Mobile Edge Computing (MEC) networks have been considered a potential solution for coping with the challenges imposed by novel latency-sensitive applications. In this work, we address the time-evolving Task Distribution Problem (TDP) to efficiently allocate network resources, guaranteeing high throughput and meeting the strict latency requirements. We model TDP as a two-stage stochastic programming problem and solve it by handling its components in different time scales. Due to its reactive characteristics, this approach has the potential to achieve good results in practice.