On Bandwidth Reservation in a Multirate Loss Model of Batched Poisson arrival processes

First, we review an extension of the Erlang Multirate Loss Model (EMLM), namely the multirate batched Poisson loss model, in which calls of each service-class arrive in a link of certain capacity following a batch Poisson process and compete for the available link bandwidth under the Complete Sharing (CS) policy. The batch size is generally distributed while the batch blocking discipline is the partial batch blocking, i.e. depending on the available link bandwidth a part of an arriving batch can be accepted while the rest of it is discarded. Second, we propose the multirate batched Poisson loss model under the Bandwidth Reservation (BR) policy. The importance of this proposal is that we can guarantee specific QoS at call-level for each service-class. For the application of the BR policy in the multirate batched Poisson loss model, we study two methods already used in the EMLM: a) the Roberts’ and b) the StasiakGlabowski (S&G) method. The new model (with BR) does not have a product form solution and therefore we propose approximate but recursive formulas for the calculation of various performance measures, such as time and call congestion probabilities and the link utilization. By comparing the analytical with simulation results we show that the S&G method performs better than the Roberts’ method in the case of QoS equalization.