Mobile edge computing (MEC) has recently been introduced as a key technology, emerging in response to the increased focus on new heterogeneous computing applications, resource-constrained mobile devices, and the long delay of traditional cloud data centers. Despite considerable research attention to understanding how the heterogeneous latency-critical application requirements can interact with a MEC system, there remains a dearth of literature on deploying a flexible MEC infrastructure at the mobile operator to meet the demands of an anticipated heterogeneous mobile traffic.
From a dual perspective, this paper addresses the design and dimensioning of a machine-driven Network Function Virtualization-enabled MEC infrastructure problem. The proposed approach leverages a neural network model, a subset of machine learning, to predict the number of service function chains (SFCs) required for the time-varying mobile traffic load and to proactively auto-scale the different types of virtual service instances. A Mixed-Integer Linear Program (MILP) is then employed to create a physical MEC system design by mapping the predicted virtual SFC networks to the MEC nodes while minimizing deployment costs. The numerical results demonstrate that the machine learning model achieves a high prediction accuracy of 95.6%, highlighting the added value of using the ML technique at the edge network. This approach reduces deployment costs while ensuring delay requirements for different latency-critical applications and high acceptance rates.