Abstract :
Two developing technologies that have become prominent in applied systems architectures are parallel processing and distributed networks. However, the integration of parallel computing devices within commercial and industrial networks is an emerging field with opportunities that have not been sufficiently exploited. One of the reasons for this lag can be found in the conventional thinking about the capabilities of parallel processing. Another has been the perceived lack of adequate communication links to handle real-time parallelism over the distances typical for local area networks. This and other implementation difficulties can be overcome quite efficiently through the use of the transputer family of microprocessors with the universal INMOS serial link. With such a transputer based system, one has the advantage of a hardware and software environment that is able to take immediate advantage of the traditional virtues of parallel processing at the individual node level, including the ability to expand the number of processors at a node without affecting the performance of other nodes or requiring major changes to the programs operating at the affected node. The net result of such a design, affords the designer and operator of a wide-area network the capabilities of run-time reconfiguration and fault tolerance, combined with relative simplicity in the architecture and in the operational requirements
