Multi-FedLS: a Framework for Cross-Silo Federated Learning Applications on Multi-Cloud Environments

TL;DR

This paper introduces Multi-FedLS, a comprehensive framework for efficient cross-silo federated learning across multi-cloud environments, leveraging preemptible VMs to reduce costs and execution time while maintaining fault tolerance.

Contribution

It formally describes the Multi-FedLS resource management framework and demonstrates its effectiveness through experiments on real multi-cloud setups.

Findings

Multi-FedLS reduces execution time and costs in federated learning.

The framework effectively manages preemptible VMs with fault tolerance.

Experimental validation on AWS and GCP confirms practical applicability.

Abstract

Federated Learning (FL) is a distributed Machine Learning (ML) technique that can benefit from cloud environments while preserving data privacy. We propose Multi-FedLS, a framework that manages multi-cloud resources, reducing execution time and financial costs of Cross-Silo Federated Learning applications by using preemptible VMs, cheaper than on-demand ones but that can be revoked at any time. Our framework encloses four modules: Pre-Scheduling, Initial Mapping, Fault Tolerance, and Dynamic Scheduler. This paper extends our previous work \cite{brum2022sbac} by formally describing the Multi-FedLS resource manager framework and its modules. Experiments were conducted with three Cross-Silo FL applications on CloudLab and a proof-of-concept confirms that Multi-FedLS can be executed on a multi-cloud composed by AWS and GCP, two commercial cloud providers. Results show that the problem of executing Cross-Silo FL applications in multi-cloud environments with preemptible VMs can be efficiently resolved using a mathematical formulation, fault tolerance techniques, and a simple heuristic to choose a new VM in case of revocation.