Why Trading Time Is Different: TSC, CLOCK_MONOTONIC, HPET, and the Lies Each Tells
A practitioner's guide to clock sources in trading systems: TSC drift, CLOCK_MONOTONIC overhead, HPET calibration, and why TAI beats CLOCK_REALTIME for audit logs.
EU AI Act Article 14: 90 days until enforcement. 2 Q2 2026 slots left. Check compliance →
PTP, TSC, and the lies your p99 tells you
Clock synchronization for trading: TSC drift, PTP/IEEE 1588 with Solarflare, MiFID II RTS-25 compliance, and coordinated omission — the measurement artifact that hides your real tail latency.
You cannot optimize what you cannot measure correctly. Coordinated omission routinely makes systems look 40x faster than they are. MiFID II mandates ±1µs clock accuracy. These five posts build the measurement discipline needed before any latency optimization work.
A practitioner's guide to clock sources in trading systems: TSC drift, CLOCK_MONOTONIC overhead, HPET calibration, and why TAI beats CLOCK_REALTIME for audit logs.
How I implemented PTP hardware timestamping on Solarflare SFN8522 at Gemini to achieve ±30ns clock sync and sub-100µs MiFID II RTS-25 compliance.
MiFID II RTS-25 timestamp requirements explained for practitioners, with the full implementation pipeline I built at Upside and what it means for UAE/DIFC crypto firms.
Gil Tene's coordinated omission problem explained with real numbers: how a naive benchmark showed 50µs P99 when the actual P99 was 2ms, and how to fix it with HDRHistogram.
How to systematically engineer sub-100µs P99.9 latency for always-on crypto trading: CPU isolation, GC elimination, request hedging, thermal throttling, and ZeroCopy's Nitro Enclave signing numbers.