Modeling of fault-tolerant techniques in hierarchical systems

The authors consider both centralized and distributed fault-tolerant schemes. Based on stochastic Petri net models, they investigated the performance of these two approaches by considering the levels in the hierarchical system independently. In the case of decentralized fault tolerance, they considered two different checkpointing strategies. In the first scheme, called the arbitrary checkpointing strategy, each process does its checkpointing independently; as a result, there is the possibility of domino effect. In the planned strategy, checkpointing is done in a manner which ensures that there is no domino effect. The results show that for certain cases, the arbitrary checkpointing strategy can perform better than the planned strategy. The authors also studied the effect of integration on the fault-tolerant strategies of the various levels.<>