Guided Search for Deadlocks in Actor-Based Models

Abstract

The success of model checking is based on its ability to uncover errors in designs of software and protocols. Even a small reactive concurrent system can exhibit complex behavior. Such systems may have state-spaces larger than explicit state model checkers can verify. In practice, finding an error with a model checker is more useful than proving a property. Informed search algorithms use heuristic strategies with problem-specific knowledge to find solutions more efficiently than uninformed algorithms. Generally, such heuristics estimate the distance from a given state to a goal state. We present seven heuristics for guiding search algorithms through the state-space of actor-based models to a deadlock. Our methods can find a deadlock more efficiently than uninformed searches for some actor-based models. The A* search algorithm guarantees an optimal solution and returns the shortest counterexample. These methods are supported by a tool that performs directed model checking of the deadlock property. The objective is to detect difficult errors that might not be found by simulation or by conventional model checkers before reaching an upper bound or state-space explosion.