Caching Strategies for Run-time Probabilistic Model Checking

For software systems that need to adapt to their environment at run-time, run-time verification is useful to guarantee the correctness of their behaviors. Probabilistic model checking using discrete time Markov chain (DTMC) model has been applied to implement run-time verification. A current existing approach provides an efficient run-time verification mechanism by pre-generating expressions for model checking at design time. In case that a system model is changed, the system is required to re-generate the expressions. In order to expand the applicability of the approach, we propose three strategies, caching, prediction, and reduction, for reducing computational time for re-generated expressions at run-time. We conduct preliminary experiments and demonstrate that our approach could expand the applicability of run-time verification by reducing the computational cost at runtime.