On Distributed Communications Series

IV. Priority, Precedence, and Overload

I. Introduction

Simulation studies of switched communications network vulnerability, such as the RAND NATCOM model,[1] invariably include the assumption that any surviving link can carry all the traffic generated in the residual network. Since such vulnerability models have been so widely adopted, it is appropriate to consider some of the implications of the effects of overload on communications systems that are often ignored in the interest of simplification.

In this Memorandum we briefly touch upon the over-load problem in present-day networks, reserving the bulk of our comments for improvements to a future system. Though this may appear to be ignoring the present in favor of a system that does not exist, the suggestions raised are intended primarily for a far-future time-frame. The general principles, however, might be used in such present-day problems as determining the order of program handling in a time-shared command-control computer.

In the process of describing the methods for automating the handling of mixed traffic of different levels of importance, data rate, and priority, simple mechanisms are mentioned which create an illusion of mechanized "judgment." Nothing magical is proposed; merely that a small set of executive policies can be automatically executed, with retention of the right to rapidly change the weights of these policies by human intervention. In essence, management's fundamental law of exception is mechanized so as to unburden humans processing information--nothing more.

We are dealing with two separate types of overload: terminal and network. If a called party is busy on the telephone, it may be said that the call is not completed because of an overload at the terminal. If the call is blocked by an overload at an intermediate switching center, or all circuits are busy, then the network is said to be overloaded.

There are two general categories of communications networks: store-and-forward, and line-switched. A torn-tape telegraph switching center which stores messages until a desired circuit is open is an example of the store-and-forward network. The conventional telephone network, which closes switches to provide a "real-time" path from user to user, is a line-switched network.

The overload phenomena have much in common, whether they be terminal or network, in store-and-forward or lineswitching networks, and can be discussed on a general conceptual basis.

[1] Eldrige, F. R., The Effectiveness of Command Control in Strategic Operations for the Mid-Sixties (U), The RAND Corporation, RM-3152-PR, October 1962.

Reinertsen, R. W., The IBM 704 Computer Communications Vulnerability Model, The RAND Corporation, S-135, August 15, 1960.

Next chapter