PDEVS Protocol Performance Prediction using Activity Patterns with Finite Probabilistic DEVS

Zeigler Bernard P. (1) , Laurent Capocchi (2) , Jean-François Santucci (2)

1 ECE - Department of Electrical and Computer Engineering [Tucson]
2 SPE - Sciences pour l'environnement

This paper deals with PDEVS (Parallel Discrete Event System specification) protocol performance prediction using activity patterns with Finite Probabilistic DEVS. In 2015, Zeigler et. al showed that the PDEVS protocol is a simple-to-implement distributed simulation protocol that can provide speedups to match the best performance of classical algorithms for most situations. We propose to model the PDEVS protocol using a Markov Continuous Time Model (CTM) which is set up using two parameters: (i) the probability of a state giving an output according to input patterns; (ii) the rate of the coordinator's release of imminents. These parameters are computed using Activity metrics and inserted into the PDEVS modeling scheme in order to predict the performance of a PDEVS protocol in the framework of distributed simulations. The approach is supported by the DEVSimPy and Ms4Me environments and is highlighted by a pedagogical example.

Mots clés: Finite Probabilistic DEVS; Markov chain; Ms4Me; parallel simulation; PDEVS Protocol

Domaines: Modélisation et simulation; Calcul parallèle, distribué et partagé [cs.DC]; Probabilités [math.PR]; Statistiques [stat]

Zeigler Bernard P., Laurent Capocchi, Jean-François Santucci. PDEVS Protocol Performance Prediction using Activity Patterns with Finite Probabilistic DEVS. Spring Simulation Multi-Conference (SpringSim'16), May 2016, Pasadena, United States. ⟨hal-01314277⟩