Communication strategies for heartbeatstyle failure detectors in wireless ad hoc networks

Heartbeat-style failure detectors are a commonly used building block in practical fault-tolerant distributed systems over unreliable and asynchronous networks. A basic requirement to implement such failure detectors is to diffuse heartbeat information across the underlying network. In wireless ad hoc networks, however, the dynamics of mobility and lack of resource make information dissemination a formidable task. Moreover, as a middleware service, the total bandwidth used for failure detection should be constrained. This paper describes several communication strategies on which heartbeat-style failure detectors can be developed in wireless ad hoc networks. The design goal is to support an effective and robust means of gossiping under a fixed message transmission rate. We show through simulations that the proposed gossiping schemes are resilient to message losses and topology changes. The simulation results also show that the performance of gossiping can be improved by introducing the concept of transient hierarchy under varied network characteristics. 1