Performance Forecasting of Discrete-Time Priority Retrial Queue With Its Application in Cognitive Radio Networks

Abstract

Queueing modeling and optimization of high-speed digital systems provide a tool to the system operators and administrators to build an economic system and to analyze overcrowding situations in various digital systems such as computer systems, cellular mobile webs, cognitive radio networks (CRNs), and many more. CRNs enable a more efficient and flexible use of the radio spectrum by allowing unlicensed users (ULUs) to dynamically access and share frequencies with licensed users (LUs), ensuring that spectrum resources are fully utilized while avoiding interference with licensed operations. This study aims to focus on complex, real-world challenges faced in CRNs and to resolve its few congestion issues through a queue-theoretic approach. The congestion issues faced by CRN can be resolved by modeling the Geo^X/G/1 priority retrial model with multielective services under the Bernoulli vacation schedule wherein the server's time can be better allocated to users to improve their grade of service. We can see how CRN can be seen as a discrete-time retrial queueing system according to the following formulation. In CRNs, there are two types of users: LUs and ULUs. The former is given priority over the latter in the sense that LUs can forestall the transferences (transmissions) of ULUs. In this perspective, the LU channel acts as a server that can be approachable by ULUs practically. The LU and ULU data packets, links, or sessions act as customers, which usually attach to the virtual track of blocked users if they do not get instant entrance. Moreover, each LU channel either provides access to the entering user or may cease providing service for some span of time called vacation time. This queueing process is termed as Bernoulli vacation (BV). Also, we apply an admission control policy (ACP), which controls the number of arrivals. In this study, we perform a numerical simulation through which we can conclude that the average number of data packets in CRN and expected total cost increases linearly by upgrading either the admission control probability or arrival rate. Also, our study suggests that the average system size decreases with an increase in the probability that a licensed unit joins the system. Further, multicriteria optimization is used to obtain Pareto optimal solutions of expected total system cost and expected system waiting time in CRN.