Multiplexers, often called muxes, are extremely important to telecommunications. Their main reason for being is to reduce network costs by minimizing the number of communications links needed between two points. As with all other computing systems, multiplexers have evolved. Each new generation has additional intelligence, and additional intelligence brings more benefits. The types of benefits that have accrued, for example, include the following:
The capability to do data compression so that you can encode certain characters with fewer bits than normally required and free up that additional capacity for the movement of other information.
The capability to do error detection and correction between the two points that are being connected to ensure that data integrity and accuracy are being maintained.
The capability to manage transmission resources on a dynamic basis, with such things as priority levels. If you have only one 64Kbps channel left, who gets it? Or what happens when the link between San Francisco and Hong Kong goes down? How else can you reroute traffic to get the high-priority information where it needs to go? Multiplexers help solve such problems.
The more intelligent the multiplexer, the more actively and intelligently it can work on your behalf to dynamically make use of the transmission resources you have.
When you're working with network design and telecommunications, you need to consider line cost versus device cost. You can provide extremely high levels of service by ensuring that everybody always has a live and available communications link. But you must pay for those services on an ongoing basis, and their costs become extremely high. You can offset the costs associated with providing large numbers of lines by instead using devices such as multiplexers that help you make more intelligent use of a smaller group of lines.
Figure 2.12 illustrates a network without multiplexers. Let's say this network is for Bob's department stores. The CPU is at Location A, a data center that's in New York that manages all the credit authorization functions for all the Bob's stores. Location B, the San Francisco area, has five different Bob's stores in different locations. Many customers will want to make purchases using their Bob's credit cards, so we need to have a communications link back to the New York credit authorization center so that the proper approvals and validations can be made. Given that it's a sales transaction, the most likely choice of communications link is the use of a leased line from each of the locations in San Francisco back to the main headquarters in New York.
Figure 2.12 A network without multiplexers
Remember that the use of leased lines is a very expensive type of network connection. Because this network resource has been reserved for one company's usage only, nobody else has access to that bandwidth, and providers can't make use of it in the evenings or the weekends to carry residential traffic, so the company pays a premium. Even though it is the most expensive approach to networking, the vast majority of data networking today still takes place using leased lines, because they make the network manager feel very much in control of the network's destiny. With leased lines, the bandwidth is not affected by sudden shifts of traffic elsewhere in the network, the company can apply its own sophisticated network management tools, and the network manager feels a sense of security in knowing who the user communities are at either end of that link. But leased lines have another negative attribute: They are mileage sensitive, so the longer the communications link, the higher the cost. And in a network that doesn't efficiently make use of that communications link all day long, leased lines become overkill and an expensive proposition.
The astute network manager at Bob's tries to think about ways to make the network less expensive. One solution, shown in Figure 2.13, is to put in multiplexers. Multiplexers always come in pairs, so if you have one at one end, you must have one at the other end. They are also symmetrical, so if there are five outputs available in San Francisco, there must also be five inputs in the New York location. The key savings in this scenario comes from using only one leased line between New York and California. In San Francisco, short leased lines, referred to as tail circuits, run from the centrally placed multiplexer to each of the individual locations. Thus, five locations are sharing one high-cost leased line, rather than each having its own leased line. Intelligence embedded in the multiplexers allows the network manager to manage access to that bandwidth and to allocate network services to the endpoints.
Figure 2.13 A network with multiplexers
Various techniquesincluding Frequency Division Multiplexing (FDM), Time Division Multiplexing (TDM), Statistical Time Division Multiplexing (STDM), intelligent multiplexing, inverse multiplexing, and Wavelength Division Multiplexing (WDM)/Dense Wavelength Division Multiplexing (DWDM)enable multiple channels to coexist on one link. The following sections examine each of these techniques.
FDM is an environment in which the entire frequency band available on the communications link is divided into smaller individual bands or channels (see Figure 2.14). Each user is assigned to a different frequency. The signals all travel in parallel over the same communications link, but they are divided by frequencythat is, each signal rides on a different portion of the frequency spectrum. Frequency, which is an analog parameter, implies that the type of link you see with FDM is usually an analog facility. A disadvantage of frequency division muxes is that they can be difficult to reconfigure in an environment in which there's a great deal of dynamic change. For instance, to increase the capacity of Channel 1 in Figure 2.14, you would also have to tweak Channels 2, 3, and 4 to accommodate that change.
