In this paper, the authors report on an ongoing project involving the implementation of a multimicroprocessor system at the Brussels Free University. The geographical distances among various nodes and nonuniform characteristics seem to have warranted the design of a hybrid solution incorporating a mixture of broadband, baseband, and wide-area networks.
The design rationale is described as the following:
(1) A set of network interface modules will be loosely coupled via an existing high-band network.
(2) The network interface consists of multimicroprocessor systems tightly coupled via a memory-bus.
(3) The software environment is based on MODULA 2. There are several concurrency primitives for communication among local processors and their bus-master, as well as for communication between different bus-masters. In addition, the bus-masters will be provided with a distributed operating system kernel which will be accessible from the other processors as a result of the usage of a set of Definition Modules.
(4) The communication interface is hoped to be universal.
(5) In order to provide privacy protection and configurability, each part is connected with a Network Administration Server which will implement all of the configuration and validation functions.
The software interface is based on abstract data typing feature of MODULA 2; these are provided in the form of coroutines. For example, “TRANSFER (Process1, Process2)” transfers control from Process1 to Process2. This is viewed as a coroutine crossover and implements a process context switch. In the network interface configuration, basic interface consists of: (1) a processor board with the monitor in ROM (bus-master), and (2) slots for CPU/peripheral processor boards. The ROM-resident monitor will contain a set of basic functions; for example, RequestBuffer, ReleaseBuffer, TransmitBuffer, ReceiveBuffer, etc.
A reader looking for details will be disappointed. I assumed that the lack of such could be attributed to the limitation of space, or it could be due to the fact that the project is still in the developing stage. The list of references provided is not satisfactory--out of five, four references are on MODULA, MODULA 2, *MOD, and EDISON. Some words are consistently misspelled (e.g, “equiped”); and broken words (e.g., “a n,” “commni cation”) are annoying. This paper may be of interest to a practitioner in or an observer of the areas of distributed system design.
