A Decoupled Federate Architecture for Hla-based Distributed Simulation

A number of research issues arise in executing large scale High Level Architecture (HLA) based distributed simulations. In order for simulation to provide an effective decision support tool, it is necessary to be able to evaluate alternative scenarios quickly and effectively. Distributed simulation cloning is designed to improve the performance of “what-if” analysis by enabling the examination of alternative scenarios concurrently within the same execution session. Fault tolerance is also required to support runtime robustness in HLA-based simulations. A generic approach to state saving and replication is necessary for these purposes. Nowadays, complex simulation applications may require huge computing resources and the data sets may also be geographically distributed. Such simulations may benefit from Web and Grid services in enhancing interoperability and execution flexibility. Furthermore, in order to effectively use sharable computing resources, it is desirable to introduce load balancing into large scale HLA-based distributed simulations. We propose a decoupled federate architecture in order to facilitate solutions to the above issues. A normal federate is decoupled into two processes, which execute the simulation model (virtual federate) and the Local RTI Component (physical federate) respectively. This approach interlinks the two processes together via external communication channels. The virtual federate interacts with the RTI through the standard RTI interface supported by a customized library. Benchmark experiments have been performed to study the extra overhead incurred by the decoupled federate architecture against a normal federate. The encouraging experimental results indicate that the