Staggered Checkpointing and Recovery in Cluster Based Mobile Ad Hoc Networks

Checkpointing uses stable storage available in the distributed system for saving the consistent states of processes to which they can rollback at the time of recovery. But the checkpointing techniques for wired and cellular mobile systems are not trivially applicable to ad hoc networks as these networks have limited stable storage and wireless links are of low bandwidth. Moreover if synchronous checkpointing is employed, the processes contend for these limited resources at the time of checkpointing. This paper addresses the application of checkpointing to ad hoc networks and proposes a staggered approach to avoid simultaneous contention for resources. The staggering causes events, which would normally happen at the same time, to start or happen at different times. The proposed protocol does not need FIFO channels and logs minimum number of messages. It supports concurrent checkpoint initiation and successfully handles the overlapping failures in ad hoc networks.