Independent node and process recovery in message passing distributed systems

Consistent recovery from process failures is an essential component of reliable distributed systems. Many existing recovery techniques use asynchronous message logging and checkpoints. Most of the present approaches depend on logged states of non-fail processes for recovery. A model of recovery based on the current active states of processes has been proposed. The algorithm considers recoverable state of the failed process and current states of the non-failed processes. Each process recovers to a consistent system state independently