What model and what conditions to implement unreliable failure detectors in dynamic networks?

Abstract

Failure detectors are classical mechanisms which provide information about process failures and can help systems to cope with the high dynamics of self-organizing, unstructured and mobile wireless networks. Unreliable failure detectors of class ◊S are of special interest because they meet the weakest assumptions able to solve fundamental problems on the design of dependable systems. Unfortunately, a negative result states that no failure detector of that class can be implemented in a network of an unknown membership; but full membership knowledge as well as fully communication connectivity are no longer appropriate assumptions to the new scenario of dynamic networks. In this paper, we provide a discussion about the conditions and model able to implement failure detectors in dynamic networks and define a new class, namely ◊SM, which adapts the properties of the ◊S class to a dynamic network with an unknown membership.