Queuing Models for Prioritized Packets

Abstract

In modern data communications networks, one of the most important problems is to meet the quality of service (QoS) for all traffic. Real time traffic, e.g. live audio or video is delay sensitive, whereas nonreal time traffic, e.g. file transfer, is loss sensitive. One method of solving this QoS problem is the use of priority control. That is, packets are classified as either class-!, high priority, or class- 2, low priority. In this thesis, the researcher studies the buffer behavior of a fileserver in a network to meet the QoS for all traffic. In particular, he provides a discrete-time queuing model under priority control that is analyzed using queueing theory. This model is also useful for the design and the performance evaluation of many communications systems. The assumptions of the model have been made so general that many previous models can be considered special cases of it. In this model, the researcher assumes that the arrival process is deterministic of one batch. That is, in each slot a batch containing packets of only class-! or only class-2 arrives. The batch size is assumed to be random. The service time has geometric distribution and the buffer size is infinite. The priority control is nonpreemptive and the approach of analysis is a random variable. The results of the analysis include deriving the PGF of sys­ tem occupancy and the marginal PGFs of the number of class-!