Robust Assertions and Fail-Bounded Behavior

Paula PrataMario RelaHenrique MadeiraJoão Gabriel Silva

In this paper the behavior of assertion-based errordetection mechanisms is characterized under faults injectedaccording to a quite general fault model. Assertionsbased on the knowledge of the application can bevery effective at detecting corruption of critical data causedby hardware faults. The main drawbacks of that approachare identified as being the lack of protection of data outsidethe section covered by assertions, namely during inputand output, and the possible incorrect execution ofthe assertions.To handle those weak-points the Robust Assertionstechnique is proposed, whose effectiveness is shown byextensive fault injection experiments. With this techniquea system follows a new failure model, that is called Fail-Bounded, where with high probability all results producedare either correct or, if wrong, they are within a certainbound of the correct value, whose exact distance dependson the output assertions used.Any kind of assertions can be considered, fromsimple likelihood tests to high coverage assertions suchas those used in the Algorithm Based Fault Tolerance paradigm.We claim that this failure model is very useful todescribe the behavior of many low-cost fault-tolerant systems,that have low hardware and software redundancy,like embedded systems, were cost is a severe restriction,yet full availability is expected.

Caso o link acima esteja inválido, faça uma busca pelo texto completo na Web: Buscar na Web

Biblioteca Digital Brasileira de Computação - Contato:
     Mantida por: