Consenso Distribuído Eficiente no Modelo Síncrono Particionado

Sérgio GorenderRaimundo Macêdo

The Partitioned Synchronous distributed system model has been introducedto take advantage of synchronous partitions of hybrid distributed systems, as suchsynchronous partitions are implementable in many real scenarios. In this paperwe present a consensus algorithm for such a system under the crash-recovery faultymodel. The proposed algorithm tolerates up to n - k process crashes, where k is thenumber of synchronous partitions and n the total number of processes, and terminatesin up to s-k+1 communication steps, s being the number of processes in synchronouspartitions. The consensus protocol utilizes a failure detector and a leader electionmechanism built for the Partitioned Synchronous model. The main advantage of theproposed solution, when compared with consensus for partially synchronous models,is the existence of an upper-bound for termination and its better resilience. Besidesformal proofs for the consensus, an implementation is briefly presented.

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: