Hybrid flow cuts training time in federated learning across networks
HybridFLow: SDN-Orchestrated Client Partitioning for Hybrid Federated Learning
Networking and Internet ArchitectureMachine Learning
Summary
Federated learning lets different places train a shared AI model without sharing their private data. However, slow network connections can cause delays and make some participants lag behind. The authors created HybridFLow, which uses a central controller to predict communication speeds and split participants into groups that train either together or at their own pace. This approach speeds up training and balances delays and outdated updates. Experiments show it reaches good accuracy faster and with shorter communication rounds than other methods, especially when data is unevenly distributed.
What this means in practice
- •For network operators: Use centralized network insights to group participating sites and optimize training speed in federated learning setups.
- •For cloud service providers: Improve federated learning services by orchestrating client training modes based on communication time predictions to reduce delays and speed convergence.$Commercial implications: Enables cloud platforms to offer federated learning with faster training times and improved accuracy guarantees to enterprise customers.
Authors
Osama Abu Hamdan, Rabin Pandey, Hao Che, Engin Arslan, Md Arifuzzaman
Abstract
Cross-silo Federated Learning (FL) enables geographically distributed institutions to collaboratively train machine learning models without sharing raw data. In wide-area deployments, however, communication delays often dominate round completion time and exacerbate the straggler effect. Hybrid FL addresses this challenge by combining synchronous and asynchronous client participation, but effective partitioning requires visibility into network conditions such as shared bottlenecks, link utilization, and path contention that individual clients cannot observe. We present HybridFLow, a closed-loop SDN-driven orchestration framework that integrates network-layer intelligence directly into hybrid FL. Leveraging the SDN controller's global topology view, HybridFLow generates calibrated per-client communication-time estimates before each training round and uses them to partition clients into synchronous and asynchronous groups while balancing round latency and update staleness. After each round, measured communication times are fed back to the controller to continuously refine future predictions. Experimental results across multiple network topologies show that HybridFLow reaches 80% target accuracy 33-40% faster than SmartFLow and reduces average round duration by 30-40 seconds, while FedAsync fails to reach the target accuracy under non-IID data distributions.