Nothing Special   »   [go: up one dir, main page]

skip to main content
Parallel simulation techniques for large-scale discrete-event models
Publisher:
  • Carleton University
  • 1125 Colonel-By Drive Ottawa, Ont. K1S 5B6
  • Canada
ISBN:978-0-494-87762-3
Order Number:AAINR87762
Pages:
128
Reflects downloads up to 16 Dec 2024Bibliometrics
Skip Abstract Section
Abstract

The Discrete Event System Specification (DEVS) provides a general methodology for hierarchical construction of reusable models in a modular way and has been used to simulate complex systems in a variety of domains. This dissertation addresses software design and performance issues that arise in parallel simulation of large-scale DEVS-based models on multiprocessor cluster architecture.

Parallel simulation of complex DEVS-based models requires a robust simulator with low synchronization overhead. Recent researches focused on optimistic parallel simulation of DEVS-based systems. In this research three conservative parallel DEVS protocols (Lower-Bound-Time-Stamp (LBTS), Chandy-Misra-Bryant (CMB), and Global-Lookahead-Management (GLM)) are proposed, allowing pure conservative simulation of DEVS-based systems. The protocols are based on the classical Chandy-Misra-Bryant synchronization mechanism, and they extend the DEVS abstract simulator, providing means for lookahead computation and null message distribution. A purely conservative simulator, called CCD++, is presented designed for running large-scale DEVS and Cell-DEVS models in parallel and distributed fashion.

An extensive comparative performance analysis is presented, analyzing the performance of CCD++ compared to an optimistic DEVS simulator. Several DEVS-based environmental models with different characteristics are studied. The experiments indicate that the conservative simulator improves performance in terms of execution time, memory usage, operational cost, and system stability for large models.

Contributors
  • Embry-Riddle Aeronautical University
Please enable JavaScript to view thecomments powered by Disqus.

Recommendations