Maximizing Service Reward for Queues with Deadlines

In this paper, we consider a real-time queuing system with rewards and deadlines. We assume that the packet processing time is known upon arrival, as is the case in communication networks. This assumption allows us to demonstrate that the well-known earliest-deadline-first policy performance can be improved. We then propose a scheduling policy that provides excellent results for packets with rewards and deadlines. We prove that the policy is optimal under deterministic service time and binomial reward distribution. In the more general case, we prove that the policy processes the maximal number of packets while collecting rewards higher than the expected reward. We present simulation results that show its high performance in more generic cases compared to the most commonly used scheduling policies.