Accelerated time discrete event simulation in a distributed environment
作者:
DAVIDA. COOK,
UDOW. POOCH,
期刊:
International Journal of Systems Science
(Taylor Available online 1993)
卷期:
Volume 24,
issue 3
页码: 451-478
ISSN:0020-7721
年代: 1993
DOI:10.1080/00207729308949500
出版商: Taylor & Francis Group
数据来源: Taylor
摘要:
In distributed discrete event simulation, accelerating the simulation time by accelerating or warping time is a means of speeding up the execution time of the simulation. One method of accelerating time is the time warp or virtual time methodology, where processors advance their local clocks at will, communicating with other processors using messages. A processor must rollback to recover when inconsistent or unsafe states occur, using antimessages. Problems associated with accelerating time in a distributed simulation include large buffer spaces for saving state information, and the potential for ripple effects, which occur when one rollback triggers others. In addition, simulations that must produce irrevocable actions cannot allow a rollback to occur that nullifies the action. This research presents a methodology for synchronizing event-driven simulations in a distributed environment where processors can be arranged into a logical token ring. This methodology, known as event jumping, prevents non-local rollbacks and reduces the amount of buffer space used for state saves. This paper also shows the results of implementing this algorithm using both a network simulator and an actual distributed environment.
点击下载:
PDF (907KB)
返 回