Conference | Live Schedule | Talks | Get your ticket | Slides | Set your timezone
Concurrency in Low-Latency Trading Systems
Sarthak Sehgal
On Day 1 at 16:00 (CET/Berlin) in Track D online
Most software can tolerate occasional stalls. In high-frequency trading, a single unexpected wait can mean missing a chunky opportunity. Concurrency in these systems isn’t about maximizing throughput, it’s about shaving off microseconds without introducing tail latency or jitter. In this talk, we’ll explore the real-world synchronization strategies and hardware-aware techniques that help trading systems stay ahead.
We'll dive into why memory alignment matters, how to pin threads for cache locality, when spin-waiting beats locking, and how NUMA architectures affect latency. We’ll go over a lock-free single-producer single-consumer (SPSC) queue tuned for modern CPUs, and explore advanced concurrency patterns like Read-Copy-Update (RCU) for scalable, wait-free reads.
This talk aims to demonstrate real-world usage of common concurrency patterns and is targeted at beginner to intermediate developers who want to sharpen their mental model of what it takes to write C++ that doesn’t wait.
Please login to comment