Performance analysis of non‐preemptive hol GE/G/1 queue with two priority classes of sip‐t signaling system in carrier grade VoIP network

Abstract

For analytic simplicity, the priority queueing of a VoIP network is modeled under the assumption that the traffic arrival process is Poisson distributed and the service time is exponentially distributed, while knowing that the holding times of calls are not an exponential distribution. Under this condition, the non‐Markovian process (e.g. Supplementary Variable Technique) is often considered. However, the analysis for priority processes has always been difficult to tackle directly, particularly when the number of priorities is greater than two. Earlier studies were therefore restricted to two priority classes (e.g. M1 B1, M2 B2/G1, G2/1). Although the analytical results are available, very complicated closed‐form expressions of mean queue length are still required for each class. The objective of this paper is to obtain a simple and analytic closed‐form for the mean length of an HOL (Head of Line) GE/G/1 queue with more than two priority classes. Compared with M 1 B1, M2 B2/G1, G2/1, the HOL GE/G/1 priority queue computations of an SIP‐T signaling system require only a few parameters (e.g. mean and variance) since measurements of actual interarrival or service times are often constrained. The performance analyses from mathematical calculations and computer simulations further confirm the accuracy and robustness of an HOL GE/G/1 queue. Hence, the optimum values of interarrival time SQVs (Squared Coefficient of Variations) and service time SQVs can be determined for GM (Guaranteed SIP‐T messages) and NGM (Non‐guaranteed SIP‐T messages).