Skip to main content

Federated Deployment

Use federated deployment when the server and clients should run as separate Flower processes, usually on separate machines or containers. This is closer to a real multi-site collaboration than local simulation.

One coordinating server runs SuperLink and submits the app. Each participating client (institution / data holder) runs a SuperNode with its own center config and local PLINK data. Genotype-level work stays on the client that holds the data.

FedGWAS provides two equivalent ways to start that runtime. Both wrap flower-superlink, flower-supernode, and flwr run . local-deployment. Pick one mode for a given run; do not mix them.

ModeEntry pointWhen to use
CLIfedgwas-deployPackage commands on each node
Script deploymentcluster_deployment/scripts/Shell wrappers from a repository checkout

Full role, port, and check details are in Installation and Overview.

Prepare each node​

Install the same FedGWAS version and Flower 1.19.x on the server and every client:

python -m pip install FedGWAS
python -m pip install "flwr[simulation]>=1.19.0,<1.20"

Script deployment also needs a repository checkout on each node:

git clone https://github.com/idsla/Fed-GWAS.git
cd Fed-GWAS
python -m pip install -e .

Each client needs PLINK and only its own genotype files:

plink --version
plink2 --version
client-k/
  configs/center_k.yaml
  data/center_k/study_center_k.{bed,bim,fam}

Confirm the chosen entry point:

fedgwas-deploy --help
ls cluster_deployment/scripts/

CLI​

On the server:

fedgwas-deploy server start --host 0.0.0.0 --daemon --log-file /tmp/superlink.log

On each client, after SuperLink is up (<k> is 1, 2, …):

fedgwas-deploy client start \
  --server <server-ip> \
  --center-id <k> \
  --config configs/center_k.yaml \
  --daemon \
  --log-file /tmp/supernode_k.log

On the server, after every SuperNode has registered:

fedgwas-deploy server run --server <server-ip> --rounds 20 --scale tiny

Stop on each node separately:

fedgwas-deploy server stop    # server node
fedgwas-deploy client stop    # each client node

See CLI Deployment for checks, --num-centers, ports, and troubleshooting.

Script deployment​

Run these from the repository root. On the server:

nohup cluster_deployment/scripts/cluster-start-server.sh --server-ip 0.0.0.0 \
  > /tmp/superlink.log 2>&1 &

On each client:

cluster_deployment/scripts/cluster-start-client.sh \
  --server-ip <server-ip> \
  --client-id <k> \
  --config configs/center_k.yaml

On the server, after every SuperNode has registered:

cluster_deployment/scripts/cluster-run-app.sh \
  --server-ip <server-ip> \
  --rounds 20 \
  --scale tiny

Stop on each node separately:

cluster_deployment/scripts/cluster-stop-all.sh

See Script Deployment for data checks, clock skew, and troubleshooting.

Next steps​