Message delays at packet-switching nodes serving multiple classes

A packet-switching node is modeled as a single-server queueing system. Messages in class c have J_c packets, each of which has a length which may be drawn from a general service time distribution prescribed on the basis of the packet´s position in the message. All of the messages´ packets arrive at the instant of message arrival; each packet within the message has a preassigned priority. The first packet of the message joins a transmission queue at the message arrival instant, there being one queue for each distinct priority. Upon completion of transmission of a given packet within a message, its successor packet joins the queue of appropriate priority. Packets which are waiting in the queues are serviced by the trunk under a nonpreemptive priority discipline. A method is presented for finding average message system times for the service discipline just described. This class of service disciplines includes several service disciplines which have been reported in the literature including the PR+RR, the HOL, and HOLP, the RR, and others that are discussed herein. Some numerical examples are presented