Marginal productivity index policies for scheduling multiclass wireless transmissions

We address the problem of dynamically scheduling wireless transmissions of multiple heterogeneous users in a centrally controlled single-hop mobile network, with the goal of designing well-grounded and tractable dynamic scheduling policies which come close to optimizing performance objectives of concern. We model the system as a discrete-time Markovian multiclass queue with random binary connectivities, where each class has a dedicated finite buffer. Different levels of tolerance to delay and loss of traffic classes are modeled via corresponding linear holding and/or rejection cost rates. We develop new dynamic index policies, prescribing to give higher service priority to connected classes with larger index values, where the index of a class measures the marginal productivity of work at its current state. Such marginal productivity index (MPI) policies furnish new, insightful structural results. Computational experiments are reported, showing that the proposed MPI policies are nearly optimal and outperform conventional policies across the instances considered