FBAS Reward Distributor

Framework for the computation of a node's influence and reward distribution in a Federated Byzantine Agreement Systems (FBASs) like Stellar.

The framework is suitable for usage as both a binary and a library and can mainly do the following:

• read node data in stellarbeat's JSON format;
• rank the nodes using a variety of implemented algorithms;
• based on the rankings, compute a reward distribution for each node.

Required toolchain

• A working Rust environment

  • Install: https://www.rust-lang.org/tools/install

• A minimal gcc and g++ toolchain is required by some of the dependencies. Should be covered by the build-essential, m4 packages.

Build and optionally run tests

1. Build:

cargo build --release

2. Tests

cargo test --release

Usage as a binary

1. Command line arguments

cargo run --release -- {distribute | rank} [-i -p -r reward] <fbas-path> {node-rank|power-index-approx|power-index-enum}

    - fbas-path: Path to file describing the FBAS.
        If no path is passed, the program will attempt to read from the command line.
    - i: Ignore inactive nodes in the FBAS. Optional. Default = false.
    - r reward: reward value that is to be distributed - only used with the rank subcommand. Default = 1.
    - p: Include the nodes' public keys in the output. Default = false.

The rank subcommand is similar to distribute with the exception that it only calculates the nodes' weights without allocating rewards. The output is always a sorted list of tuples: (NodeID, Public Key (where available), Ranking, [Reward]).

2. Compute a reward distribution for the nodes in the mobilecoin_nodes_2021-10-22.json FBAS using

   1. the Shapley-Shubik power index

          cargo run --release -- rank test_data/mobilecoin_nodes_2021-10-22.json power-index-enum
      

      This algorithm computes the players' Shapley-Shubik indices via enumeration in O(2^n) time, and is therefore not recommended for larger FBASs.

   2. As an alternative, we provide a polynomial time approximation algorithm using Castro et al.'s algorithm based on sampling.

      cargo run --release -- rank test_data/mobilecoin_nodes_2021-10-22.json power-index-approx 1000
      

   3. Distributions can also be computed based on a graph-theoretic (NodeRank) metric:

      cargo run --release -- rank test_data/mobilecoin_nodes_2021-10-22.json node-rank
      
      

Usage as a library

[dependencies]
fbas-reward-distributor = { git = "https://github.com/cndolo/fbas-reward-distributor", default-features = true}

See the fbas-graph-generator for some examples.

Performance and approximation measurements

1. Build with

cargo build --release --features "measurements"

2. then run performance measurements using the selected ranking algorithm

target/release/performance_tests -m $MAX_TOP_TIER --no-quorum-intersection -r $ITERATIONS -o $OUTPUT_FILE -j $JOBS -u $FBAS_TYPE $RANKING_ALGO

3. and/or approximation measurements

target/release/approximation_tests -m $MAX_TOP_TIER --no-quorum-intersection -r $ITERATIONS -o $OUTPUT_FILE -j $JOBS -u $FBAS_TYPE