Model based evaluation of policies for end-node driven fault recovery

In evolving ubiquitous networking environments some end-nodes will rely on highly critical end-user services in infrastructure and ad-hoc networks. As future networks will be dominated by mobility and unreliable wireless links, providing such critical end-user services is a challenge. In this work we target the problem of managing faults in such networking systems by fault management driven by the end-node. We introduce a system model to evaluate recovery decision policies based on fault diagnosis capabilities. The model method is applied in a scenario of hand-over between wireless access points. Evaluated policies are (1) no fail-over, (2) fail-over at fault diagnosis and (3) time restricted fail-over. Comparison to detailed simulations show that the developed Markov model is suitable to maximize end-user service reliability - in this case for the example of an SCTP file transfer. The numerical results from the model allow to determine the best strategy and its parameters depending on properties of the fault-diagnosis component.