Theoretical Model of Optimistic Recovery in Distributed Systems.

摘要

The research sponsored by this contract has produced the following results: 1) A model for recovery using message logging and checkpointing, resulting in a major theorem establishing the uniqueness of a maximum recoverable system state; 2) An implementation and a performance evaluation of sender-based message logging; 3) A batch and an incremental algorithm for optimistic recovery, and an implementation and a performance evaluation of optimistic recovery; 4) Preliminary results in extending these methods to nondeterministic processes; and 5) A new definition of distributed breakpoint, called causal distributed breakpoint. (KR)