Optimizing model interoperability in parallel discrete event simulation for cluster environment

ABSTRACT

Model interoperability enables independently developed models to interoperate with each other and has been a key technology in distributed simulations such as HLA-supported simulations. Parallel Discrete Event Simulation (PDES) also uses object models to represent real world entities thus it can also benefit from adopting a model interoperability mechanism. The work in this paper developed a model interoperability layer in the PDES environment by conforming to its event-object paradigm. The model interoperability layer uses a proxy-based approach to enable object models in a PDES to "publish" and "subscribe" their attributes similar to mechanisms in HLA. Especially, this design is optimized for executing in cluster-like environment. Preliminary performance tests show that the optimization improves Data Distribution Management performance up to 12 times in a cluster environment compared with unoptimized design.

REFERENCES

• 1: Richard M. Fujimoto, Parallel and Distribution Simulation Systems, John Wiley & Sons, Inc., New York, NY, 1999
• 2: Steinman J. S. "The Distributed Simulation Management Services layer in SPEEDES" {P}. 04S-SIW-098
• 3: Jeff S. Steinman, Incremental state saving in SPEEDES using C++, Proceedings of the 25th conference on Winter simulation, p.687-696, December 12-15, 1993, Los Angeles, California, United States [doi>10.1145/256563.256816]
• 4: Ivan Tacic , Richard M. Fujimoto, Efficient synchronized data distribution management in distributed simulations, Georgia Institute of Technology, Atlanta, GA, 2005

INDEX TERMS

Keywords:
DDM optimization, cluster computing, data distribution management, high performance simulation, model interoperability, parallel discrete event simulation