| Real-time Distributed Simulation is analyzed as a class of loosely-coupled, distributed computation. A formal model of the system is developed which provides a context for inductive and deductive analysis techniques. A synchronization approach is developed to induce repeatable execution without changing the nature of the computation. A data collection approach is developed to accommodate the fundamental temporal properties of the computation and support deductive analysis of the simulation message logs.; In the synchronization research, we examine real-time models of execution, and show ways to induce repeatability in real-time simulations. We explore a technique to address certain overhead issues in making real-time distributed simulation processes meet hard real-time constraints.; The performance demands of real-time simulation are also used as motivation for research in high-performance temporal databases. We address efficient storage techniques for real-time temporal data, and develop a prototype implementation of a high-performance log-based temporal database. We explore basic concepts in mapping the simulation data collection problem to a temporal database model, and show how to exploit the specialized bi-temporal relationship that exists between simulation time and wall-clock time. Finally, we confirm the practicality of the log-based approach by presenting performance data for various strategies to manage memory and disk space requirements. |
