Model-Driven Engineering for Reversible Parallel Discrete Event Simulation

Simone Bauco and Alessandro Pellegrini

Abstract:
This paper introduces a Model-Driven Engineering approach to automatically generate reverse event handlers for optimistic Parallel Discrete Event Simulation. Optimistic synchronisation, such as Time Warp, relies on the ability to roll back the simulation state to a previous consistent point. Traditionally, this is achieved through manual implementation of reverse handlers or full state checkpointing, both of which are error-prone or computationally expensive.
The proposed methodology uses model-to-model transformations to derive a reverse model from a high-level forward event model defined in the DESL language. This transformation transparently handles data and structural irreversibility by automatically inverting reversible operations and injecting checkpointing logic for destructive operations. We also address the limitations of existing reverse computing solutions regarding dynamic memory by introducing a reference implementation of a reversible memory allocator specifically designed for PDES. The approach is validated using the ROSS runtime environment, demonstrating that higher-level abstractions can simplify model development while maintaining execution consistency and performance.

BibTeX Entry: