Failure detectors, which provide (maybe erroneous) information about process failures, have been shown to be a very useful mechanism to solve, among others, the consensus problem. Communication efficiency, denned in [1] as having a linear number of links carrying messages forever, has been a subject of research when designing failure detectors for the crash model [7], but, as far as we know, not yet for the general omission model. This model covers situations in which message omissions can not be fully controlled, e.g., buffer overflows, restricted capacity of re-sending or malicious behaviour like such of [2]. Failure detection in omission environments was first addressed in [5], and more recently in [2, 4], All these failure detectors have a permanent all-to-all communication pattern, and thus are not communication-efficient. In this brief announcement, we first define a new failure detector for the general omission model in terms of completeness and accuracy properties, and then propose a communication-efficient algorithm that implements it.
展开▼
