Study of Queueing Systems with phase Distributions
Abstract
The paper presents the obtained results of the analysis of general type queuing systems (QS), which are described by the second-order Cox phase distributions C2/C2/1. The use of phase distributions in queueing theory has been known for a long time. This class of distributions is suitable as a mathematical apparatus for analyzing teletraffic in future generation networks, which is characterized by high variability and has a self-similarity effect. Method. To find the average waiting time for requirements in the queue, the spectral method was used to solve the Lindley integral equation with finding the zeros and poles of the spectral expansion in the form of a ratio of rational functions, which in this case are reduced to fourth-degree polynomials. This method is less labor-intensive than the classical matrix-geometric method, which uses Kronecker operations and iterations. Results. The paper presents a numerical-analytical model for the average waiting time of requests in a queue for a system with second-order Cox distributions. To assess the accuracy of the obtained solution, computational experiments were conducted for low, medium, and high values of the variation coefficients of time intervals and system load. The obtained data allow us to estimate the performance of switching nodes under increasing stochastic uncertainty in telecom traffic. Consequently, neglecting the phase structure of the distribution laws of the QS when designing highly loaded systems for various purposes leads to an underestimation of packet delays in switching nodes by many times. Comparison with similar systems demonstrates the complete adequacy of the obtained results.