Robust Stabilizing Leader Election

Carole Delporte-Gallet (1) , Stéphane Devismes (2) , Hugues Fauconnier (1)

1 LIAFA - Laboratoire d'informatique Algorithmique : Fondements et Applications
2 LaRIA - Laboratoire de Recherche en Informatique d'Amiens

In this paper, we mix two well-known approaches of the fault-tolerance: robustness and stabilization. Robustness is the aptitude of an algorithm to withstand permanent failures such as process crashes. The stabilization is a general technique to design algorithms tolerating transient failures. Using these two approaches, we propose algorithms that tolerate both transient and crash failures. We study two notions of stabilization: the self- and the pseudo- stabilization (pseudo-stabilization is weaker than self-stabilization). We focus on the leader election problem. The goal here is to show the implementability of the robust self- and/or pseudo- stabilizing leader election in various systems with weak reliability and synchrony assumptions. We try to propose, when it is possible, communication-efficient implementations. In this work, we exhibit some assumptions required to obtain robust stabilizing leader election algorithms. Our results show, in particular, that the gap between robustness and stabilizing robustness is not really significant when we consider fix-point problems such as leader election.

Mots clés: Distributed systems; self-stabilization; pseudo-stabilization; robust algorithm; leader election

Domaines: Informatique

Carole Delporte-Gallet, Stéphane Devismes, Hugues Fauconnier. Robust Stabilizing Leader Election. 2007. ⟨hal-00167935⟩