master-thesis-code

Dependencies

go v1.20.5

Docker

Java 21 (for BFT-SMaRt)

Deployment

All files for deployment are located in the subdirectory deploy. Deployment scripts are provided for local and cluster deployment (TU Darmstadt DM Cluster). All scripts should be run with the project root as working directory.

Local

For local deployment, use the script deploy/local/run-etcd-local.sh. It will also install dependencies and build the required images before deploying them. After the first usage, if no rebuild is required, flags can be passed to skip rebuilding. The following flags are accepted:

--skip-install      - Skips installing etcd, but still rebuilds the docker image.

--skip-etcd-build   - Skips installing and building etcd.

--skip-da-build     - Skips building the da application and image.

--cluster-size <n>  - Deploys <n> etcd nodes and da instances. (default 2)

Cluster

For cluster deployment, use the script deploy/cluster/deploy-etcd-cluster.sh for etcd and deploy/cluster/deploy-cluster-bftsmart.sh for BFT-SMaRt. It uses the config file deploy/cluster/deployment-cluster.yml to determine the remote hosts to deploy on and the ssh user to use. Make sure the identity file is set up, or if password login is enabled, it needs to be provided during deployment for each host.

Experiments

Checkout branch performance-experiments to

run scripts for performance experiments: produces CSV files with client request timings
deploy etcd cluster or BFT-SMaRt cluster for fault experiments

Last relevant commit before thesis submission for the experiments: a4eca30b8966fc87bcc15ea8b688c5b8cbd1e724

To reproduce the fault experiments from the thesis, make sure install the forked projects instead of the original implementation:

Make sure to change REMOTE_DEPLOYMENT_DIR in deploy/cluster/.env according to your remote directory when using a cluster setup. Also, change the location of the BFT-SMaRt fork in build-bftsmart-client.sh and adjust cluster-specific information, i.e. user and host. For the BFT-SMaRt client, a Dockerfile is provided Dockerfile-bftsmart-client. For etcd, the client is run without docker.

Consensus UI

Checkout branch consensus-ui and run the script deploy/local/run-etcd-education.sh. This deploys the backend locally with an etcd cluster. For the frontend, please refer to https://github.com/FatProteins/consensus-ui.

Last relevant commit before thesis submission for the Consensus UI: e817b1e31f3d672012d0d9d2033bd80d79096a8d

Plotting Scripts

Additional scripts for plot creation can be found in https://github.com/FatProteins/master-thesis-scripts.

fatproteins / master-thesis-code Goto Github PK