Towards cost-effective and resource-aware aggregation at Edge for Federated Learning
Ahmad Faraz Khan, Yuze Li, Xinran Wang, Sabaat Haroon, Haider Ali, Yue Cheng, Ali R. Butt, Ali Anwar
TL;DR
This work tackles the bottlenecks of cloud-centric FL aggregators when deployed at edge data centers by proposing a resource-aware adaptive aggregator that can switch between three methods—Numba-based multi-core, Spark-based multi-node, and Serverless tree-reduction—based on workload and resource state. A RL-driven policy optimizes the trade-off between completion time and cost, achieving substantial gains in latency and cost across diverse model sizes and participation levels. The approach demonstrates that no single method is ideal for all edge workloads, and that dynamic selection yields up to significant improvements in QoS, with practical deployment guidance for edge environments. Overall, the study provides a concrete framework for scalable, cost-efficient FL aggregation at the edge, supported by an extensive empirical evaluation and an emulator for cost-benefit analysis.
Abstract
Federated Learning (FL) is a machine learning approach that addresses privacy and data transfer costs by computing data at the source. It's particularly popular for Edge and IoT applications where the aggregator server of FL is in resource-capped edge data centers for reducing communication costs. Existing cloud-based aggregator solutions are resource-inefficient and expensive at the Edge, leading to low scalability and high latency. To address these challenges, this study compares prior and new aggregation methodologies under the changing demands of IoT and Edge applications. This work is the first to propose an adaptive FL aggregator at the Edge, enabling users to manage the cost and efficiency trade-off. An extensive comparative analysis demonstrates that the design improves scalability by up to 4X, time efficiency by 8X, and reduces costs by more than 2X compared to extant cloud-based static methodologies.