Sub-1% Overhead Tail-Latency Histograms With Per-CPU eBPF Maps
Why user-space OpenTelemetry spans miss kernel socket-buffer queueing delays, and how log-linear BPF per-CPU arrays capture every microsecond stall at 420,000 requests per second.
Complete Essay Index
Published Essays
Production post-mortems, kernel telemetry studies, and reproducible hardware benchmarks—filterable by topic or keyword.
- Articles
- 8
- Topics
- 6
- Latest
- Oct 28, 2026