Quality of Information aware scheduling in task processing networks

We investigate Quality of Information (QoI) aware scheduling in task processing networks. Specifically, we consider the scenario where a network sequentially receives tasks from an end user, utilizes its resources to process them, and sends back its response. The utility derived by the end user from this response depends on both the accuracy and the freshness of the information. There is often a trade-off between these two attributes and we present a model that quantifies this dependence. Using dynamic programming and optimal stopping theory, we characterize the optimal scheduling policy that maximizes the time average utility delivered by the network. We show that for many scenarios of practical interest, the optimal policy has a simple threshold structure. We also propose a method to approximately compute the threshold in closed-form. This work takes a step towards incorporating application aware objectives in making optimal scheduling decisions.