Open Source Tools
The Microseconds You're Losing Are Not Mysterious.
We don't guess. We measure. Production-grade tools for HFT infrastructure validation and latency profiling.
Built from HFT experience
Production-tested
MIT Licensed
"One misconfigured sysctl cost a trading desk 340μs per trade"
---
- [x] Swappiness (should be 0)
- [x] Transparent Hugepages (should be `never`)
- [x] Kernel preemption model
- [x] Frequency Governor (should be `performance`)
- [x] C-States (should be disabled)
- [x] Core Isolation (`isolcpus` configuration)
pip install latency-audit ## See Also
- eBPF tracing with <1% overhead
- No code instrumentation required
- Percentile breakdowns (p50, p99, p999)
- Export to flame graphs
pip install latencyscope Why These Exist
After seeing the same misconfigurations at three different trading firms, I automated the audit process. latency-audit is the tool I wish I had on day one.
latencyscope came from debugging a 47μs spike that only appeared under production load. Traditional profilers added too much overhead. eBPF was the only way.