Figure 2.14 FDM
If an enterprise has a high degree of moves, additions, and changes, FDM would be an expensive system to maintain because it would require the additional expertise of frequency engineering and reconfiguration. Given the environment today, we don't make great use of FDM, but it is still used extensively in cable TV and in radio. In cable TV, multiple channels of programming all coexist on the coax coming into your home, and they are separated based on the frequency band in which they travel. When you enter a channel number on your set-top box or cable-ready TV, you're essentially indicating to the network what portion of the frequency band it's traveling on.
The second muxing technique to be delivered to the marketplace was TDM. There are various levels of TDM. In the plain-vanilla TDM model, as shown in Figure 2.15, a dedicated time slot is provided for each port or point of interface on the system. Each device in a predetermined sequence is allotted a time slot during which it can transmit. That time slot would enable one character of data, or 8 bits of digitized voice, to be placed on the communications link. The allocated time slots have to be framed in order for the individual channels to be separated out. A problem with a standard time-division mux is that there is a one-to-one correlation between each port and time slot, so if the device attached to Port 2 is out for the day, nobody else can make use of Time Slot 2. Hence, there is a tendency to waste bandwidth when vacant slots occur because of idle stations. However, this type of TDM is more efficient than standard FDM because more subchannels can be derived.
Figure 2.15 TDM
FDM and TDM can be combined. For example, you could use FDM to carve out individual channels and then within each of those channels apply TDM to carry multiple conversations on each channel. In fact, this is the way that some digital cellular systems work (for example, Global Systems for Mobile Communications [GSM]). Digital cellular systems are discussed in Chapter 14, "Wireless Communications."
STDM was introduced to overcome the limitation of standard TDM, in which stations cannot use each other's time slots. Statistical time-division multiplexers, sometimes called statistical muxes or stat muxes, dynamically allocate the time slots among the active terminals, which means that you can actually have more terminals than you have time slots (see Figure 2.16).
Figure 2.16 STDM
A stat mux is a smarter mux and it has more memory than other muxes, so if all the time slots are busy, excess data goes into a buffer. If the buffer fills up, the additional access data gets lost, so it's important to think about how much traffic to put through the stat mux to ensure that performance variables are maintained. By dynamically allocating the time slots, you get the most efficient use of bandwidth. Additionally, because these are smarter muxes, they have the additional intelligence mentioned earlier in terms of compression and error-control features. Because of the dynamic allocation of time slots, a stat mux is able to carry two to five times more traffic than a traditional time-division mux. But, again, as you load the stat mux with traffic, you run the risk of delays and data loss occurring.
Stat muxes are extremely important because they are the basis on which packet-switching technologies (for example, X.25, IP, Frame Relay, ATM) are built. The main benefit of a stat mux is the efficient use of bandwidth, which leads to transmission efficiencies.
An intelligent multiplexer is often referred to as a concentrator, particularly in the telecom world. Rather than being a device used in pairs, it is used as a singular device, a line-sharing device whose purpose is to concentrate large numbers of low-speed lines to be carried over a high-speed line to a further point in the network. A good example of a concentrator is in a device called the digital loop carrier (DLC), which is also referred to as a remote concentrator or remote terminal. In Figure 2.17, twisted-pairs go from the local exchange to the neighborhood. Before the advent of DLCs, you needed a twisted-pair for each household. If the demand increased beyond the number of pairs you had available out of that local exchange, you were out of luck until a new local exchange was added.
Figure 2.17 Intelligent multiplexing: Concentrators
With digital technology, you can make better use of the existing pairs. Instead of using each pair individually per subscriber from the local exchange to the subscriber, you can put a DLC in the center. You use a series of either fiber-optic pairs or microwave beams to connect the local exchange to this intermediate DLC, and those facilities then carry multiplexed traffic. When you get to the DLC, you break out the individual twisted-pairs to the household. This allows you to eliminate much of what used to be an analog plant leading up to the local exchange. It also allows you to provide service to customers who are outside the distance specifications between a subscriber and the local exchange. So, in effect, that DLC can be used to reduce the loop length.
