On the allocation of computing tasks under QoS constraints in hierarchical MEC architectures

In this study, a model for the allocation of processing tasks in Mobile Edge Computing (MEC) environments is put forward, whereby a certain amount of workload, coming from the base stations at the network edge, has to be optimally distributed across the available servers. At first, this allocation problem is formulated as a centralized (offline) optimization program with delay constraints (deadlines), by keeping into account server qualities such as computation speed and cost, and by optimally distributing the workload across a hierarchy of computation servers. Afterwards, the offline problem is solved devising a distributed algorithm, utilizing the Alternating Direction Method of Multipliers (ADMM). Selected numerical results are presented to discuss the key features of our approach, which provides control over contrasting optimization objectives such as minimizing the energy consumption, balancing the workload, and controlling the number of servers that are involved in the computation.