Some properties of communication behaviors in discrete-event simulation models

This paper summarizes the profile data captured from 22 discrete-event simulation models captured from 4 different simulators. The profile data is captured by instrumenting the simulation engine and does not require any modification to the models. The profile data reported focuses on the communication properties of events exchanged between the various processes (LPs) in the model. The data suggests that some models share common behaviors and this work summaries their general characteristics. This permits a presentation of the principle characteristics using only six of the studied models. This resulting information can be used to: (i) provide configuration data for synthetic model generation, (ii) provide direction for configuring and optimizing parallel and distributed simulation engines, and (iii) provide insights into model correctness. The focus of this specific study is to determine regularities among LP event communications that can be exploited for model partitioning and event scheduling in parallel simulation.