Rewriting 'the hare and the turtle': sleeping to get there faster

Abstract

When developing algorithms for dependable distributed systems one often makes several simplifying assumptions that are essential to reason about the problem in hand. It is usual to assume an asynchronous system model, unconstrained system resources and the absence of easily maskable faults such as message loss. While most of the simplifications strengthen the model and are particularly useful when proving theoretical edge results, asynchrony, on the contrary, is a "non-assumption" and it is specially appealing in practice as it yields robust solutions that are correct regardless of the actual timing behavior of the target systems.