Electronic Communication

Technology And Impacts

Madeline M. Henderson, Madeline M. Henderson

• Publisher: Routledge
• Year: 2019
• ISBN:
  9780429728358
• Edition: 1
• Topic: Sciences sociales
• Subtopic: Sociologie

Electronic communications technology and services permeate every aspect of national life. This book examines the current and expected states of the technology and considers the societal impact and policy issues arising from new technological developments. Particular attention is paid to evaluation of computerized conferencing for enhanced communication among researchers in specialized and interdisciplinary fields and to technology assessments of criminal justice and tax administration systems.

Part 1
Current and Future States of Electronic Communications Technology

1. Perspective on Trends in Electronic Communication

Ira W. Cotton

Introduction

This review and perspective will concentrate on the area of electronic data communications, where major technological advances have been made in the past ten or fifteen years. Not only technological changes but relevant regulatory concerns are of interest to us, and will be covered here. Their impact on users will also be summarized.

Communications Channels

The history of technology in data communications has been based on the premise that bandwidth is expensive and the efforts of technology are to conserve bandwidth, to pack more traffic into lines of a given capacity. The first approach to sharing the capacity of a communications channel was to divide the channel into a fixed number of subchannels that could each be used by a single user. (Figure 1). That's known as fixed-division multiplexing, where the capacity of the channel is subdivided, generally with fixed logic or "hardwired," such that the total capacity of the carrier is approximately the sum of the capacities of the individual subchannels.

In actual use, when people interact with computer systems from interactive terminals, more often than not they are scratching their heads and thinking about what they are going to do rather than actually doing anything, so the traffic through these channels is intermittent or "bursty." We'd like

Figure 1. Multiplexing

Figure 2. Multiplexing Techniques

to be able to recover some portion of this idle time and put it to good use. But since each of those individual subchannels is unique and segregated from the others, when a user is idle on a subchannel that capacity is irrevocably lost.

There are two fundamental approaches that are used to implement this kind of fixed-division multiplexing. (Figure 2). In frequency-division multiplexing, which was just described, the bandwidth of the channel is sliced up into smaller bandwidth sections, each of which can be transmitting continuously. The other approach is more like what is done in computer time-sharing, where everybody gets the full capacity of the channel but for a limited period of time; that is, one user has the full bandwidth for a brief period of time, then the next user, and the next, then back to the first user again. This is called time-division multiplexing. But the situation still holds that if a user doesn't have anything to send at a particular point in time, that entire time slice stays empty, is unused and is lost. In addition, in this method the identification of data is implicit in the order in which they are sent; that is, every n-th chunk of data will belong to a particular user. In order to recapture unused capacity by skipping a user, some identification of whose data are being transported will have to be carried along with the data.

That is precisely what is done in concentration, also called adaptive (or statistical) multiplexing, where the channel capacity is divided in a variable rather than in a fixed way. (Figure 3). Under program control, with some intelligence in the communications system, the bandwidth is divided into continuously varying pieces or allocations. In this case, there has to be some way of knowing when a new portion of data begins and when it ends; in other words, we must be able to identify to whom the data belong. This introduces some degree of overhead, so that the total capacity of the channel is somewhat less than the information-carrying capacity of the logical subchannels. On the other hand, if a group of users is sharing a channel in this way, if any of them has anything to send there won't be any idle capacity, the channel will be occupied and there can be much more efficient utilization of the capacity.

When we compare the multiplexing and concentration approaches to sharing bandwidth, we see that the theoretical capacity of the multiplexer is really the full channel capacity because there is no overhead carried. If there is continuous traffic from a number of sources where the peak-to-average ratio is very low, that is, there is quite a regular flow of information, then a multiplexer is the most efficient way to share traffic. With concentration, the capacity of the carrier is reduced by the amount of overhead that is introduced,

Figure 3. Concentration: Adaptive (Statistical) Multiplexing

Figure 4. Physical Circuit Switching

so theoretically it is not as efficient as the multiplexer. What happens in actual fact is that the capacity of the multiplexer in operation is considerably reduced by any idle time from any of the transmitters. In the concentrator, only the overhead is lost and all of the idle capacity is regained, so that in most cases a concentrator is much more efficient than a multiplexer, particularly for interactive terminals, where the peak-to-average ratio is typically high.

Switching

Switching is the process of selecting a path for information to go from sender to a number of alternative receivers. In talking about physical circuit switching, our model is the voice telephone network in this country. (Figure 4). A simplified model of the phone system shows that when a number is dialed there is first a circuit setup phase, a path through the network is found, and that route is in some sense physically connected. (In the old days, a network of operators actually plugged in the long distance calls!) This circuit or path is fixed and is held for the duration of the call. New calls are blocked on overload; that is, when the system approaches saturation it refuses to accept new traffic. There is no overhead on data, there is a fixed bandwidth allocation and there is no delay in the switch. Those are the salient characteristics of this kind of switching technology.

