Age of information (AoI) is a key metric to measure the freshness of information for IoT applications. This paper investigates a scheduling problem in a 5G network where there is an AoI threshold for each source node at the edge base station (BS). Our goal is to design a 5G scheduler to minimize the proportion of time when the information stored at the source is outdated, i.e., when the AoI is beyond its threshold. For performance benchmark, we develop an offline computational procedure to find a lower bound for the objective value. Then we derive a property called uniform fairness for an offline optimal scheduler and use this property as a guideline to develop an online 5G scheduler-Aequitas. To meet the sub-millisecond real-time requirement in 5G, we implement Aequitas on an off-the-shelf GPU. Through experiments, we show that the objective achieved by Aequitas is close to the lower bound and its running time is under 1 ms.