Sampling Simulation Model Profile Data for Analysis

The capture of data about the events executed by a discrete event simulation can easily lead to very large trace data files. While disk space is relatively inexpensive and mostly capable of storing these large trace files, the manipulation and analysis of these large trace files can prove difficult. Furthermore, some types of analysis must be performed in-core and they cannot be performed with the trace data exceeds the size of the physical RAM where the analysis is performed. Because of these limits, it is often necessary to strictly limit the simulation run time to satisfy the analysis time memory limits. Experience with the DESMetrics tool suite (a collection of tools to analyze event trace files), demonstrates that our in-memory analysis tools are limited to trace files on the order of 10GB (on a machine with 24GB of RAM). Furthermore, even when it is possible to analyze large trace files, the run time costs of performing this analysis can take several days to complete. While high performance analysis of traces data is not strictly necessary, the results should be available within some reasonably bounded time frame. This paper explores techniques to overcome the limits of analyzing very large event trace files. While explorations for out-out-core analysis have been examined as part of this work, the run time costs for out-of-core processing can increase processing time 10-fold. As a result, the work reported here will focus on an approach to capture and analyze small samples from the event trace file. The work reported in this paper will examine how closely the analysis from sampling matches the analysis from a full trace file. Two techniques for comparison are presented. First a visual comparison of analysis results between the full trace and a trace sample are presented. Second, numerical quantification of the different analysis results (between the full trace and trace sample) will be reported using the Wasserstein, Directed Hausdorff, and Kolmogorov-Smirnov distance metrics. Finally, the ability to process trace samples from a very large trace file of 80GB is demonstrated.