Consensus is Easier Than Reliable Broadcast

Carole Delporte-Gallet (1) , Stéphane Devismes (2) , Hugues Fauconnier (1) , Franck Petit (3) , Sam Toueg (4)

1 LIAFA - Laboratoire d'informatique Algorithmique : Fondements et Applications
2 VERIMAG - IMAG - VERIMAG
3 GRAAL - Algorithms and Scheduling for Distributed Heterogeneous Platforms
4 DCS - Department of Computer Science [University of Toronto]

We consider asynchronous distributed systems with message losses and process crashes. We study the impact of finite process memory on the solution to consensus, repeated consensus and reliable broadcast. With finite process memory, we show that in some sense consensus is easier to solve than reliable broadcast, and that reliable broadcast is as difficult to solve as repeated consensus: More precisely, with finite memory, consensus can be solved with failure detector S , and P − (a variant of the perfect failure detector which is stronger than S ) is necessary and sufficient to solve reliable broadcast and repeated consensus.

Mots clés: Distributed algorithms; failure detectors; reliable broadcast; consensus; repeated consensus.; repeated consensus

Domaines: Informatique

Carole Delporte-Gallet, Stéphane Devismes, Hugues Fauconnier, Franck Petit, Sam Toueg. Consensus is Easier Than Reliable Broadcast. 2008. ⟨hal-00325470⟩