Delayed Preemption: A New Policy for Balancing the Age of Information in Prioritized Networks

In this paper, we study the information freshness in a network with two information classes, where class 1 packets have a priority higher than class 2 packets. We propose a new queueing policy with delayed preemptions, that is, when a class 1 packet arrives while a class 2 packet is in service, instead of the usual immediate preemption, a deadline is assigned for the service of class 2 packets. By the end of the mentioned deadline, if the low-priority packet is still in service, it is preempted. The main goal of this delayed preemption policy is to balance the freshness metrics of the information classes, which is an intrinsic attribute in some applications, e.g., caching of dynamic contents, where the freshness of some contents is more critical than the others. For exponentially distributed deadlines, by exploiting a stochastic hybrid system (SHS), we evaluate the average age of information (AoI) and average peak AoI (PAoI). We also find general formulas for the average PAoIs of the classes for arbitrarily distributed deadlines and use them to find the average PAoI for deterministic deadlines. By numerical results, we investigate the performance of this policy in balancing the average AoIs/PAoIs of information classes.