We present the results of coverage tests performed to validate our preliminary analysis indicating that determining 'appropriate' run length is more important for obtaining coverage than performing 'proper' transient analysis. Our preliminary experiment was designed with the intention of showing students the pitfalls of performing 'bad' transient analysis when estimating steady-state parameters. However, we found that for short run lengths any transient truncation diminishes coverage; and it is only beneficial to delete transient data when long runs of the output data are available. As with the preliminary analysis, two types of systems are analyzed (M/M/1/GD/∞/∞ systems and an M/M/s/GD/∞/∞); additionally, coverage tests are also conducted on 3-stage M/M/1/GD/∞/∞ queuing systems. The coverage analysis supports our preliminary conclusion: when first exposing students to the subject of output analysis on non-terminating systems, strong emphasis should be placed on choosing proper run length and the time devoted to transient analysis can be reduced.
Unfortunately, ACM prohibits us from displaying non-influential references for this paper.
To see the full reference list, please visit http://dl.acm.org/citation.cfm?id=1218524.