN services, interactive and distribution:

· Interactive services involve bidirectional transmission and include three classes of service. Conversational services include voice, interactive data, and interactive video. Messaging services include compound document mail and video mail. Retrieval services include text retrieval, data retrieval, image retrieval, video retrieval, and compound document retrieval.

B-ISDN user equipment is an extension of that described for N-ISDN. Broadband Terminal Equipment 1 (B-TE1) is defined as B-ISDN-compatible CPE. While the exact nature of the B-TE1 remains undetermined, it is likely that it will take the form of ATM-compatible B-ISDN communications servers. Those servers will combine the functions of a PBX, data switch, and video switch. At the extreme, they will take the form of ATM-based multimedia communications servers.

Broadband Terminal Equipment Type 2 (B-TE2) is defined as terminal equipment that supports a broadband interface other than B-ISDN. Terminal Equipment Type 2 (TE2) continues to be defined as terminal equipment that supports an interface other than ISDN. Both B-TE2 and TE2 equipment will interface with the network through a Broadband Terminal Adapter (B-TA). All of this may well be moot now, as IP seems to have relegated B-ISDN to an historical pipe dream.

N-ISDN was intended to be available universally. This did not happen, of course, as many nations and regions did not see the value in upgrading the infrastructure to support it. B-ISDN seems headed for the same fate, although for a different reason. There certainly has been no lack of interest in and commitment to the underlying technologies but, over time, the low cost and inherent simplicity of other technologies seem to have rudely shoved them aside. ATM is on the decline, at least in the backbone, in favor of a combination of IP and MPLS. Ethernet is gaining ground over ATM in PON local loops and in the MAN. SS7 is on the decline in favor of SIP.

There is no question that B-ISDN has a lot of potential, but as a famous author (that would be me) once said, "Potential means you haven't done anything yet." That statement is perhaps a bit harsh when it comes to B-ISDN, as all of the underlying technologies and standards were in place and made considerable impact on telecommunications. It is highly unlikely, however, that B-ISDN will ever be fully specified, and much less that it will ever take full form as the ubiquitous integrated network solution, end to end.

10.6 ADVANCED INTELLIGENT NETWORKS (AINs)

Once upon a time, the networks were truly dumb. Through the 1960s and even into the 1970s, they remained fairly dumb. In other words, the networks were composed largely of switches and pipes that could do little more than connect calls as directed based on a hierarchical order of limited switching intelligence. That intelligence was in the form of very limited programmed logic housed in databases that interacted at a minimal level. Each switch performed its own job, with little thought of the network as a whole.

The linchpin of the intelligent network is the Service Creation Element (SCE), which is a set of modular programming tools permitting services to be developed independently of the switch, the nature and capability of which can vary by manufacturer and software generic. The SCE divorces the service-specific programmed logic from the switch logic, thereby enabling the independent development of the service. The service, therefore, is available to all switches in the network. The concept of a sparse network is one of dumb switches supported by centralized intelligence with connectivity between distributed switches and centralized logic provided over high-speed digital circuits. In the context of a complex network, this concept is highly viable in terms of both performance and economics.

Bellcore defined AIN in the early 1980s as AIN Software Release 1.0. This release is intended to provide a generic and modular set of tools that enable the creation, deployment, and management of services on a flexible basis. The software tools yield a suite of service offerings that are accessible to all network switches but which operate independently from the switch logic. The services, therefore, can be defined, developed, and deployed quickly and in a multivendor environment.

Characteristics of AINs include service creation toolkits, which enable the creation of centralized logic residing in centralized databases for the development and delivery of features across the network. AINs support all ISDN features, including caller ID, selective call blocking, and distinctive ringing. Finally, AINs are intended to provide support for Personal Communications Services (PCS), which permit subscribed features to be supported across networks of all types. PCS, not to be confused with the PCS cellular service, currently is conceptual.

The Service Creation Environment (SCE) is the key distinction of an AIN. The SCE offers the carrier a toolkit for the development of service offerings, which can be provided on a network basis. The services can be generic or customer specific. Often, you can create the customer-specific services through linking generic services and varying the available options and parameters. Additionally, the SCE can be opened to the user organization, which then can customize the service offering as desired.

· Common Channel Signaling System 7 (SS7) is an out-of-band signaling system for communication between devices in the carrier networks, as is discussed in several previous chapters. SS7 is an absolute requirement for both ISDN and AIN. SS7 is deployed throughout advanced voice networks worldwide.

· Service Switching Points (SSPs) are PSTN switches that act on the instructions dictated by AIN centralized databases. SSPs can be end offices or tandem switches, as is defined in the discussion of the PSTN.

· Signal Transfer Points (STPs) are packet switches that route signaling and control messages between SSPs and SCPs and between STPs.

· Signal Control Points (SCPs) contain all customer information in databases that reside on centralized network servers. SCPs provide routing and other instructions to SSPs, as requested and required.

· Service Management Systems (SMSs) are network control interfaces that enable the service provider to vary the parameters of the AIN services. Under certain circumstances, the user organization may be provided access to a partition of the SMS.

· Adjunct Processors (APs) are decentralized SCPs that support service offerings limited either to a single SSP or to a regional subset of SSPs. APs might support routing tables or authorization schemes specific to a single switch or regional subset of switches.

