Byzantine fault tolerance was introduced to implement systems capable of tolerating arbitrary component failures, where n - f replicated state machines maintain their state consistent despite the action of up to f arbitrarily or Byzantine faulty state machines, for n ≥ 3f + 1. This notion was extended later for tolerating malicious attacks or intrusions when some of the systems components get compromised by a malicious intruder or attacker. Due to its high costs in terms of minimum needed redundancy (minimum 3f + 1 replicas), and related latency, several authors have turned their attention to alternative approaches where faulty processes can be excluded from the system adapting the current quorum of operational processes. In this paper, we explore the Partitioned Synchronous distributed system model, which suits existing real systems, such as large computational grids and distributed industrial plants, to propose a new Byzantine failure detector and related consensus algorithm for such a model. From our simulations we show that our approach indeed improves resilience when compared to static quorum approaches, with the continuous detection of faulty processes.
展开▼