Detector de Falhas Impact em uma Abordagem Requisição-Resposta Tolerante à Instabilidade do Sistema

• Anubis Graciela de Moraes Rossetto
• Marcelo Felipe Guarani Fernandes
• Cláudio F. R. Geyer
• Luciana Arantes
• Pierre Sens

Resumo

The Impact failure detector (FD), introduced in [Rossetto et al. 2015a] outputs a trust level value which expresses the degree of confidence in the system. An impact factor is assigned to each process and the trust level is equal to the sum of the impact factors of the processes not suspected of failure. Moreover, an input threshold parameter defines a lower bound value for the trust level, over which the confidence in the system is ensured. However, in the definition of the impact factor, every node has an impact factor value which does not change during execution. In this article, we propose a message-pattern implementation of the Impact FD (query-response) in which the concept of impact factor is extended to become dynamic. Its value is periodically re-evaluated based on the history of false suspicions of the nodes. Performance results of experiments conducted with real PlanetLab traces confirm that the dynamic impact factor increases significantly the confidence in the system, when compared to the original static impact factor.