We discuss stability of a FIFO two-station queueing network of the type introduced by Bramson (1995), in which the "usual" traffic conditions do not ensure stability. We present further results and simulations showing how stability can be effected by the probability distributions of components involved; expected values (e.g., rates) alone are not enough to determine stability. Whereas this compliments some results of Dai, Hasenbein and Vande Vate (2001), our FIFO examples offer an opposite kind of result: higher variability can make the stability region larger (as opposed to smaller). Some conjectures are given too.
Last Updated: February 7, 2002