🧮 arb-solver

arb-solver is a Rust engine for efficiently detecting profitable arbitrage cycles in directed weighted graphs. It uses an optimized Bellman–Ford variant (SPFA) for fast negative cycle detection, essential for finding arbitrage opportunities where the product of edge weights (rates) exceeds one (∏ rᵢ > 1).

The project also employs a Compressed Sparse Row (CSR) graph representation, ensuring memory and computational efficiency.

---

📦 Project Structure

arb-solver is a multi-crate workspace, designed for modularity and performance analysis.

Crate Name	Description
core	Implements the main logic, including the CSR graph structure and the SPFA arbitrage detection algorithm.
common	Provides shared utilities used across other crates.
executor	Handles project execution, input parsing (CSV or simulation), and coordinates the core logic via an async pipeline.
perf-bench	Dedicated crate for benchmarking different data layouts, such as Array-of-Structs (AoS) vs Struct-of-Arrays (SoA).
[root-level workspace]	Top-level workspace that aggregates and manages the other crates.

---

🔨 Build and Usage

Clone & Build the Repository

git clone git@github.com:0xphen/arb-solver.git
cd arb-solver

cargo build

---

🧪 Running Tests

Execute unit tests across all crates:

cargo test

---

📊 Running Benchmarks

Compare performance of different data layouts with the perf-bench crate:

Layout	Command
Struct-of-Arrays (SoA)	cargo run --release --bin bench_soa -p perf-bench
Array-of-Structs (AoS)	cargo run --release --bin bench_aos -p perf-bench

---

🚀 Running the Executor

The executor crate can be run in two main modes:

1️⃣ Simulation Mode

Runs the executor using a randomly generated internal graph :

cargo run --release -p executor -- sim

2️⃣ CSV Input Mode

Runs the executor using a graph provided via a parsed local CSV file.

The CSV file must be formatted as:

from,to,rate

Example:

0,1,0.92
1,2,150.5
2,0,0.0074

Run the executor:

cargo run --release -p executor -- csv <path_to_csv_file>

⚠️ Important: The Config.toml file configures various aspects of the executor system, including the searcher, writer, simulator, producer, and executor. It controls batch sizes, processing intervals, backpressure behavior, and simulation parameters, making it the central configuration for the system in all modes—including CSV input and simulation.