In this paper, we investigate how polling systems with intelligent policies perform compared to the optimal service policy. To this end, we introduce two new policies: an intelligent server idling policy incorporating information about shortterm future arrivals, and an informed switching policy that uses queue lengths to switch to the next non-empty queue. For benchmarking purposes, the optimal service schedule is obtained by formulating the problem as a Mixed Integer Program, with full arrival information. Our results show that with these intelligent policies, polling systems significantly improve in performance, thereby narrowing the gap between polling policies and optimal schedules.
Joshi et al. (Tue,) studied this question.