A hysteresis type strategy for server reservation in a multi-server queueing system with priority and retrials

This work deals with a possible model of a cell of cognitive radio architectures via a multi-server queueing system with two different types of requests and preemptive priority of one type of requests on the other. Service demands arrive according to Markovian Arrival Processes in order to capture traffic correlation. Cognitive engine tries to find a solution or optimize a performance goal based on the received inputs that define the radio’s current internal state and operating environment. Hence, beside priority mechanisms, a possible deactivation of information to transmit is considered via a hysteresis type strategy for the access to the service, with the aim of smoothing the effects of service interruptions for low priority requests. Fixed values for the thresholds of the strategy are assumed. The system is statistically described by a level-dependent multi-dimensional Markov chain, which allows the computation of stationary probabilities and robust performance indices. Numerical results are presented to compare the system performances in the case of uncorrelated flows and flows with different correlation levels.