Trust Layer Benchmarks – Baseline (Q4 2025)¶

Generated via:

cargo bench --bench trust_layer_bench -- --sample-size=20 --warm-up-time=1 --measurement-time=5

Summary¶

Benchmark	Mean	Std Dev	Throughput	Notes
`registry/record_archive/100`	0.92 s	0.15 s	109 ops/s	JSON snapshot on tmpfs (100 entries per iter)
`proof/prove_loss_delta`	1.32 µs	0.12 µs	0.76 M ops/s	Deterministic hash-based placeholder proof
`proof/verify_loss_delta`	1.01 µs	0.08 µs	0.99 M ops/s	Receipt hash check + serde decode
`ipfs/upload_bytes`	209 µs	9 µs	4.8 k ops/s	Mock HTTP server (no TLS)
`end_to_end/prove+upload`	242 µs	19 µs	4.1 k ops/s	Prove + upload pipeline

Raw Criterion Outputs¶

target/criterion/registry/record_archive/100/new/estimates.json
target/criterion/proof/prove_loss_delta/new/estimates.json
target/criterion/proof/verify_loss_delta/new/estimates.json
target/criterion/ipfs/upload_bytes/new/estimates.json
target/criterion/end_to_end/prove+upload/new/estimates.json

Machine-readable baselines live in docs/performance/baseline_trust_layer.json.
The CI pipeline runs tools/check_bench_regressions.py and fails if any mean drifts by more than 20%.