On describing the behavior and implementation of distributed systems

Abstract

A simple, basic and general model for describing both the (input-output) behavior and the implementation of distributed systems is presented. An important feature of the model is the separation of the machinery used to describe the implementation and the behavior. This feature makes the model potentially useful for design specification of systems and of subsystems.

The implementation model relies on the basic notions of process and variable, assuming indivisibility of variable access. Long-distance communication is modelled by a special process representing a "channel." Process executions are considered to be completely asynchronous; this consideration is reflected in the fairness of the operations for combining processes. The primitivity and generality of the model make it an apparently suitable basis for cost comparison of various message-passing protocols and other higher-level programming constructs, as well as of complex distributed system implementations.

A system's (input-output) behavior is modelled by a set of finite and infinite sequences of actions, each action involving access to a variable.

Basic definitions, examples and characterization results are given. An extended example, involving specification and implementation of an arbiter system, is presented. For this example, equivalent implicit and explicit specifications are given. Several different implementations are described, each of which exhibits the required behavior.

General remarks are made about cost comparison of distributed system implementations.

This research was supported in part by the National Science Foundation under grants MCS77-02474 and MCS77-15628.