Technology Distillation

David B. Hill
April 2000
Copyright Hill Associates, Inc. All Rights Reserved

Access Technologies

It is typical to divide a communications network into two distinct elements, the access network and "the cloud," or backbone network. On this page we will consider the technologies of most interest in the access portion, i.e., we will consider technologies used to bring user information (voice and data) to the backbone structure for transport and switching.

ISDN
The classical telco loop carried information in analog form, and used tones in the voice channel itself for signaling purposes. ISDN is an access technology that alters both these characteristics. ISDN loops carry all traffic in digital form, and use a separate channel for signaling and user data packets. In the so-called Basic Rate Interface (BRI) there are two B (or Bearer) channels offering 64kbps data rates, and a D (or Delta) channel that carries signaling packets, and user data packets, at 16kbps. The Primary Rate Interface (PRI) offers 23 B channels at 64kbps, and a single D channel, also at 64kbps.

The BRI is primarily aimed at residential users, with telecommuting as the "killer application." The two B channels could be used, for example, to carry a voice conversation while transmitting data or fax simultaneously. The D channel in this application would principally be used to obtain B channel services from the provider, i.e., principally as a signaling channel.

PRI is aimed at business users, who typically use it to provide trunking for premises PBXs, and to support higher-speed dial up services such as teleconferencing. PRI has been a far more successful offering than BRI, since it often reduces trunking costs, while providing a rich panoply of services.

Today, most BRI customers use the two B channels "strapped together" to provide a single 128kbps connection to an Internet Services provider. In a second application, the D channel is used to provide an always-on, low speed, packet-mode connection to an ISP, with the B channels being used to support higher speed data bursts. This approach is called Always On, Dynamic ISDN, or AO/DI.

DSL
Typical analog local loops use only the band of frequencies from 0 to 4kHz to carry traffic, generally voice or modem traffic. The actual twisted pair loop, unless it has some form of impairment, can carry a significantly wider band of frequencies.

In most DSL technologies, the analog local loop is frequency division multiplexed into a collection of channels. The lowest frequency band (0 - 4kHz) is used for POTS (plain old telephone service). Above the voice band a so-called guard band is provided to minimize cross-talk between the POTS channel and higher frequencies. Above the guard band are the data channels, with bands often being asymmetrically assigned so as to provide higher data rates from the network to the customer (downstream) than from the customer to the network (upstream).

At both the customer premises and the central office, splitters are used to separate voice and data channels. Voice traffic is routed to the voice switch, while data traffic in the high frequency bands is run through a multiplexer to an ISP router or an ATM switch. This approach offers always-on data connectivity, while removing much of the pressure on the central office voice switch that Internet traffic often brings.

In the strictest sense, the term DSL includes any facility that offers digital connectivity. One should therefore include under the general heading ISDN, which provides digital capability, as well as HDSL which provides a digital facility equivalent to the T-1 carrier, using the ISDN BRI signal transmission technology. Generally, however, the term DSL is used to describe frequency division multiplexed subscriber loops offering asymmetric data services, as discussed above.

Wireless - Narrowband
In this category one finds cellular and PCS systems. Although they go by different names, the two systems use essentially the same technologies, and are differentiated principally by the fact that they operate at different locations in the electromagnetic spectrum. The cellular principle is based on the concept of frequency re-use. A cellular operator divides the region into physical entities called cells, and allocates some of the spectrum available to each cell. Adjacent cells, of course, use different frequencies, but the frequencies allocated to a given cell are re-used in cells farther away. By varying transmitter power the size of the cells can be varied. Clearly, small cells offer substantial frequency re-use opportunity, allowing many users to be served simultaneously. On the other hand, small cells complicate the problems of "hand-off" of a call as a user crosses a cell boundary.

Early cellular systems used strictly analog techniques, and created multiple channels in a cell by frequency division multiplexing. This technique, called FDMA, continues to be used in many parts of the U.S. More modern cellular, and all PCS systems, use digital techniques. That is, voice is digitized, and then sent using a modem operating in the specified portion of the spectrum. The voice signal is typically compressed rather substantially, leading to high system capacities. Channel allocation is done in a number of ways, perhaps the most common is to use frequency division multiplexing to create multiple channels within a cell, and then time division multiplexing to carry multiple signals in a given frequency channel. This technique is called TDMA (for Time Division Multiple Access). The Global System for Mobile Communications (GSM) which is the standard in most of the world for cellular telephony is a TDMA system. Unfortunately it uses different frequency and time allocations than does the North American TDMA standard, so the two unfortunately cannot interoperate.

