We propose an approach to modeling a production facility that makes many products in large, discrete batches, when demands and the production process are stochastic. This approach combines standard inventory and queueing submodels into classical optimization problems. The models developed in this paper to illustrate the approach represent the production facility by a single-server queueing system or by a network of queues; other systems can also be handled within the approach. The models optimize the approximate operating cost of a given facility over certain simple, plausible control policies; furthermore, the models are tractable enough to be used routinely for design studies.