Read an Excerpt
Chapter 1: Evolution of Data NetworksThe historical perspective in this chapter will help you understand where and why improvements were made as data networks have evolved and scaled. This chapter also illustrates a more salient point: The use of ideas in development is often cyclic. This cyclical nature of development places a whole new meaning on the so-called "wheel of invention."
Improvements in technology enable you to renew your outlook on solutions that were previously dismissed as technically or economically unfeasible. An extreme example is the use of optical communications. Although successful many years ago in the form of torches and flares, in the last century, optical communication was disbanded because it was much easier to guide lower-frequency electromagnetic radiation using simple copper cable. With the advent of optical fiber, however, this is no longer true. The switching of data packets also has historically been accomplished in the electrical domain for similar reasons. However, recent innovations in wave division multiplexing and optical networks soon may obviate the need for electronic switching in many highbandwidth communication infrastructures.
Overview of Communications History
Networks are now a core component of our business and personal lives. Today, businesses that may hobble along with the loss of telephone service can be rendered nonfunctional by the loss of their data network infrastructure. Understandably, corporations spend a great deal of time and money nursing this critical resource.
How and why did this dependency occur? Simply because networks provide a means to amplify all the historical communication mechanisms.Nearly 50,000 years of speech, at least 3500 years of written communication, and many thousands of years of creating images all can be captured and communicated through a network to anywhere on the planet.
In the 1980s, fiber optics improved the distance, cost, and reliability issues; CB radio taught us about peer-to-peer communication and self-regulation without a centralized communications infrastructure provider. The needs of the military led to the development of network technologies that were resilient to attack, which also meant that they were resilient to other types of failure. Today's optical switching and multiplexing again demonstrate that to take the next leap in capability, scalability, and reliability, businesses and individuals cannot afford to cling to tried-and-true techniques.
Data communications grew from the need to connect islands of users on LANs to mainframes (IBM's Systems Network Architecture [SNA] and Digital's DECnet) and then to each other. As time passed, these services were required over wide geographical areas. Then came the need for administrative control, as well as media and protocol conversion. Routers began to become key components of a network in the 1980s, which was within the same time that Asynchronous Transfer Mode (ATM) cell switching was being developed as the technology for the deployment of worldwide networks supporting multimedia communications.
The designers of ATM were constrained by the need to support the traditional voice network. This is not surprising, however, for at that time voice revenues exceeded other forms of communications infrastructures. If new applications were to develop, many people thought they were likely to take the form of video, either for communication or entertainment.
Very few people predicted the coming of the Internet. After all, it was not real-time voice or video that stole the show, but the ubiquity of home personal computers coupled with a few applications. These included the simple one-to-one or one-to-many communication applications, such as e-mail and chat groups, and the powerful Web browsers and Internet search engines that turned the Internet into a virtual world in which people could journey, learn, teach, and share. Users did not need megabits per second to enter this world: 32 Kbps was happiness, 64 Kbps was bliss, and 128 Kbps was heaven.
The increases in desktop computing power and in peer-to-peer applications had a fundamental impact on network architectures. Modem network architectures expect intelligence and selfregulation at the user workstation, and they provide a network infrastructure that maintains only the intelligence sufficient to support packet forwarding. This contrasts significantly with the approach of connecting simple terminal devices to intelligent mainframes using a complex networking device that is used by many proprietary solutions, notably IBM's SNA.
The impact of technological development, and the changing needs of businesses, consumers, and society in general, is clear. Data networking is growing at 25 percent per year, traditional voice is increasing by only 6 percent, and the Internet is doubling every few months. (The term traditional voice is used because the Internet now carries voice, and in the last year or so several providers have announced their intent to supply voice over IP services.)
The revenue to be gained from carrying data packets is close to, or perhaps even exceeds, that of carrying traditional telephone voice circuits. Voice is rapidly becoming just another variation on the data-networking theme-just packets for another application.
Ultimately, competition drives design and development, whether it be the telegraph versus the telephone, or traditional telephony versus Internet telephony. Providers attempt to gain a commercial advantage over their competitors by adopting new technologies that will provide a service that is either cheaper, better, or more flexible. Naturally, a network that is well designed, planned, and implemented will be in a position to take on new technologies.
Evolution of the Internet
Socially, economically, culturally, and technologically, for many of us the Internet already has changed our lives dramatically. For many more of us, it soon will. Along with telephones, televisions, and automobiles, Internet connectivity is rapidly becoming a commodity in every home. Yet, as dramatic as the changes have been, it is worth remembering that-initially, at least-the Internet grew at a considerably slower pace than the telephone network (although some would argue that this is merely because the former is dependent on the latter).
In the 1960s, the idea of a ubiquitous communication network was not new, but the angle of using the network for more than just personal communications-and, in particular, for the exchange of computer programs and other forms of arbitrary data-was fairly radical. For one thing, communications technology at the time was not flexible enough to allow it to happen. Then, ARPANET entered the picture.
Technology researchers working independently at MIT, RAND, and NPL from 1961 through 1967 conducted a number of experiments in what was later termed packet networking. One of the papers about those experiments, published in 1967, was a design for an experimental widearea packet-switched network, called the ARPANET.
This was the technological turning point. It would have a profound impact on the capability of networks to grow and evolve to meet the changing needs of their users-and, in particular, to support the world of peer-to-peer networking and multimedia communications. While replacing the traditional time-division multiplexing in the LAN environment, ARPANET provided the capability, through layering, of exploiting the benefits of TDM systems in a wide area, and seamlessly interconnecting the two.
In 1969, after several years of observing this technology in the lab, the U.S. Department of Defense commissioned ARPANET, connecting four nodes from SDS, IBM, and DEC at 50 Kbps rather than the originally planned 2.4 Kbps. The Information Message Processors (IMPs) used in the network were Honeywell 516 minicomputers, with a whopping 24 KB of memory, with code supplied by BBN Inc. These were followed by BBN C-30s and C-300s, and subsequently were renamed Packet Switch Nodes in 1984. Aside from its historical significance, the ARPANET was characterized by two traits that remain true of the Internet to this day: The end hosts came from different manufacturers, and the initial bandwidth proposed was far less than necessary...