Reliability Analysis of Cognitive Radio Networks

Abstract

The current paper deals with the simulation of a queuing model developed to evaluate the performance of a cognitive radio network and its reliability.We take into consideration two interconnected subsystems, the first one is dedicated to primary users (PU). The number of sources is finite, and each source generates a primary request after an exponentially distributed interval of time, these requests then are then sent to a single server called Primary Channel Service (PCS), under the assumption that the service times are distributed exponentially as well. The service is carried out in the order of the arrivals of primary calls. The second part of the model is associated to secondary users (SU), which has also a finite source with exponentially distributed request generation time and it is assumed that the service time at the Secondary Channel Service (SCS) with a single server is exponentially distributed. However, the two subsystems operate in the following way. Each generated primary request is directed to the primary server in order to check its accessibility, in case that the service unit is free, the service starts instantly. If the primary unit is already busy with another primary request, the call joins a FIFO queue. However, if the primary unit is busy by processing a service for a secondary user, this latter service stops immediately and should be sent back to the (SCS), based on the availability of the secondary server this postponed task either starts the service again or joins the orbit.In the other hand, the secondary requests are directed to the secondary server to verify its availability, if the aimed server is accessible, the service starts immediately, otherwise these secondary requests try to join the (PSU) and if it is idle the service starts. If not, they join to the orbit at SCS. Postponed requests in the orbit retry to be served after an exponentially distributed interval of time.In the current work both service units are subject to some random breakdowns, in such case the interrupted requests are sent either to the queue or to the orbit, respectively. It is assumed that the operation and repair times of the given server are generally distributed. We use Hypo-Exponential, Hyper-Exponential and Gamma distributed times because by assuming the same means and variances the effect of the distribution could be analyzed. Due to the page limitations we deal with the effect of the distribution of operation and service time distributions, only on the mean response time of the secondary server.Assuming that all the random times concerned in this model are independent of each other the primary aim of the present paper is to show the impact of the distribution of the operation and repair time on the mean response time of the secondary users as a function of the failure and repair intensity, respectively Several Figures are generated to visualize the difference due to the distributions.