On Distributed Communications Series

V. History, Alternative Approaches, and Comparisons

Contents
Preface
Summary

I. Introduction

The term "distributed networks" is commonly used as a broad generic term encompassing a number of communications systems with different properties. A brief history of the development of "distributed networks" is provided in the following sections, with emphasis on the development of "heuristic routing" doctrines that seek "perfect switching"; i.e., those able to find "best" surviving paths in a heavily damaged network. The discussion is restricted to those systems using "locally" implemented switching rules, and which do not need a single, highly critical, control center.

Such systems have different properties, require different equipment, and exhibit different survivability characteristics. But, they can be factored into categories for comparison.

The work of at least six separate disciplines is germane, but inasmuch as the members of these separate disciplines reside in different communities of interest, the all too common problem of a lack of communication between one another is quite noticeable. And, when there is an interchange of ideas, the words used take on different meanings to the representatives of the various fields of study. The term "distributed networks" is one of these very ambiguous entities. Sometimes it is simply a meaningless, "ok" word used to spice up hardware proposal brochures. To others, the meaning is exact, and narrow.

One writer has defined distributed switching as any switching medium noncollocated with the communications subscriber; another has implied that collocation is almost a necessary condition in a distributed system. Though these two definitions contain explicit delineations, they are at opposite ends of a spectrum. In our usage, collocation has no bearing whatsoever in determining whether or not a system is "distributed."

It might be helpful to enumerate the fields concerned in these studies. Distributed networks, of one sort or another, are of interest to:

  1. Those concerned with "artificial intelligence";

  2. Those concerned with communications within organisms and organizations;

  3. Mathematicians working with optimization of flow in networks;

  4. Mathematicians using dynamic programming to optimize incompletely understood and changing systems;

  5. Those connected with civilian common carrier telephone plant switching;

  6. Military systems planners, especially those dissatisfied with existing communications network techniques.

The present Memorandum is written primarily from the viewpoint of the latter group. And, while significant contributions have come forth from each and every one of the above listed disciplines, the present work has taken its examples principally from projects concerned with study of military systems.

Just as there are various definitions of "distributed networks," there are many ways to build such networks; in fact, there are probably as many ways of building them as there are designers. Therefore, several properties have been factored out, and our attention will be focused on distributed networks which possess these characteristics:

  1. A certain degree of routing flexibility;

  2. Capability for implementation by real-world hardware at a "reasonable" price;

  3. Compatibility with existing input devices;

  4. Survivability in an unfriendly environment;

  5. Capacity for orderly system growth;

  6. Operational routine of sufficient simplicity--that is, one that can be maintained by reasonably intelligent personnel.
Design of such systems tends to be quite complex, and the systems themselves are inherently intolerant of poor design, due to the practical difficulties encountered in preventing or minimizing propagation of system malfunctions. But, the potential promise of a high degree of survivability and the potential use of low-reliability elements makes the labor of a complex design worthwhile.

The switching criteria will receive considerable attention since survivability is a function of switching flexibility. The resulting detailed inspections of system proposals are necessary to accurately determine survivability properties.

The following section is devoted to a presentation of the distributed network concept and its characteristics, while Sections III and IV detail some of the history of distributed networks, both in theoretical studies and experiments and in hardware implementation proposals. To those interested in the evolution of modern systems and proposed systems, Appendix A contains several charts which summarize the stages of this development, using specific proposals as signposts along the evolutionary path.

II. The Distributed Network Concept
III. Early History
IV. Specific Hardware Proposals
V. Conclusions
Appendix A. Summary Charts
Appendix B. The DDD System
List of Publications in the Series