Distributed implementation of a stable storage for MANET checkpointing protocols
Haroun Benkaouha, Abdelkrim Abdelli, Jalel Ben‐Othman, Youcef Zaffoune, Lynda Mokdad · 2016
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.