Benchmarking

This repository includes phpbench profiles for common ABAC policy shapes and for degraded latency conditions.

Available benchmark scenarios

• Flat AND chain with multiple resource checks.
• Nested OR chain with two child branches.
• Mixed resource + actor checks.

Bench sources:

• benchmarks/AbacEvaluationBench.php for normal in-process evaluation.
• benchmarks/AbacLatencyEvaluationBench.php for latency-simulated evaluation.

Run benchmarks

From repository root:

composer bench

Quick local feedback:

composer bench:quick

Latency-simulated runs:

composer bench-latency

Quick latency feedback:

composer bench-latency:quick

Notes

• Benchmarks use in-memory SQLite for repeatable local runs.
• The normal suite measures package overhead and stays in the low-ms range.
• The latency suite adds 100 ms of synthetic delay per evaluation to model bad conditions.
• Results are best used for relative comparisons between changes.
• For release comparisons, run on a quiet machine with consistent CPU settings.