Building a Low Latency Risk Engine in C++

Market-data handlers and matching engines get the attention. The risk engine deserves it just as much.

Every order a trading firm sends passes through a risk check first. In most systems that check is an afterthought: a position limit and little else, kept small so it doesn't slow anything down. Yet it is the one component that stands between every order and the market, and its failures are never small. It has to be exactly right and fast, at the same time, while other threads change the data it reads.

This book builds that engine from an empty directory to a tested, measured system in C++23: a lock-free pre-trade and portfolio risk engine for high- and medium-frequency trading. Every order is checked against size, notional, fat-finger and price-collar limits in tens of nanoseconds, with no locks and no heap allocation on the order's path. Positions, resting exposure and Value at Risk stay current on every fill. Stress scenarios run against the live book, and an audit trail records every decision so the past can be rebuilt exactly.

What you'll learn

- Lock-free building blocks (seqlocks, single-producer rings, RCU publication, fixed-capacity containers) and when each is the right tool
- How to time a 50-nanosecond operation honestly, and prove the hot path never allocates
- The finance from first principles: positions, limits, Value at Risk and stress testing, each with a worked example
- Keeping VaR current on every fill, with expensive recomputation moved off the critical thread
- What a risk check should do when it can't answer in time
- Testing systems whose bugs don't crash: reference implementations, randomized input, sanitizers, and breaking the code on purpose
- Measuring what a component really costs a system, from the outside, against a noise floor

Every chapter shows the design that looks right and the input that breaks it, then the design that works and what it costs, with the exact command to reproduce every number.

Who it's for: engineers who read modern C++ and want to understand low-latency design, finance, or both. No background in risk management is assumed.

Included with the book: access to the complete source repository: the engine, a simulated exchange, every test and benchmark, every code listing as a file, and a recorded trading session you can inspect without building anything.

Chapter 1 is available as a free sample.