Analysis of a multi-priority dual queue with preemptive and non-preemptive service disciplines

Abstract

Traffic management in communication environments like the Internet is vital in meeting the demand of users. The recent introduction of various dynamic scheduling algorithms has been to target improvements in QoS (Quality of Service). The Multi-Priority Dual Queue (MPDQ) was designed to reduce congestion levels in communications models with finite buffers. By splitting a single finite buffer space into two queues, known as dual queues, and by administering a priority scheme to arriving customers, the MPDQ has shown its advantages to communications providers in QoS. The MPDQ is especially relevant for the Internet’s development, and also has merits for use in certain manufacturing systems where a second holding area is available. In this paper we use the matrix analytic methods to define the state rate transitions and uniquely implement them into an algorithm to obtain a solution. We analyze the differences in loss between a MPDQ and a basic single queue with preemptive and non-preemptive priority service disciplines and then prove the results in Bedford and Zeephongsekul (EJOR, 2003) are not correct. And we analyze the differences in loss between preemptive and non-preemptive service disciplines and show that non-preemptive service discipline is superior than the preemptive scheme.