Abstract

Simulation-based fault injection is an important experimental means for evaluating system dependability and for testing the effectiveness of embedded fault tolerance mechanisms during the design phase of a system. A crucial problem in simulation-based fault injection is the number of fault injections to perform. In particular, when testing embedded fault tolerance mechanisms it is of main interest that an injected fault really forces an erroneous system behavior. Operational-profile-based fault injection is a method in which faults/errors are injected in the detailed knowledge of the system dynamics.

This enables a significantly decrease of the number of fault simulation runs since injection is reduced to the important fault scenarios. Discrete-event simulation and event-driven monitoring, joined together, provide detailed data which underline the basis to determine the operational system profile. After an initial fault-free simulation run this data is used to fix interesting fault scenarios. During fault injection, the data allows detailed observation of the effects of an injected fault/error. The proposed fault-injection technique requires the introduction of massively tool-support. This paper presents the techniques and concepts used by

The real man behind the paper

Sometimes, when he is kissed by the Muse, he turns on his old sampler composing some pieces or just playing some good old songs of Depeche Mode. When there is time enough, he performs some really amazing Techno gigs with his friends.