An interesting phenomenon over the past several years has been the spontaneous growth of interest in multiple-microprocessor computer systems in many universities and research laboratories. This interest is not hard to understand given the inexpensive computational power offered by microprocessors today and the cost-performance improvements promised by those to be delivered in the near future. Microprocessors have had a dramatic impact on applications that require a small amount of computing. They have been used in instruments, industrial controllers, intelligent terminals, communications systems as special function processors in large computers, and, more recently, in consumer goods and games.

The question naturally arises as to whether the microprocessor, which has proved so successful in these diverse applications, can be used as a building block for large general purpose computer systems. In other words, can a suitably interconnected set of microprocessors be used for tasks that currently require large uniprocessors capable of executing millions of instructions per second? At present, there is no definitive answer to this question, but there are several reasons to believe that multiple-microprocessor systems might indeed be viable.

A strong argument for a microprocessor-based system is its potential cost-effectiveness. This point is graphically demonstrated in Figure 1 which shows cost/performance as a function of computer system size.* Each point in this figure represents a (uniprocessor) system currently available and introduced between 1975 and 1977 [GML Corp., 1977]. For example, the computer represented by the point labelled A has a purchase price of about $10,000. It is capable of transferring data between memory and the central processor at about 200 Mbits/second, yielding a figure of merit of 2 X l0⁴ bits/second/dollar. The figure shows that with conventional methods of organizing computers, the cost/performance of a

_______________

*The measure of system size used here is its purchase price.