Latency-Constrained Task Distribution in Multi-Access Edge Computing Systems

Multi-Access Edge Computing (MEC) paradigm has been widely studied as a potential solution to cope with the challenges emerging from new generations of mobile networks. By processing applications’ data closer to the users, service providers are able to offload origin servers and their underlying network infrastructure, which consequently reduces users’ experienced latency. In this paper, we consider internet-based applications with strict latency tolerance which are primarily enabled by the MEC architecture, e.g., Virtual Reality (VR) control systems, cloud gaming, and smart healthcare. Moreover, nodes at the edge may host application-related tasks as well as assist in their provision. We introduce the Task Distribution (TD) problem, where the objective is to maximize the overall Quality of Service (QoS) while ensuring that tasks’ latency requirements are satisfied. The TD problem is modeled as an Integer Programming problem, taking into account three components: (i) tasks’ priority assignment, (ii) placement and (iii) routing through the MEC network. We propose to approach the TD problem through an approximation scheme, called Shortest-Flow Approximation (SFA), which considers pre-determined network flows selected according to a Betweenness-based criterion. In our experiments, we first extract useful insights on the TD problem that help characterize its solution. Then, we evaluate SFA’s performance empirically across different experimental settings, showing numerical results confirming its proximity to the optimal solution.