EneA-FL: Energy-aware orchestration for serverless federated learning

Andrea Agiollo, Paolo Bellavista, Matteo Mendula, Andrea Omicini

Abstract: Highlights•EneA-FL targets serverless computing for FL in Fog and constrained environments.•EneA-FL introduces a resource-conscious worker selection policy in FL.•EneA-FL optimises resource consumption while achieving efficient model convergence.•EneA-FL achieves a 3x reduction in energy consumption and 50% faster convergence.•EneA-FL is flexible and can be applied to a diverse array of learning tasks.•EneA-FL uses containers and microservices to shift from the Cloud to the Edge.