Reducing Delay in Preemptive Repeat Priority Queues

The classical model of the preemptive repeat priority queue assumes that there is a single distribution from which service times for a given class are selected, regardless of the number of preemptions. This paper presents three alternatives for improving service progressively to preempted customers: (1) changing the service distribution, (2) preventing further preemptions, and (3) promoting customers to the next higher priority class. The Laplace-Stieltjes transform of the flow time distribution is found for each class in a model that can employ these three alternatives in arbitrary combinations.