search

dqian3 / DOM-BFT

MIT License
0 stars 0 forks source link

readme

DOM-BFT

Repository for WIP project on applying Deadline Ordered Multicast to a BFT protocol.

Note this repo is based off of Nezha.

Installing Bazel

I used Bazelisk as a wrapper for Bazel, which allows for us to specify a version of Bazel to use (5.2.0). The way third party dependencies are setup in this repo means that more recent versions of Bazel are not compatible.

Building Repository

To build the client

bazel build //processes/client/...

The same syntax is used to build the proxy, receiver and other components. You can also build all of them at the same time with something like

bazel build //processes/...

Running

The built executables will be created by default in bazel-bin, inside of which is the same directory structure as the source, containing the executables.

The /scripts directory also contains a number of helpful scripts for running the executables both locally and remotely. See the README there for more details.

Tests

Tests exist for this code, though I make no guarantees about how comprehensive they are. The tests can be built with 

bazel build //test/...

The files starting with test_ are gtest files for various components that I've built.

Basic Docker Setup

This repo includes a super basic Docker setup which can be used for development. Future work is planned to make it more realistic for testing and deployment.

The current way to use this setup is to simply run

docker compose run --rm dev

This will give you a bash shell in a conatiner where you can run the above commands to compile the code. Docker compose mounts the source code into the container as well, so any changes you make locally (i.e. in an editor) will be reflected in your conatiner.

Profiling

To get CPU profiles of the processes, we use the gperftools CPU profiler. The steps for setting this up are:

  1. Build and install the repository https://github.com/gperftools/gperftools on the machine you plan to run the profile on

  2. Follow the instructions here: https://gperftools.github.io/gperftools/cpuprofile.html to run the process with the profiler. Specifically, append env LD_PRELOAD='.../libprofiler.so' CPUPROFILE=profile.prof to the command.

    • Note it isn't reccomended in the above link to do this runtime linking. We should probably figure out how to incoporate this library into our build
    • Also, the invoke gcloud-run command includes a --profile argument that adds this to processes already, for running on the cloud.

  3. Visualize the profile results with the pprof tool: https://github.com/google/pprof.

    pprof -web [main_binary] profile.prof