Stochastic modeling of dynamic distributed systems with crash recovery and its application to atomic registers

In a dynamic distributed system, processes can join and leave the system. We consider such a system in which processes are subject to crash failures from which they may recover. Assuming a stochastic model for joining, leaving, crashing, and recovering of processes, we provide a probabilistic analysis of the long-term behavior of the system. As an example of the utility of our modeling, we provide a specification and implementation of an atomic register in such a system. The dynamic nature of the system can cause all active processes to leave or crash, leaving the system in a dormant state. We analyze the average time spent in dormant states that can give us some insight into the behavior of the register system. © 2012 Springer-Verlag.