[en] Large High Energy and Nuclear Physics (HENP) databases are commonly stored on robotic tape systems because of cost considerations. Later, selected subsets of the data are cached into disk caches for analysis or data mining. Because of the relatively long time to mount, seek, and read a tape, it is important to minimize the number of times that data is cached into disk. Having too little disk cache will force files to be removed from disk prematurely, thus reducing the potential of their sharing with other users. Similarly, having too few tape drives will not make good use of a large disk cache, as the throughput from the tape system will form the bottleneck. Balancing the tape and disk resources is dependent on the patterns of the requests to the data. The authors describe a simulation that characterizes such a system in terms of the resources and the request patterns. The authors learn from the simulation which parameters affect the performance of the system the most. The authors also observe from the simulation that, there is a point beyond which it is not worth investing in additional resources as the benefit is too marginal. The authors call this point the 'point-of-no-benefit' (or PNB), and show that using this concept we can more easily discover the relationship of various parameters to the performance of the system