TL;DR
- ▸Ultra-low latency means different things in different industries — single-digit microseconds in HFT, single-digit milliseconds in 5G URLLC, sub-frame in broadcast IP video.
- ▸The common thread is that the application's value depends on responding to an event as quickly as possible after observing it.
- ▸Precision timing is the substrate underneath any honest latency engineering — without it, you can't measure what you're optimising.
What ULL actually is
Ultra-low latency is a fuzzy term that means different things in different industries. In high-frequency trading it means single-digit microseconds end to end — the time between an order packet entering the trading system and the matching engine processing it. In 5G URLLC it means single-digit milliseconds — the radio-to-radio latency for safety-critical wireless control. In broadcast IP video it means sub-frame latencies — the time between a camera capture and the broadcast feed showing the captured image. In each case the absolute latency targets differ, but the common thread is that the application's value depends on responding to an event as quickly as possible after observing it.
The engineering work to achieve ULL is similar across industries: dedicated low-latency hardware (FPGA-based processing, hardware-timestamped NICs, kernel-bypass networking), optimised software stacks (lock-free data structures, latency-aware scheduling, careful memory layout), and infrastructure that's engineered specifically for the latency requirement rather than for general-purpose throughput.
Where time fits in
Latency measurement requires precision timestamps at both ends of the path being measured. Without PTP-grade synchronisation between the measurement points, the latency you measure is dominated by the clock skew between the measurement clocks rather than by the actual network behaviour. A latency measurement reported as "5 microseconds" means very different things if the measurement clocks are synchronised to 100 nanoseconds versus to 10 microseconds. Precision timing is the substrate underneath any honest latency engineering.
This is one of the underappreciated reasons why ULL applications adopt precision timing infrastructure even when the application itself doesn't strictly need synchronised clocks. The latency optimisation work depends on being able to measure what's being optimised, and the measurement depends on synchronised clocks at the precision tier of the latency target. HFT firms run hardware grandmasters partly for compliance and partly because they couldn't measure their own latency optimisation without them.
Frequently asked questions
What is ultra-low latency?+
Why does ULL need precision timing?+
Related reading
Blog · Finance
Ultra-Low-Latency Trading: Capturing Timestamps in Nanoseconds
How modern HFT systems capture nanosecond-resolution timestamps for every order event, why the precision matters for both regulation and competitive performance, and what the hardware stack actually looks like.
Blog · Finance
High-Frequency Trading: Speed, Fairness and Nanosecond Precision
Why HFT firms are pushing the precision floor of clock synchronisation into single-digit nanoseconds, what venue parity actually means at that precision, and where the next competitive frontier lives.
Blog · Standards
Understanding IEEE 1588 PTP: How Precision Time Powers Industrial Ethernet
What IEEE 1588 actually defines, how the protocol works at the message level, and why it's the foundation under every modern industrial Ethernet, telecom and broadcast timing fabric.