An efficient coordinated checkpointing scheme based on PWD model

Abstract

In this paper, we will prove that the new consistency condition for checkpoints must be needed in systems based on PWD model and propose an efficient coordinated checkpointing scheme. In our scheme, whenever process constructs a consistent global checkpoint set, it must obey the new consistency condition instead of previous consistency condition. That is, an execution of process is divided into an unique state interval occurred by non-deterministic event and checkpoint is taken in only processes happened a state interval transition. Consequently, proposed coordinated checkpointing scheme removes an unnecessary overhead emerged in previous works and guarantees the limited rollback propagation on the occurrence of a failure in systems based on PWD model.