Brief Announcement: Sharing Memory in a Self-stabilizing Manner

摘要

A core abstraction for many distributed algorithms simulates shared memory [3]; this abstraction allows to take algorithms designed for shared memory, and port them to asynchronous message-passing systems, even in the presence of failures. There has been significant work on creating such simulations, under various types of permanent failures, as well as on exploiting this abstraction in order to derive algorithms for message-passing systems (see [2].) All these works, however, only consider permanent failures, neglecting to incorporate mechanisms for handling transient failures. Such failures may result from incorrect initialization of the system, or from temporary violations of the assumptions made by the system designer, for example the assumption that a corrupted message is always identified by an error detection code.