An Adaptive Framework for Synchronization of Distributed Simulations

Open Access
Author:
Altuntas, Bertan
Graduate Program:
Industrial Engineering
Degree:
Doctor of Philosophy
Document Type:
Dissertation
Date of Defense:
November 22, 2004
Committee Members:
  • Richard Allen Wysk, Committee Chair
  • Catherine Mary Harmonosky, Committee Member
  • Vittaldas V Prabhu, Committee Member
  • David J Hall, Committee Member
Keywords:
  • simulation
  • discrete event
  • synchronization
  • distributed simulation
Abstract:
Increased complexity of simulation models has created the need for distributing the execution of simulations on multiple processors. While distributed simulation promises rapid sub-model development and faster execution speeds, it poses two challenging problems: (a) synchronization of distributed simulations and (b) rapid development and deployment of integrated distributed simulation systems. In this doctoral research, a new adaptive synchronization framework was developed to address both of the aforementioned problems. Formal models of distributed discrete-event simulation were created based on the finite-state machine representations of discrete-event systems. These models were used for: (a) analyzing the requirements of distributed simulation synchronization and (b) creating an automated synthesis mechanism for rapid development and deployment of distributed simulation systems. The developed synchronization framework features a novel synchronization protocol, which is based on adaptive pacing of the execution of distributed simulations during interactions. This new synchronization protocol, called Adaptive Partial Pacing (APP) synchronization, relies on synchronized real-time clocks at distributed simulations for coordinated pacing and also features an adaptive pace determination scheme for greater synchronization efficiency. The developed synchronization framework also features an automated synthesis mechanism which allows rapid creation of a synchronization system from simulation models. This capability allows users with limited specific knowledge of distributed simulation synchronization to create and run distribute simulation systems in a short amount of time. The synchronization framework was successfully implemented using the ArenaTM simulation package from Rockwell Automation. The research shows that for loosely coupled simulations, a significant speed-up of the system can be realized. It also shows that for systems with little variability more significant improvements can be achieved.