Analysis of Scheduling Policies for a M/G/1 Queue with Rework.

摘要

This thesis analyzes a multi-class M/G/1 priority queueing system in which distinct job types require one service cycle and, with non-zero probability, require a second service cycle. The main objective is to find a new heuristic scheduling policy that minimizes the long-run expected holding and preemption costs. Arrival rates, service rates, and the probability of undertaking second service are all class specific. A mean value analysis (MVA) approach was employed to derive the long- run mean time in queue for each job type under each policy, thereby providing the appropriate cost equations. Numerical experiments suggest that the preemptive resume scheduling policy yields the lowest cost most frequently.