030.50.1 Gigabit Networks Robert E. Kahn* <kahn@nri.reston.va.us>
A small number of research sites will soon be gaining limited access to experimental networks which support end-end user communication rates on the order of a gigabit per second (i.e. a billion bits per second). While many technical problems must be solved before wide-area common-user networks are available with this capability, it cannot be assumed that they will simply happen because all the technical problems are solved. High speed "network backbones" have been cost effective for multiplexing large numbers of lower speed users. Backbone speeds on the order of Terabits per second and higher will be necessary to support a large user community with gigabit access speed requirements in a wide-area common user network. Many factors, including sustained throughput requirements, acceptable levels of delay, the need for real-time delivery guarantees and overall size of the market will surely be taken into account in determining the price of gigabit network services.
Quantitative changes often lead to qualitative changes when the magnitude of change is sufficiently large. Substantial investment is being made in the development of 1) gigabit technology for computing and communications and 2) lightwave technology for all- optical networking at much higher speeds. However, a critical concern is identifying and developing applications that require or can make effective use of such high data rates to the end user. For over twenty years, network users have relied principally on text-based forms of communication due to limitations in speed of existing computer communication networks to the range of megabits per second and less. With the advent of common user networks operating at speeds on the order of a gigabit per second or more, new visual forms of communication becomes practical including the regular use of images, graphics and real-time video. End-user data rate requirements can also approach gigabit per second speeds with real-time high-resolution large screen display systems. The impact on the user of such visual communication systems may be hard to quantify, but the practical result will be a radical alteration of the relationship between a user and data output from a computer or network. Visualizing the pictorial output of a simulation can take place in a fraction of a second, while scanning reams of printed computer data may never afford a clear picture of the results no matter how much time is taken to absorb it.
High Performance Computers may also need to interact with each other at gigabit per second speeds. This is currently a subject of considerable interest within the scientific community. The value of such high speed computer interactions, once they have been demonstrated, may eventually be of interest to many other sectors of society.
Other uses of gigabit networks may not lead to qualitative change immediately. It is clear that higher speed networks will enable computers to exchange files more quickly, but this may not necessarily produce any real or perceived benefit to the end user. In those cases where storage exists to hold extremely long files and databases, one-time transfers of large data sets on demand or by pre-arranged distributions of these data sets at selected user sites on a network may be desirable. However, these transfers need not necessarily occur at gigabit per second speeds and remote use of the files may even be possible initially. The speed of such network transfers can be technologically impressive and may not disadvantage the application, but it may be unnecessary in many cases as well as unaffordable.
In weighing the value of gigabit networks to the end user, one must include the basic economics, the competitive value, the qualitative value and the changes (real and perceived) wrought by such new technological capabilities. For many applications and users, lower speed network alternatives may well suffice. However, as technology matures and advances, the relative cost of gigabit technology and services can be expected to decrease and we should ultimately see the widespread deployment of the technology to all sectors of society.
Gigabit technology will be expensive to build and deploy for many more years. When the technology is widely available and networks are deployed with end-end gigabit speeds, users can begin to explore possible new uses and applications that may create a significant marketplace for gigabit services. But, without an existing or nascent market for gigabit speeds to the end-user, industry will find it difficult to justify many of the requisite initial investments. We plan to explore various aspects of this "chicken-egg" problem along with other matters relating to gigabit networks in future columns.
* President, Corporation for National Research Initiatives, Reston, Virginia, USA.
