The use of priorities allows us to improve the quality of service of inhomogeneous customers in telecommunication networks, inventory and health-care systems. An important modern direction of research is to analyze systems in which priority of a customer can be changed during his/her stay in the system. We considered a single-server queuing system with a finite buffer, where two types of customers arrive according to a batch marked Markov arrival process. Type 1 customers have non-preemptive priority over type 2 customers. Low priority customers are able to receive high priority after the random amount of time. For each non-priority customer accepted into the buffer, a timer, which counts a random time having a phase type distribution, is switched-on. When the timer expires, the customer with some probability leaves the system unserved and with the complimentary probability gains the high priority. Such a type of queues is typical in many health-care systems, contact centers, perishable inventory, etc. We describe the behavior of the system by a multi-dimensional continuous-time Markov chain and calculate a number of the stationary performance measures of the system including the various loss probabilities as well as the distribution function of the waiting time of priority customers. The illustrative numerical examples giving insights into the system behavior are presented.

A model of a cell of a communication network, divided into zones, with the dependence of the users' service time on the zone, in which they are located, is considered. The arrival flow of users is defined by a marked Markovian process. The number of users that can receive service in a cell simultaneously is finite. If the number of users reached the limit, then new users are lost, except the users that come to the cell already during the service (handover users). For the short-term storage of such users, there is a buffer of a finite capacity. Cell operation is modeled by a multi-server queueing system. The behavior of this system is described by a multi-dimensional Markov chain. The algorithms for calculating the stationary distribution of this chain and the main characteristics of the quality of service have been developed.