Lock-Free Limit Order Book

Live Demo: lob-engine-lock-free.vercel.app

Overview

A production-grade C++20 limit order book and matching engine built around lock-free data structures, targeting sub-microsecond P99 matching latency with zero heap allocation on the critical path.

Technical Highlights

• Matching Engine: Engineered a C++20 price-time priority matching engine processing 2.6M+ orders/sec via a lock-free MPSC queue (Vyukov exchange) and wait-free SPSC ring buffer (Lamport), achieving sub-microsecond P99 matching latency with zero heap allocation on the hot path.

• Cache-Line Isolation: Eliminated cross-core false sharing using alignas(64) cache-line isolation on all concurrent structs (Order pool, queue heads, EventLog slots), reducing thread contention across MPSC producer threads.

• Pre-Trade Risk Engine: Implemented atomic net-position tracking at ~5ns overhead per order, enforcing per-order notional, quantity, and max long/short limits without a single mutex on the critical path.

• Zero-Allocation EventLog: Architected a zero-allocation LMAX Disruptor-style EventLog (64K-slot ring buffer) enabling lock-free audit trail with single atomic::store writes and independent multi-reader cursors, decoupling the matching core from all downstream consumers.

• NASDAQ ITCH 5.0 Parser: Parsed NASDAQ TotalView-ITCH 5.0 binary protocol (big-endian, A/D/E/U/X message types) to replay real exchange data through the engine, validating IOC/FOK/GTC order semantics against production-grade market feeds.

• Live Trading Dashboard: Built a React/TypeScript trading dashboard streaming live order book depth, fill tape, VWAP, and book imbalance at 60Hz over a self-implemented RFC 6455 WebSocket server with zero external C++ dependencies.

Key Results