Parallel Event–Driven Simulation Based on Application Global State Monitoring

Abstract

Discrete event simulation is a well known technique used for modeling and simulating complex parallel systems. Parallel simulation introduces multiple simulated event queues processed in parallel. A proper synchronization between parallel queues must be introduced. Program global state monitoring is a natural way to organize global simulation state monitoring and control. Every queue process reports its progress state, being the timestamp of the most recently processed event, to a global synchronizer. Reporting is done asynchronously and has no influence on the simulation process. A global simulation state can be defined as the vector containing timestamps of the most recently processed event in every queue. The paper presents the principles of parallel simulation designed by the use of a system infrastructure for global states monitoring. Comparison to existing parallel simulation methods is provided.