Closed queuing networks have been widely used in recent years to model manufacturing systems, especially modern flexible manufacturing systems. These models have tended to be descriptive. In this paper, we explore the application of optimization procedures to closed queuing network models for the purpose of selecting optimal parameter values for manufacturing system operation. Optimal levels of work-in-process inventory are determined jointly with workcentre processing rates. A simple procedure is developed for obtaining the optimal work-in-process levels and processing rates for balanced workload systems. The tradeoff between work-in-process levels and processing speeds is examined. In general, investigation of model results indicates the importance of jointly determining processing rates at all workcentres and shop loading. Small facilities are seen to have an advantage over large facilities.