Validação distribuída do algoritmo paxos no modelo arbitrário não malicioso

Abstract

Distributed algorithms have been increasingly used because of commercial and technical factors, especially when the aim is to obtain scalability or high availability of data. Regarding high scalability, stands out the algorithms in which a set of processes interact through the exchange of messages in order to perform the same actions and remain in the same state, called replication algorithms. However, to meet this high demand, it is not enough for such algorithms to be functional, but they must also be fault tolerant, in order to avoid data corruption. Fault tolerance in distributed algorithms is not a trivial topic, and algorithms that tolerate arbitrary failure tend to be costly and complex to implement. In order to tolerate several fault types while keeping the complexity low, the model of arbitrary non-malicious faults was created, where only faults caused by intentional invasions to the environment are not tolerated. This dissertation proposes a distributed validation mechanism for the Paxos algorithm in order to guarantee the properties of the model, making it tolerant to non-malicious arbitrary faults.