A Fault Tolerance protocol for ASP calculus: Design and Proof
Résumé
This research report first details a communication induced checkpointing fault tolerance protocol adapted to ProActive, a Java library that implements the ASP model. This model is based n a request/reply mechanism. In order to prove the correctness of this protocol, we introduce a local partial order between events occurring on a given process. This order is extended into a global order by the Lamport's happened-before relation. Finally, we prove that from a cut that is ''consistent enough'', a second execution is constrained to go equivalently to the first one until a consistent global state of the first one (the history closure). Thus, the protocol described in this report ensures that, even from a inconsistent recovery line, a reexecution cannot lead to an inconsistent state that could not exist in a normal execution.
Domaines
Autre [cs.OH]
Loading...