Daemon conversions in distributed self-stabilizing algorithms

Abstract

We consider protocols to transform a self-stabilizing algorithm for one daemon to one that can run under a different daemon. In the literature, there are several daemons, and several possible attributes of those daemons, and it is customary to detail the choice of daemon one is using in designing a specific self-stabilizing algorithm. The choice of daemon plays an important role in designing self-stabilizing algorithm in terms of correctness and convergence time analysis; techniques and complexity vary widely with the type of daemons used. In order to simplify algorithm development in a systematic way, it would be useful to have to consider only one "canonical" daemon and then to use a relatively mechanical procedure to convert the algorithm to any other daemon when needed. We give the first (full) proof that, provided there are IDs, any algorithm that self-stabilizes only under a fair central daemon can be converted to one that self-stabilizes under an unfair read/write daemon. © 2013 Springer-Verlag.