Finally, CDMA (for Code Division Multiple Access) uses different pseudo-random coding sequences to differentiate channels, rather than different frequencies or time slots. Adjacent cells use the same code sequences and frequencies, but the code sequences are offset from one another in time. This requires exquisite timing synchronization in a CDMA network, the global positioning system satellites are often used for this purpose. Next generation cellular systems (and even some today) will certainly use CDMA, and there is some hope that a global standard will arise around this technology.

Narrowband wireless systems are principally used for voice, but data transmission is certainly possible. Because FDMA and TDMA systems divide the spectrum into narrow channels, data rates are typically low, and applications are restricted to low volume tasks such as mobile credit card checking (e.g., in taxicabs). Because CDMA systems make the entire band available to each cell, significantly higher data rates can be achieved. Third generation cellular systems will support data rates of 2mbps, and toll quality voice, and may significantly impact the wireline access providers. Applications for high speed cellular access clearly exist in both the residential and business marketplaces.

Wireless - Broadband
A number of techniques exist for using spectrum in the microwave region to carry voice and data signals. MMDS (Multichannel Multipoint Distribution System) is a technique originally proposed to carry "wireless cable" TV payloads, but now more widely used for Internet access. Sprint has purchased a number of MMDS providers; this may be their major thrust into the local market. MMDS is used for both business and residential access.

LMDS (Local Multipoint Distribution System) is a relatively new technique, and LMDS license holders have been granted a huge amount of spectrum. Like the proprietary systems from Teligent and Winstar, LMDS systems are targeted at business customers, particularly those occupying buildings that are not served by fiber facilities. Modems place data streams on carriers operating in the microwave band and "shoot," typically from an antenna located on the building rooftop to an antenna provided by the LMDS (or other microwave) carrier on another rooftop. The central antenna connects to fiber optic landline facilities; the broadband wireless systems are strictly designed for access to the landline backbone. LMDS is thought by some to be the most likely breakthrough wireless technology, and even companies such as Winstar are buying LMDS licenses to supplement their existing licenses at 38gHz.

Backbone Network Technologies

This page will discuss technologies typically found within "the cloud." These fall into two broad categories: pure transport technologies used to carry information from point to point, and switching technologies used to add value to the transport links by providing connectivity on an as needed basis between end user nodes. 

Transport Technologies - SONET
SONET stands for the Synchronous Optical NETwork, and is a set of standards for fiber optic transmission systems. SONET is an extension to the classical T-carrier network, intended to remedy difficulties that are present in T-carrier systems.

Two T-carriers arriving at a multiplexer can be out of synch. This makes multiplexing and demultiplexing very difficult, and the provisioning of new services out of a T-carrier network can be complex and costly. In contrast, SONET uses higher clocking standards and a different approach to multiplexing to make the provisioning of services significantly less difficult. In particular, add-drop multiplexing is easily accomplished with SONET, so that constituent streams in a SONET flow can be identified and dropped off at a customer premises, or network device, without requiring the demultiplexing of the overall flow. 

SONET contains a rich set of overhead channels that are used for operations, administration, management and provisioning. The add-drop multiplexing alluded to above could, for example, be initiated and managed from a remote terminal device, using the SONET OAM&P channels to pass instructions to the appropriate add-drop multiplexer. These overhead channels are also used by SONET devices themselves to detect and report errors, and if necessary to automatically switch to backup facilities. In contrast, T-carrier has essentially no OAM&P capability, making the management of such networks difficult at best.

Finally, SONET provides a standard for multiplexing that extends to 10gbps, well above the top of the T-carrier range. The standardization of the system provides for "mid-span meet," which refers to the ability of links created by multiplexers from different sources to meet in the middle and transmit information correctly. T-carrier had no mid-span meet capability at high speeds, since there are no real standards for T-carrier systems at speeds in excess of T-3 (44mbps). 

A SONET network would typically consist of terminal multiplexers dealing with legacy T-carrier systems and embedding T-carrier information in SONET frames. These TMs would then connect to add-drop multiplexers or ADMs, connected to one another and charged with distributing the information contained in the SONET stream. For reliability, redundant ring topologies are generally used to connect ADMs.

