ABSTRACT
This chapter surveys the state of the art in executing discrete event simulation programs on a parallel computer. It discusses fundamental concepts used in discrete event simulation and examine why this application is so difficult toparallelize. The chapter also discusses several simulation strategies that have been proposed, as well as the ideas on which they are based. It focuses attention on discrete event simulations of asynchronous systems, that is, systems that are not synchronized by a global clock. Parallelization of discrete event simulation problems contrasts sharply with parallelization of continuous simulation models. The communication network simulation also maps very naturally to logical processes that communicate by exchanging time-stamped event messages. Many conservative schemes require knowledge concerning logical process behavior to achieve good performance. Conservative mechanisms strictly avoid any possibility of violating the local causality constraint, but optimistic approaches allow violations to occur but provide a mechanism for detecting and recovering from them.
