advanced high-performance matching engine for financial trading systems

plutus, named after the greek god of wealth, is a robust high-performance matching engine, similar to those found in stock exchanges today. it's loosely inspired by the Jane Street talk on building exchanges, and extends it with a more modern and advanced architecture for real-time order matching, risk management, and trade execution.

plutus supports a wide range of trading scenarios and incorporates advanced order types (limit, market, stop-loss, iceberg), time-in-force (GTC, IOC, FOK) options, and symbol-specific configurations. i tried to emphasize low-latency and high-throughput processing through efficient data structures and a multithreaded API to mimic how production systems work from what I could gather from various blog posts, articles, papers, and discussions.

at its core, plutus is a matching engine that processes, validates, and matches incoming orders against the order book, with support for partial fills, self-trade prevention, and other advanced handling. the order book maintains structures for bids and asks and the event loop manages the API with non-blocking I/O for connections and sockets. it uses a volume-weighted average price algorithm to calculate the average price of a symbol over a period of time.

the client communicates with the server with a rudimentary text-based protocol. eg:

ADD|1|1640995200000|1001|AAPL|150.25|10|BUY|GTC|LIMIT|123|0|0

the library in client/ is a simple order management system wrapper over plutus that serves as a lightweight demo and validation for changes.

i tried to comment parts i found more difficult and write self-documenting code to make it easier to navigate and learn, especially to simplify the design choices i made. make builds the entire project with c++20 with only a kqueue dependency for macos. the build outputs a binary which launches a server.