EneA-FL: Energy-aware Orchestration for Serverless Federated Learning


Federated Learning (FL) represents the de-facto standard paradigm for enabling distributed learning over multiple clients in real world scenarios. Despite the long strides reached in terms of accuracy and privacy awareness, the real adoption of FL in real-world scenarios, in particular in industrial deployment environments, is still an open thread. This is mainly due to privacy constraints and to the additional complexity stemming from the set of hyperparameters to tune when employing AI techniques on bandwidth-, computing-, and energy-constrained nodes. Motivated by these issues, we focus on scenarios where participating clients are characterised by highly- heterogeneous computing performance and energy budgets, and propose EneA-FL, a novel scheme for serverless smart energy management aimed at facilitating the training process and the interaction between IoT and edge nodes. In particular, the proposed middleware provides FL nodes with a containerised software module that monitors the local energy budget, computational capability, and accuracy; at the same time, it efficiently manages the interaction with the central aggregator to determine which node has to be included in training rounds, transparently, also to simplify the design and development of FL applications on top of our middleware. Finally, extensive experiments over multiple scenarios demonstrate how the proposed serverless middleware can lead to three-times lower energy consumption and less than half of overall time required for training when compared with popular solutions available in the literature.

keywordsServerless, Federated Learning, Energy Management, Internet of Things, Resource-constrained Learning
origin event
journal or series
book Future Generation Computer Systems (FGCS)

Partita IVA: 01131710376 — Copyright © 2008–2023 APICe@DISI – PRIVACY