· Intelligent Peripherals (IPs) provide intelligent peripheral capabilities to enhance the delivery of certain services by offloading processing demands from the SCPs and providing a basic set of services to the SCPs. The role of the intelligent peripheral typically includes collection of digits, collection and playing of voice prompts, collection of voice responses and their conversion to digits, menu services, and database lookups. Voice processing and voice recognition, for example, might be implemented in support of the processing of "collect" calls. Services such as Verizon's 1-800-COLLECT enable the processing of a collect or third-party, billed call without operator intervention. Voice recognition also can be used for voice printing, which permits user authentication in highly secure applications. A number of LECs provide automated directory assistance services, accomplishing database lookups based on voice recognition inputs through intelligent peripherals. As the term IP (Intelligent Peripheral) has become confused with IP (Internet Protocol), the term Special Resource Function (SRF) often is used to describe these peripherals and the functions they perform.

AIN services truly are open to the imagination. But the availability of AINs is very uneven within the United States and, certainly, around the world. AIN services are of wide variety, including the following:

· Follow-me service provides for call forwarding on a predetermined schedule. A telecommuter, for example, might have the network forward calls to the home office three days a week during normal business hours. Calls would be directed to the traditional office two days a week. Calls clearly outside of normal business hours automatically would be directed to a voice mail system.

· Computer security service automatically would serve to provide secure access to networked hosts based on Calling Line ID (CLID) and supplemented by authorization codes. Additionally, the network automatically would keep a log of all access and access attempts, thereby providing an audit trail.

· Call pickup service also known as call notification service, provides for calls to be answered automatically by a voice processor. The called party can be notified of a deposited voice message by pager, fax, e-mail, or other means. The caller can enter a privilege code provided by the called party to distinguish the priority of the calling party. To pick up the call, the called party dials a DISA (Direct Inward System Access) port on the network switch and enters password codes in a manner similar to that used to access contemporary voice mail systems.

· Store locator service also known as single-number dialing, provides the ability to advertise a single number. The network routes calls to the closest store location in terms of either geography or time zone based on the originating address (i.e., telephone number) of the caller. This service is deployed widely in the United States in support of businesses offering delivery services. It also is used widely for directing calls to networked incoming call centers such as reservation centers (e.g., airlines, auto rental agencies, and hotel chains).

· Multilocation extension dialing provides for network routing of calls based on abbreviated numbers. This VPN service resembles a coordinated dialing plan in a networked PBX environment.

· Call blocking can work on incoming or outgoing basis. On an incoming basis, the feature allows the subscriber to program some number of telephone numbers from which incoming calls are denied. Incoming call blocking is widely available. On an outgoing basis, this feature typically supports the blocking of calls to international destinations, either in total or to specific country codes. Content blocking supports the blocking of calls to specific numbers, such as 900/976 numbers. This capability is deployed fairly commonly in many foreign networks but is not widely available in the U.S. PSTN. Cellular radio providers have made extensive use of such capability on a systemwide basis to avoid cellular fraud involving certain countries in the Middle East, The Bahamas, and South and Central America. Ameritech (now part of SBC) conducted trials of its Call Control Service as early as 1994, enabling residential customers to block calls to specific numbers or to all numbers except those on an allowed list. Widespread deployment of outgoing call blocking is unlikely in the near future.

· Enhanced call return enables the subscriber to access what is, in effect, a highly sophisticated, network-based voice mail system. Bell Atlantic (now part of Verizon) trialed an offering that enabled the user to call the system and enter security codes to gain access to the switch and data stores. A voice announcement identified the date, time, and phone number of the last incoming call, at which point the return call could be launched by pressing a number on the telephone keypad [54].

· Enhanced call routing is a network-based enhancement to toll-free calling. The callers are presented with options that enable them to specify their needs and then be connected with the offices or individuals best able to satisfy them. MCI (Verizon) and Stentor (Canada) conducted trials on such a system in early 1995, providing seamless interconnection between the U.S. and Canadian networks [55].

· Call completion service enables the directory assistance operator to extend the call automatically, perhaps at an additional charge. This capability is offered by cellular providers to avoid accidents caused by "driving and dialing."

· Number portability serves to provide portability of telephone numbers such as INWATS and 900 numbers. This function is utilized increasingly as the regulators insist on portability of local numbers between LECs (Local Exchange Carriers). Local Number Portability (LNP) is essential to facilitate local competition, as decreed by the Telecommunications Act of 1996.

The future of the AIN, at least as defined by Bellcore, is a bit cloudy. Many of the large carriers have developed their own versions of IN in support of toll-free services, calling card verification, and various other services. Whether they will move toward a standardized version is not clear, although it appears unlikely. As AIN capabilities continue to roll out on a gradual basis, you likely never will see headlines announcing the arrival of AIN. Rather, it has crept into the networks and will continue to do so. Driving forces over the last decade or so in the United States include FCC (May 4, 1995) and state Public Utility Commission (PUC) decisions in support of caller ID and name ID, which require SS7 and some level of AIN. A number of carriers have offered automated directory assistance; that trend is likely to continue. The Telecommunications Act of 1996 introduced local competition in the United States, and the states then took the initiative of requiring local number portability and equal access at a local exchange level, with AINs being the key to implementing such flexibility. Similar forces and trends have influenced the gradual deployment of AIN services elsewhere in the world. AIN is less of a factor in developing countries, as one might imagine.

These types of services do not necessarily require broadband networks to support them, although broadband is always good. For that matter and in this world oriented toward instant gratification, more is always better and faster is better still. Broadband networks certainly fill that need with more bandwidth that supports faster communications.