Traditional DLCs are not interoperable with some of the new DSL offerings, including ADSL and SDSL. For example, about 30% to 40% of the U.S. population is serviced through DLCs. And in general, globally, the more rural or remote a city or neighborhood, the more likely that it is serviced via a DLC. For those people to be able to subscribe to the new high-bandwidth DSL services, the carrier will have to replace the DLC with a newer generation of device. Lucent's xDSL Access Gateway, for example, is such a device; it offers a multiservice access system that provides Lite and full-rate ADSL, Integrated Services Digital Network (ISDN), Asynchronous Transfer Mode (ATM), and plain old telephone service (POTS)/analog over twisted-pair lines or fiber. Or, the carrier may simply determine that the market area doesn't promise enough revenue and leave cable modems or broadband wireless as the only available broadband access technique. But the nature of a concentrator is that it enables you to aggregate numerous low-speed residential lines and to multiplex them onto high-bandwidth facilities to pass off to the local exchange.
The inverse multiplexer arrived on the scene in the 1990s. It does the opposite of what the multiplexers described so far do. Rather than combine lots of low-bit-rate streams to ride over a high-bit-rate pipe, an inverse multiplexer breaks down a high-bandwidth signal into a group of smaller-data-rate signals that can be dispersed over a range of channels to be carried over the network. A primary application for inverse multiplexers is to support of high-bandwidth applications such as videoconferencing.
In Figure 2.18, a videoconference is to occur at 1.5Mbps. For a good-quality,
or PRI ISDN. (PRI ISDN is discussed in Chapter 3, "Transmission Media: Characteristics and Applications.") The channels are carried across the network separately. At the destination point, a complementary inverse mux again reaggregates, resynchronizes, and reproduces that high-bandwidth signal so that it can be projected on the destination video monitor.
Figure 2.18 Inverse multiplexing
Inverse multiplexing therefore allows you to experience a bit of elastic bandwidth. You can allocate existing capacity to a high-bandwidth application without having to subscribe to a separate link just for that purpose.
WDM was specifically developed for use with fiber optics. In the past, we could use only a fraction of the available bandwidth of a fiber-optic system. This was mainly because we had to convert the optical pulses into electrical signals to regenerate them as they moved through the fiber network. And because repeaters were originally electronic, data rates were limited to about 2.5Gbps. In 1994, something very important happened: optical amplifiers called erbium-doped fiber amplifiers (EDFAs) were introduced. Erbium is a chemical that's injected into the fiber. As a light pulse passes through the erbium, the light is amplified and continues on its merry way, without having to be stopped and processed as an electrical signal. The introduction of EDFAs immediately opened up the opportunity to make use of fiber-optic systems operating at 10Gbps.
EDFAs also paved the way to developing wavelength division multiplexers. Before the advent of WDM, we were using only one wavelength of light within each fiber, whereas the visible light spectrum engages a large number of different wavelengths. WDM takes advantage of the fact that multiple colors or frequencies of light can be transmitted simultaneously down a single optical fiber. The data rate that's supported by each of the wavelengths depends on the type of light source. Today, we have light sources that operate at a rate of OC-48, which is shorthand for 2.5Gbps. We have light sources that operate at OC-192, which is equivalent to 10Gbps. And there are systems in trial that operate at OC-768, offering 40Gbps per wavelength. In the future, we'll go beyond that. Part of the evolution of WDM is that every year we double the number of bits per second that can be carried on a wavelength, and every year we double the number of wavelengths that can be carried over a single fiber. But we have just begun. Soon light sources should be able to pulse in the terabits per second range, and in five years, light sources should pulse in the petabits per second (1,000Tbps) range.
One thing to clarify about the first use of WDM is that unlike with the other types of multiplexing, where the goal is to aggregate smaller channels into one larger channel, WDM is meant to furnish separate channels for each service, at the full data rate. Increasingly, enterprises are making use of new high-capacity switches and routers that are equipped with 2.5Gbps interfaces, so there's a great deal of desire within the user community to be able to plug in to a channel of sufficient size to carry that high-bandwidth signal end-to-end, without having to break it down into smaller increments only to build them back out at the destination.
WDM furnishes a separate channel for each service at the full rate; you cannot aggregate smaller channels into one large channel. Systems that support more than 16 wavelengths are referred to as DWDM (see Figure 2.19). Systems at the OC-48
Figure 2.19 WDM and DWDM