Congestion Probabilities in Erlang-Engset Multirate Loss Models under the Multiple Fractional Channel Reservation Policy

Abstract A communication link that accommodates different service-classes whose calls have different bandwidth requirements and compete for the available bandwidth under the Multiple Fractional Channel Reservation (MFCR) policy is considered. The MFCR policy allows the reservation of real number of channels in order to favor high speed calls. Two call arrival processes are studied: i) the Poisson (random) process and ii) the quasi-random process. In the first case, calls come from an infinite number of sources while in the second case calls are generated by a finite number of sources. To determine call blocking probabilities for Poisson arriving calls, recursive formulas are proposed based on reverse transition rates. To determine time and call congestion probabilities for quasi-random arriving calls, recursive formulas are proven based on the fact that the steady state probabilities cannot be described by a product form solution. The accuracy of the new formulas is verified through simulation.