Distributed checkpoint algorithms to avoid roll-back propagation

Checkpointing is a very well known mechanism to achieve fault tolerance. In distributed applications, a local checkpoint is useful for fault tolerance purposes only if can belong to at least one consistent global checkpoint and then, execution can be restarted from it without needing to roll back the execution in the past. The paper introduces a theoretical framework that facilitates the definition and the analysis of distributed checkpoint algorithms to avoid roll backpropagation. On this base, several algorithms are presented and evaluated in a set of testbed applications.