Checkpoint/rollback vs causally-consistent reversibility
Abstract
This paper analyzes the relationship between a distributed checkpoint/rollback scheme based on causal logging, called Manetho, and a reversible concurrent model of computation, based on the π-calculus with imperative rollback developed by Lanese et al. in [?]. We show a rather tight relationship between rollback based on causal logging as performed in Manetho and the rollback algorithm underlying the calculus in [?]. Our main result is that the latter can faithfully simulate Manetho, where the notion of simulation we use is that of weak barbed simulation, and that the converse only holds if possible rollbacks in are restricted.
Origin | Files produced by the author(s) |
---|