Understanding partitions and the `no partition' assumption

摘要

Discusses partitions in asynchronous message-passing systems. Innsuch systems, slow processes and slow links can lead to virtualnpartitions that are indistinguishable from real ones. To overcome thenimpossibility of detecting crashed processes in an asynchronous system,nthe system model incorporates a failure suspector to detect (possiblynerroneously) process failures. Based on failure suspicions, the authorsngive a definition of partitions that accounts for real partitions asnwell as virtual ones. It is shown that under certain assumptions aboutnthe process behavior, any incorrect failure suspicion inevitablynpartitions the system. It is then shown how to interpret thenabsence-of-partition assumption