No abstract

Uncoordinated checkpointing protocol is a simple protocol used in many distributed systems for fault tolerance. In this paper, we will discuss on the size of rollback it has in the presence of failures. In order to determining the recovery line in checkpoint-based recovery, we first study to common approaches: dependency graph and checkpoint graph and provide some algorithms for these approaches. Then we introduce a new approach for calculating the recovery line and making a graph (independent graph). Finally we present a solution for reducing the cost of graph when calculating the recovery line, particularly when the domino effect is occurred.