Authors: Johannes Luethi and Clemens Berchtold

  Title: Concepts for Dependable Distributed Discrete Event Simulation

     In: Proceedings of the International European Simulation Multi-Conference
	 Ghent University, Ghent, Belgium, May 23-26, 2000


Abstract:

In many situations, parallel and distributed simulation is a well-suited 
approach to overcome performance as well as capacity limitations of complex 
simulation models.  However, if distributed simulation runs take hours or 
days for termination or if distributed simulation is applied to safety 
critical domains such as e.g. military decision making, it is not acceptable 
that a failure in a single process, processing element, or communication
link stops the whole simulation experiment. Fault tolerance mechanisms which 
are popular in general purpose distributed computing are not common in the 
special case of distributed discrete event simulation.  In this work, we 
structure the necessary dependability requirements and point out approaches 
how they can be met. Two directions are discussed: the specialization of 
fault tolerance mechanisms from general purpose distributed computing to 
distributed simulation and the adaptation of special mechanisms of 
distributed simulation to support fault tolerance.  Special attention is 
also paid to approaches increasing the dependability of implementations of 
the High Level Architecture (HLA) and to possibilities to extend the HLA 
to support fault tolerance mechanisms for HLA federates and federations.