Distributed Implementation of a Stable Storage for MANET Checkpointing Protocols

To guarantee fault tolerant system, checkpointing protocols are usually run to save the last consistent state of the system. This information will be recovered whenever a node failure occurs, thereby making it possible to resume the execution of the system. In the context of MANETs, checkpoints need to be saved in a safe and a stable storage to avoid data loss. Due the the specific constraints of such networks, the implementation of a stable storage (SS) is a hard task to achieve. So far, the most used techniques are based on mirroring or replication. However, although the latter achieve a safe an a reliable implementation of the SS, the deployment and the maintenance of the SS remain very costly. We explore in this paper a solution based on a distributed implementation of the SS by using RAID5 storage scheme. Mobile agent paradigm and geographic areas repartition of nodes are exploited to provide a virtual centralisation of the system. We associate this solution with a recent checkpointing protocol to provide an efficient fault tolerance tool for MANETs. Simulation results performed on NS2 show that our solution outperforms clearly replication.