The performance of prioritized DQDB MAN under bursty tra c environment is studied in this paper. The tagged node model is adopted to simplify the analysis. The processes of the packet arrivals to the tagged node, the empty-slot pattern from upstream, and the request pattern from downstream are assumed to be Markov M o dulated Poisson Processes, a more general and appropriate model to capture the bursty characteristics existing in the tra c ows in high-speed networks. A s t o c hastic reward nets model for the prioritized DQDB MAN is presented. With the help of the Stochastic Petri Net Package (SPNP) developed in Duke University, the steady state and as well as the transient b e h a vior of packet loss ratio of tagged node are investigated. The in uence of the burstiness of the empty-slot pattern, packet arrival and request pattern on the tagged node's performance are examined.