Modelling the influence of unreliable software in distributed computer systems

The author develops a system dependability model accounting for the influence of software faults. When one software system resides on a set of distributed computers or processing units, it is necessary to comprehend the logical linkage between system units. The model is based on the concept that a logical fault in a system will cause an inconsistent (erroneous) internal state in the processing unit where the fault is activated, and that this error can cause a failure of the processing unit, be corrected without causing a failure, or propagate to other cooperating processing units. The model also accounts for the logical hardware faults and the transient faults, which have a similar pattern of manifestations. Analysis of the stationary dependability characteristics of distributed systems is discussed. An approximate procedure which allows evaluation of fairly large systems is outlined. A brief example is presented which demonstrates that even a low error propagation may have large performance consequences.<
>