Web Services-Based Parallel Replicated Discrete Event Simulation for Large-Scale Simulation Optimization

Abstract

The exciting developments in the World Wide Web (WWW) have revived interest in computer simulation for modeling, particularly for conceiving simulation languages and building model libraries that can be assembled and executed over the Internet, and for analysis, particularly for developing simulation optimization algorithms for parallel experimentation. This paper contributes to this second stream of research by introducing a framework for Optimization via Simulation (OvS) through Parallel Replicated Discrete Event Simulation (PRDES). In particular, we combine Nested Partitions (NP) and Extended Optimal Computing Budget Allocation (EOCBA) to provide an efficient framework for PRDES experiments. The number of candidate alternatives to be evaluated can be reduced by the application of NP. EOCBA, a modification of the Optimal Computing Budget Allocation (OCBA) for PRDES, minimizes the number of simulation replications required to evaluate a particular alternative by allocating computing resources to potentially critical alternatives. We deploy web services technologies based on the Java Axis and .NET framework to create a viable infrastructure for heterogeneous PRDES systems. This approach, which receives increasing attention under the banners of `grid computing' and `cloud computing,' further promotes reusability, scalability, and interoperability. Experimental results with a prototype implementation not only furnish a proof of concept but also illustrate significant gains in simulation efficiency with PRDES. The proposed concept and techniques can also be applied to simulation models that require coordination and interoperation in heterogeneous environments, such as decentralized supply chains.