Performance evaluation of testing strategies in parallel systems

This paper presents an evaluation tool, SimDiag, developed in order to allow quantitative analysis of distributed diagnosis algorithms. This tool uses a process-based model for simulating diagnosis algorithms. It allows the evaluation of many interesting diagnosis performance characteristics such as diagnosis latency, total number of generated tests and total number of transmitted messages. Four algorithms are evaluated using SimDiag. Three of them use a static testing strategy whereas the other is based on an adaptive strategy. Presented results concern two aspects. First, an evaluation of the effect of increasing diagnosis functionalities is performed. Then, a practical comparison of used testing strategies is provided