Dynamic scheduling and maintenance for a two-class queue with a deteriorating server

We consider a queueing control problem motivated by a scheduling and maintenance problem in semi-conductor manufacturing. Conditions are given under which it is optimal, relative to both infinite-horizon discounted-cost and average-cost criteria, to perform scheduling according to a priority policy, and maintenance according to a threshold policy. We also provide empirical evidence that, even when the aforementioned conditions do not hold, such policies provide nearly-optimal performance. In addition, the empirical results indicate the importance of performing preventive maintenance.