Trace Based Analysis of Process Interaction Models

Abstract

Simulation models of real world systems may have a complex dynamic behavior that has an impact on measures of interest but its causes are hard to detect by statistic measures commonly computed in a simulation run. Most simulators are able to document the details of a simulation run in a trace file. We propose a trace-based analysis and visualization method that allows us to identify reasons for large confidence intervals, high variances in lead times and correlations among delays. The technique is implemented in a stand alone tool for trace analysis that has been evaluated for the ProC/B modeling and simulation framework. We illustrate the visualization technique with the help of a ProC/B model of a warehouse.