Two-phase service systems with one input flow of requests

The two-phase service system of the paper has an arbitrary number of servers in both phases, a buffer of an arbitrary length between the phases, blocking the first phase once the buffer is full, and with feedbacks between the phases. A numerical method is shown for this system, the maximal system throughput is found, and possible applicability of the method is discussed to systems with an Erlang service duration distribution function in the second phase.