Building a Financial Exchange from first principles

18 hours ago

Speed and simplicity rarely come together, but farzi.exchange aims to achieve both in a fake stock exchange built from first principles.
The core of a financial exchange is to match orders via an Order Book and a Matching Algorithm, with rules ensuring determinism and fairness.
farzi.exchange is a price-time exchange, matching orders based on price and time, with tie-breaks considering disclosed quantity or order size.
Key order types include Limit Orders, Market Orders, and Iceberg Orders, each serving different trading purposes and scenarios.
The exchange architecture includes a Matching Engine with parallel goroutines per instrument, an in-memory Order Book using Binary Trees for speed.
An Emitter fans out events to consumers, an Aggregator creates candles from trades, and a Persistence Layer stores data asynchronously in PostgreSQL.
A MarketData Layer serves real-time data via websockets and historical data via APIs, using CQRS separation for scalability.
Market Makers with personas (Farzi Maker, Seth, Scalper, Gambler) provide liquidity and price movement in the absence of real traders.
The Farzi 10 Index is computed from underlying instrument values, updated on each trade and sampled at 100ms intervals.
The exchange processed over 28 million trades and 122 million orders in a day, hosted on AWS with Vercel for the client, showing significant scalability.

Hasty Briefsbeta