Performance Analysis of Integrated Services on a Single Server System

Abstract The integration of time-critical (TC) and non-time-critical (NTC) services on a single resource, in which the TC traffic is given a higher service priority, can be modeled by a server that alternates between a work cycle and a vacation cycle. When a stream traffic is partitioned into packets, adjacent packets are correlated. With the TC traffic given a higher service priority than the NTC traffic, the correlation present in the TC traffic has the effect of increasing the queue size of the NTC traffic. Generating functions for the delay and queue length distributions of the NTC traffic, incorporating the signal dependency (packet-to-packet correlation) within the TC traffic, are derived. Exact expressions for the means and standard deviations as functions of the throughput are then obtained and computed. The incorporation of the packet-to-packet correlation in the analytical expressions offers a unified approach for analyzing the performance of integrated services systems. Numerical results indicate that the presence of correlation due to packetization of stream traffic introduces a significant effect on the queue length distribution of the low priority (NTC) traffic.