Statistical Model Checking of Distributed Programs within SimGrid

Marie Duflot-Kremer, Yann Duplouy


In this paper, we present an approach to perform statistical model-checking over stochastic distributed programs using the SimGrid framework. The distributed programs are modeled using SimGrid, a fast and lightweight framework for the simulation of distributed programs, which we have enhanced in three ways: a cleaner description of the probabilistic evolution of the capacities of resources, a centralized random number generator, and a protocol for the observation of the simulations. We also propose a toolset for the statistical model-checking of those simulated distributed programs, and in particular a prototype tool SimGridStatMC. The toolset is illustrated to evaluate various properties of an implementation of the peer to peer BitTorrent protocol.


Paper Citation