The multiple server effect: Optimal allocation of servers to stations with different service-time distributions in tandem queueing networks

Traditionally, studies on tandem queueing networks concentrate on systems with infinite buffers, exponential service times, and/or single servers where solutions are more tractable. Less research can be found on more general, less tractable systems. We examine multiple-server systems with finite buffers and non-exponential service times, studying the effects of coefficient of variation (cv) of the service-time distribution on the throughput of these systems, where cv differs among stations. Starting with the single station, we examine the effects of cv and the number of servers at the station on the distribution of interdeparture times. This insight helps explain the differences in throughput seen in the single (fast) server vs. multiple (slow) server problem. These results, in turn, shed light on the server allocation problem when cv differs among stations. We present some observations, as well as the intuition behind them.