Dependability Evaluation of a Replication Service for Mobile Applications in Dynamic Ad-Hoc Networks

In order to increase availability and reliability of stateful applications, redundancy as provided by replication in cluster solutions is a wellknown and frequently utilized approach. For mobile services in dynamic ad-hoc networks, such replication mechanisms have to be adapted to deal with the frequently higher communication delays and with the intermittent connectivity. Dynamic clustering strategies in which the replica set is adjusted to the current network state can help to handle the network dynamicity. The paper develops a stochastic Petri net model (and its corresponding Markov chain representation) to analyze the resulting availability and replica consistency in such dynamic clusters. The numerical results are interpreted in the context of a vehicular (c2c) communication use-case and can be used to determine optimized cluster configuration parameters.