The principal investigator (PI) will explore buffer management policies for ATM (Asynchronous Transfer Mode) networks. Traffic concentrators and switch buffers will be modeled as queues with impatient customers, possibly belonging to different priority classes. The impatient customers correspond to cells or packets from time-constrained traffic such as voice or video. The focus here is the development of optimal scheduling policies which, for example, maximize the number of customers served before their respective deadlines. The techniques used will be based on sample path comparisons, stochastic dominance, and stochastic dynamic programming.