Transport Technologies - Wavelength Division Multiplexing
If a light fiber is carrying an information stream using a specific color of light (for illustrative purposes, let's say green), one could easily add a second stream by the expedient of running a red information stream along the fiber as well. Of course, one would need splitters and filters at both ends, so the "green receiver" only gets green light, and conversely for the "red receiver." Different colors of light correspond to different wavelengths, so the technique just described is called Wavelength Division Multiplexing.

Fiber optic systems operate in the infrared, and a WDM system uses different wavelengths of infrared light to carry multiple independent information streams over a single light fiber. Existing systems typically provide 16 or 32 streams, but recent announcements from Lucent, Nortel, Ciena and others suggest that 200 channel systems are not far in the future. Note that each stream could carry, for example, a SONET OC-192, operating at 10gbps. 200 streams at 10gbps per stream gives 2 terabits per second, a data rate capable of sending the entire Library of Congress in 20 seconds. Such high density systems are identified as Dense WDM (DWDM) systems. They use a somewhat different fiber from that commonly found in older carrier systems. Older fiber systems can be rejuvenated using lower density WDM, new plant such as that being developed by Qwest and Level 3 is designed for DWDM systems.

Switching Technologies - TCP/IP
TCP/IP is actually a suite of protocols, describing all aspects of a communications network, to include issues ranging from physical connectivity to such an applications related concept as e-mail formatting. TCP is the so-called transport layer portion of the TCP/IP suite, it is responsible for assuring that messages sent over a TCP/IP network are correctly received on an end to end basis. IP is the network layer protocol, it is responsible for routing and congestion control in the network. 

TCP/IP networks are packet-switched. Long communications are broken into short packages called packets, and each is routed independently through the network to the destination, where reassembly of the message takes place. Packet switched networks are subject to variable delay, since a packet may sit in a buffer waiting for the next link in the path to the destination to become available. The operative word in any packet network is "shared," every link is shared on a first come-first served basis.

The TCP/IP suite of protocols was developed in a network called ARPANET, which was built with U.S. Government funding. As a consequence, TCP/IP is in the public domain, and implementations exist for virtually every type of computing device. Since TCP/IP is ubiquitous, and since adherence to the standard assures interoperability, it follows that a global network with massive connectivity could be built around it. Of course, this network has been built, and is called the Internet. Indeed, the Internet can be defined as the globally interconnected collection of networks that use the TCP/IP protocol suite.

Switching Technologies - Frame Relay
Frame relay (FR) is also a packet switching technology, but it differs from TCP/IP in a number of important ways. First of all, FR does not define all of the layers of protocol required for end to end interoperability. The FR protocol set only defines the procedures necessary to transport information across the "cloud," i.e., it does not address end to end issues.

FR differs from TCP/IP in an additional important way. FR is connection oriented. Unlike TCP/IP which treats each packet as an independent entity, FR requires a setup phase prior to any transmission. Subsequent packets then follow a predefined path to the destination. All links are still shared, but the path itself is constant. Such a network is often called a virtual circuit (VC) system. The emphasis in FR is on delivery throughput, as a consequence errors are viewed as an end to end problem. A FR network will typically discard any errored packet, relying on a higher level protocol to recover.

Because FR does not concern itself with higher layer issues, it can carry any form of information. Higher layer information is simply encapsulated in a "frame," and delivered to the far end for interpretation. It is this "adjustable wrench" characteristic that makes FR a popular technology, a FR network can carry all types of traffic, including voice and legacy IBM SNA data. Businesses often use FR to structure private data networks, termed "virtual private networks" or VPNs.

Switching Technologies - Asynchronous Transfer Mode
Asynchronous Transfer Mode (ATM) is a packet network technology, which like FR is connection-oriented. It only defines the lower protocol layers, and so it too can carry a variety of traffic types. ATM differs from all other packet technologies in that it uses fixed length packets; each packet must be 53 octets (bytes) long. The short packet length reduces a particular type of delay, but more importantly it allows certain functions such as routing to be performed by high-speed hardware, rather than by relatively slow software.

At setup, an ATM network engages in a negotiation with the user concerning the Quality of Service (QoS) required on the virtual circuit. If a user requests a QoS that the network cannot deliver, the connection request is simply denied. The quality of service characteristic of ATM networks, together with their very high speed switching capability makes this technology optimal for multi-media applications where voice, data and even video are interwoven in the information stream. ATM is often at the core of other packet switching networks, i.e., it is often the case that FR and TCP/IP networks actually use ATM switching inside the cloud to achieve high performance.