FedFetch: Faster Federated Learning with Adaptive Downstream Prefetching

2025-04-21

AI Summary

In federated learning, jointly applying client sampling and model compression causes infrequently selected heterogeneous clients to repeatedly download stale model states, creating a severe downlink communication bottleneck. To address this, we propose the first adaptive downlink prefetching mechanism: leveraging local client state prediction and dynamic scheduling, it proactively fetches future-required model parameters multiple rounds in advance, thereby decoupling the adverse interaction between sampling and compression. Our mechanism is compatible with mainstream compression techniques (e.g., Top-k sparsification, quantization) and diverse sampling protocols (e.g., random, stratified). Extensive experiments under heterogeneous settings demonstrate an end-to-end training speedup of 1.26× and a 4.49× reduction in downlink download time. The implementation is publicly available.

Technology Category

• Machine Learning: Distributed Machine Learning & Federated Learning
• Search and Optimization: Sampling/Simulation-based Search
• Data Mining & Knowledge Management: Data Compression

Application Category

• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Federated learning (FL) is a machine learning paradigm that facilitates massively distributed model training with end-user data on edge devices directed by a central server. However, the large number of heterogeneous clients in FL deployments leads to a communication bottleneck between the server and the clients. This bottleneck is made worse by straggling clients, any one of which will further slow down training. To tackle these challenges, researchers have proposed techniques like client sampling and update compression. These techniques work well in isolation but combine poorly in the downstream, server-to-client direction. This is because unselected clients have outdated local model states and need to synchronize these states with the server first. We introduce FedFetch, a strategy to mitigate the download time overhead caused by combining client sampling and compression techniques. FedFetch achieves this with an efficient prefetch schedule for clients to prefetch model states multiple rounds before a stated training round. We empirically show that adding FedFetch to communication efficient FL techniques reduces end-to-end training time by 1.26$ imes$ and download time by 4.49$ imes$ across compression techniques with heterogeneous client settings. Our implementation is available at https://github.com/DistributedML/FedFetch

Problem

• Reduces communication bottleneck in federated learning
• Mitigates slowdown from straggling clients in FL
• Improves efficiency of client sampling and compression

Innovation

• Adaptive downstream prefetching for FL
• Reduces download time with prefetch schedule
• Combines client sampling and compression efficiently