DEPARTURE PROCESSES FROM A LOSS SYSTEM WITH TWO STATIONS AND ONE SERVER

We consider a loss system with two stations and one server. The server switches between the stations according to a policy which depends on the last station served and the state of the system just after a departure. At each station, the arrival process is Poisson and the service distribution is negative exponential. We assume independence among arrivals and services. We model this system using a Markov renewal process embedded at departure times. Using standard filtering techniques we determine the interdeparture distribution from either of the stations. Since the exact expression can be difficult to obtain, a simpler approximation is calculated through the method called One-Step Projection. Comparative studies are made between the two solutions and numerical results are given. Some of the results suggest that one effect of switching servers is to increase the variability of the departure processes over those of comparable non-switching systems.