A Swarm-based Approach for Function Placement in Federated Edges

Multi-access Edge Computing (MEC) provides cloud computing capabilities at the edge by offloading users' service requests on MEC servers deployed at Base Stations (BS). Optimising the resource allocation on such distributed units in a physical area such as a city, especially for compute-intensive and latency-critical services, is a key challenge. We propose a swarm-based approach for placing functions in the edge using a serverless architecture, which does not require services to pre-occupy the required computing resources. The approach uses a probabilistic model to decide where to place the functions while considering the resources available at each MEC server and the latency between the physical servers and the application requester. A central controller with a federated view of available MEC servers orchestrates functions' deployment and deals changes available resources. We compare our approach against the Best-Fit, Max-Fit, MultiOpt, ILP and Random baselines. Results show that our approach can reduce the latency of applications with limited effect on the resource utilisation.