What is done in the way of physical connecting of the switches in setting up a fixed path, can also be done with a minicomputer. (Figure 5). Rather than having the input line and the output line physically connected, they can be connected logically or virtually with a program in a minicomputer. A network in which the physical switches are replaced with stored program devices still has a circuit setup phase, then the route through the network is logically connected by tables in memory, rather than being physically connected by plug boards or cross-box relays. New calls are still blocked to prevent overload. One difference in this approach is there is probably some overhead on data in order to identify to whom the data belong. Bandwidth can be allocated when the call is placed, because the amount of time allocated to a user is a function of the program in the minicomputer, and different amounts of bandwidth can be allocated.

Another significant difference between physical circuit switching and virtual circuit switching is that in the latter case data are stored in the switch: a block of data comes in, it is stored in memory while a decision is made on how to handle it and then it is re-transmitted. This stored delay is cumulative so for longer distances the data are increasingly delayed in transmission from sender to recipient.

Figure 5. Virtual Circuit Switching

Figure 6. Message Switching

Given a minicomputer in place, one is not constrained to have the data follow the same path all the time, so we can progress to what is called message switching. (Figure 6). Here the circuit setup is abolished, and data can be routed through this network on a block-by-block basis according to the address. There's no holding time in this kind of a system, that is, there is no fixed route that's reserved for an individual user that would serve to block other users. When the system becomes overloaded it simply stores the data in memory; service will gradually degrade to all users as the system becomes more and more heavily used. There may be considerable overhead due to the addressing information and the functioning of the routing program, but the bandwidth is continuously adaptable. So long as the system is not saturated, users get as much or as little bandwidth as they need and the system can adapt to selective component failure. If a switching center or a circuit fails, data can be sent through alternate routes quite easily to the ultimate destination.

Packet switching is a subset of message switching with a number of characteristics, no single one of which is sufficient to completely discriminate between the two, but which when taken together describe a quite different kind of service. First of all, the block length or the transmission slice is kept relatively small, perhaps no more than a hundred characters, for example. The switch is designed for forwarding, not for storing: people talk about "hot potato routing," wherein a packet of data comes in and has to be either passed on immediately or thrown away. There is no backup storage, no extensive, records of traffic, and again, input is refused to avoid overload. When the system becomes saturated, information is simply thrown away, undeliverable packets are discarded, and it is up to the sender and the receiver to maintain their own control scheme to guarantee the receipt of information in its proper sequence.

Now a data communications network might be called upon to carry different kinds of traffic: interactive (transaction) traffic, with a user at a terminal, intermittent or bursty flow and, since users of terminals don't like to wait, the requirement for rapid response from the system. In contrast, bulk traffic or remote batch entry involves large jobs that may be transmitted at night, that have a continuous moderate-to-high bandwidth demand, and lessened response requirements so that bulk traffic is much easier to engineer for than interactive traffic. The final category of real-time traffic combines the most perverse characteristics of the first two: high bandwidth bulk traffic also requiring high response, the kind of traffic in command and control systems, in sensor applications and laboratory automation. Data that cannot be delivered in an adequate time period are probably useless for the application

Let's consider how these types of traffic can be handled by circuit switching (either physical or virtual) or packet switching. Interactive traffic is somewhat unsuited for circuit switching and that's really what led to the concentration and multiplexing schemes described earlier. Without a well-defined multiplexing scheme, there is going to be a lot of wasted bandwidth. Packet switching can have potentially high overhead but there is not going to be any idle bandwidth. It turns out that packet switching is ideal for terminal-to-computer traffic where people are occasionally transmitting messages through a computer system. Bulk traffic on the other hand is most suitable for circuit switching; with a lot of continuous information to send, utilization of the line is quite good. The suitability of bulk traffic for packet switching is really a function of price. With current pricing, it is most cost effective for moderate quantities over long distances; over short distances you are better off dialing a call. Over long distances or with lots of information, you are better off leasing a line. Real-time traffic is very suitable in the circuit switching case with dedicated facilities: a leased line, guaranteed bandwidth and certainty that the transmitted data will get through. There have been experiments in the research community to carry real-time traffic over packet switching systems; it is only possible with some new control mechanism added to the network to give certain packets priority over others, to guarantee that they get through in a reasonable period of time.

Regulatory Background

For data communications, one of the first significant regulatory proceedings in this country came, back in the late 1960's and early 1970's, when the Federal Communications Commission (FCC) considered the issue of specialized common carriers. The question on the FCC docket asked specifically whether carriers other than the established ones could offer competitive services. It took eight years for the FCC to finally set the course for free competitive entry into selected areas of the communications marketplace. MCI was permitted to erect its own microwave transmitter between Chicago and St. Louis; based on this decision, there have been a substantial number of new entrants offering different kinds of service in the communications area in this country.

Another significant issue is the interdependence of computers and communications. Here the question ...