SAN-based modeling of fault tolerant protocols for MANETs

With the advent of new technologies in the area of wireless systems and communications, there is an increasing need to develop distributed applications that take into account mobility constraints while being fault tolerant. Indeed, this environment is very sensitive to failures. Among the existing techniques, we focus on the rollback recovery that is best suited for this kind of situations. The rollback recovery requires to save regularly (in the background of the distributed application), local states of nodes; this task is called checkpointing. Thus, the system will resume (after repairing or replacing the failing node) from one of these states (checkpoints). To launch the rollback recovery, it is necessary to define a protocol that has the ability to detect all the failures while reducing the rate of false suspicions. Therefore, the definition, the verification and the performance evaluation of failure detection protocols is a crucial stage to achieve. In this paper, we show how to model the behavior of already defined fault tolerant protocols: FDAN (failure detection protocol), 2P ACA (checkpointing protocol) by using stochastic Automata Networks (SAN). This makes it possible to verify their coherence and evaluate their performances by using dedicated tools.