local-docker-setup

Local docker instance for dpsa4fl

This file describes how to setup and start a local dpsa4fl infrastructure instance, consisting of two aggregation servers. With it, you can test differentially private, federated ML projects, such as our example project, locally on your development machine.

Requirements

The only requirement is a working docker installation. The build process itself runs inside a container.

Downloading

To get all required files, you have to clone this repository. It contains git submodules, so the --recursive option has to be used to clone them as well:

git clone --recursive https://github.com/dpsa-project/dpsa4fl-infrastructure.git

Building

Switch into the local-docker-setup subfolder. Inside, build all required docker images by running the following command:

~/dpsa4fl-infrastructure/local-docker-setup> ./build.sh

Docker will build all required janus server executables, this might take some minutes.

Running

To start all required containers, use docker-compose up:

~/dpsa4fl-infrastructure/local-docker-setup> docker-compose up

To stop this infrastructure instance, run docker-compose down. This will remove all containers, volumes and networks created by the previous command.

Updating

If you want to update to a new commmit of this repository, you have to also update the submodules. Thus, run:

git pull && git submodule update

After updating, you have to build all docker images again.