Distributed Task Offloading in MEC Networks for Temporary Peaks in Demand

Abstract

Multi-Access Edge Computing (MEC) network planning is performed considering a forecast of estimated workload in each coverage zone with the aim of offloading computationally expensive tasks from user's devices to the nearest MEC Data Center (MEC-DC). Nevertheless, in some scenarios, these forecasts are exceeded temporarily due to sudden peaks in demand in a determined MEC-DC, making its planned computing resources (i.e., MEC servers) scarce, and introducing the need of a strategy to face this increment in demand. In this paper, we propose and evaluate an Integer Linear Programming (ILP) model for optimizing the task offloading considering a previously defined MEC network topology. Our model is based on the possibility of offloading some tasks to MEC-DCs different to the initially planned (nearest to the user) one, as long as the latency requirements are met, and the allocated server has enough idle computing power. Results show that the proposed strategy considerably increases the capacity of the network to face sudden workload increments compared to an approach that only assigns the nearest MEC server to every user.