Docs

Aetower performance harness

Use the local harness to measure the core sampling pipeline cost on the current machine.

Run

./scripts/measure-overhead.sh --iterations 30

Fail the command when the current machine exceeds the default local budget:

./scripts/measure-overhead.sh --iterations 30 --enforce

The output is JSON and includes:

• measured iterations and warmup iterations
• average process count seen per tick
• average entity count produced per tick
• average / p95 / max milliseconds for:
• collection
• identity resolution
• attribution
• friction scoring
• total pipeline time
• resident memory before and after the run
• the default benchmark budget
• pass/fail plus any budget violations

Notes

• This harness measures the core Rust pipeline only.
• It does not include SwiftUI rendering or optional adapter enrichments.
• The harness now warms up the pipeline before timing so the local gate is less sensitive to first-tick cache effects.
• Run it on an idle machine and again while stressing the system to compare overhead envelopes.
• The default local budget currently expects:
• collection average <= 5 ms
• identity average <= 2 ms
• total pipeline p95 <= 10 ms
• resident memory growth <= 24 MiB across the run

For the packaged app itself, use runtime-profiling.md.