Performance Modeling of Cellular Systems with Finite Processor Sharing Queues in Random Environment, Guard Policy and Flex Retrial Users

We investigate a two-station retrial queueing system to model the access in modern cellular networks managed by two service providers. Each provider owns a single access point, which operates under processor sharing discipline, and accepts three types of users: the handover users and, the originating subscribers and the originating flex users. At the arrival epoch, a flex user connects with the provider, which offers the largest data rate. Each access point can admit a limited number of users and employ a guard bandwidth policy in order to prioritize the handover users. Both blocked handover and originating subscriber users are lost. Blocked flex users join a virtual orbit queue of infinite capacity from where they retry independently to connect with the service provider that offers the largest data rate at their retrial time. Moreover, the system operates in varying environmental conditions. Using the matrix analytic formalism we construct a four-dimensional Markovian model, which allows to represent accurately the types of user behavior and the environmental aspects in cellular networks. We perform a steady-state analysis and a study of the main performance metrics.