This repository contains a blueprint and a reference architecture to implement and realize federated learning use cases on Google Cloud, such as:
- Cross-silo federated learning: where the participating members are organizations or companies, and the number of members is usually small such as within a hundred.
- Cross-device federated computation: is a type of federated computation where the participating members are end user devices such as mobile phones and vehicles. The number of members can reach up to a scale of millions or even tens of millions.
This reference architecture is based on Google Kubernetes Engine (GKE) and follows security best practices that help you secure your Google Cloud environment.
This reference architecture configures separated, isolated runtime environments. Each runtime environment gets:
- A dedicated Kubernetes namespace
- A dedicated GKE node pool
You can run potentially untrusted federated learning workloads in these runtime environments, granting these workloads only the minimum permissions that they need.
This reference architecture assumes that you are familiar with GKE and Kubernetes.
For more information about federated learning, see Cross-silo and cross-device federated learning on Google Cloud.
For more information about deploying this reference architecture in your Google Cloud environment, see Deploy the Federated learning reference architecture on Google Cloud