Skip to content

Infrastructure Audit · Latency Profiling

Signing Latency Audit Template. 14 measurements most shops have never taken

Fill in your p50/p95/p99 for each measurement under realistic production load, not synthetic benchmarks. Compare against the industry targets on the right.

How to use this template

Run each measurement with your actual signing key type (KMS / HSM / TEE / software) under a load profile matching your peak production throughput. Record p50, p95, and p99 separately. Averages hide tail latency. Any p99 over 1ms warrants immediate investigation.

1 Signing path 3 measurements

Measurement Your result Target

2 Key management overhead 3 measurements

Measurement Your result Target

3 Execution loop 3 measurements

Measurement Your result Target

4 Kernel path 4 measurements (often never measured)

Measurement Your result Target

5 Network path 2 measurements

Measurement Your result Target

Your overall p99 signing latency

[total wall-clock p99] µs

If your signing p99 exceeds 1ms, your key management choice is costing you alpha on every trade. AWS KMS baseline is ~130ms. CloudHSM is ~5ms. TEE (ZeroCopy pattern) is 42µs.

If your p99 is worse than 1ms, book a call . calendar.app.google/W1CEqo8GsoGtjJX49 · or see the prop shop engagement page

Nikhil Padala · Ex-Gemini SRE · Ex-Akuna Capital · ZeroCopy Systems