Implementing parallel and distributed DEVS and cell-DEVS simulation in a windows platform.

摘要

Research advances in modeling and simulation emphasize the need for parallel and distributed methodologies and environments. The growing demand for executing complex models by general users has directed researchers to implement parallel simulation with commodity PC machines. This research presents an effective approach to executing a parallel and distributed Discrete Event System Specification (DEVS) and Cell-DEVS application in Windows cluster environments.;DEVS is a modular and hierarchical formalism for modeling and analyzing general systems that can be described by discrete events. Cell-DEVS is a DEVS-based formalism used to model complex physical systems as cell spaces. Parallel DEVS provides a way to handle simultaneously scheduled events, while keeping all the major properties of the original DEVS formalism. A logical process (LP), as a basic entity in parallel DEVS environments, receives and generates timestamped events to communicate with other LPs. The communication mechanism of the LP can be implemented with a distributed paradigm.;In this research, PCD++Win and PCD++/.NET simulation systems are presented. Both of them follow a conservative approach and use a set of commodity Windows PC machines to execute parallel DEVS and Cell-DEVS simulations. PCD++Win is based on Windows MPI and allows users to setup a Windows cluster and execute parallel DEVS and Cell-DEVS simulations with a GUI. PCD++Win can be exposed as a Web service, while another application can consume PCD++Win over the Internet. PCD++/.NET is based on Microsoft.NET. It presents an approach to combining .NET Remoting objects with a simulation engine to execute parallel and distributed DEVS and Cell-DEVS simulations. PCD++/.NET supports several communication protocols and runs on the Common Language Runtime. The performance analysis shows that the speedup of PCD++/.NET can be achieved for the simulation, which has a modest inter-LP communication load.