Consistent modeling of distributed mutual exclusion protocol using optimistic simulation

This paper discusses the problem of risk in optimistic simulation protocols, using as an example, simulation of a distributed mutual exclusion protocol with strong consistency properties. The simulation model is augmented to detect model inconsistency errors resulting from risky optimistic simulation. While the model runs sequentially without consistency errors, errors occur when the model is executed in parallel optimistically. Some of the errors entirely violate the fundamental mutual exclusion properties of the model itself. To address this problem, we extend the optimistic simulation library to eliminate these inconsistencies. We discuss the details of these extensions and the performance tradeoff for adding them.