11.6 Summary

< Free Open Study >

In this chapter we introduced some simple rules to consider when selecting a modeling tool for a specific modeling project. We indicated that if time and money were not a factor we would use all methods. First, we would apply analytical modeling to quickly eliminate alternative designs that would not meet the needs of the target system. Second, we would apply Petri net models to further compare and remove alternatives from consideration. Third, we would use simulation to study a few alternative components or systems. Simulation provides for very detailed modeling of components or operations if so desired. The fourth tool to apply would be testbeds. These are much more complex, and we would use this alternative when we are down to only a few alternatives, possibly only one, that need to be validated.

Since it is not a perfect world, time and money do count; therefore, our modeling tool selection would be driven by these considerations. If cost is of paramount importance, we may look to analytical modeling, since it is relatively cheap if we happen to have queuing analysts on our staff. If cost is not a problem, then building testbeds would be the way to go. If cost falls somewhere between this, we would choose simulation or Petri nets. If time is of the essence, we would also recommend queuing theory over the others, since a model can be developed and analyzed. If time is available, then simulation or testbeds would be appropriate choices.

After this discussion, the chapter moved on to examining some of the components of a modeling project that also assist us in deciding on which modeling tool to apply. The metrics we need and their fidelity or accuracy will also push us toward specific tools. If we need very accurate information, we may wish to use testbeds and empirical models, since we are measuring the real system or a prototype of it. If we are less concerned with accuracy, we may wish to use analytical models, since they can be easily constructed and provide coarse-grained analysis.

The chapter then goes on to discuss some of the implications of modeling a system-for example, how to determine if the model's data are correct, or if the results are good or bad. Interpretation of results is dependent on knowing the measurements being taken and their relationship to important systems metrics, such as throughput, utilization, and response time.

< Free Open Study >