Efficient heuristics for simulating rare events in queuing networks

In this thesis we propose state-dependent importance sampling heuristics to estimate the probability of population overflow in queuing networks. These heuristics capture state-dependence along the boundaries (when one or more queues are almost empty) which is crucial for the asymptotic efficiency of the change of measure. The approach does not require difficult (and often intractable) mathematical analysis or costly optimization involved in adaptive importance sampling methodologies. Experimental results on tandem, parallel, feed-forward, and a 2-node feedback Jackson queuing networks as well as a 2-node tandem non-Markovian network suggest that the proposed heuristics yield asymptotically efficient estimators, sometimes with bounded relative error. For these queuing networks no state-independent importance sampling techniques are known to be efficient.