ai-agent/benchmarks/performance_metrics.md

Performance Benchmark Metrics

Overview

This document describes the performance benchmark suite for core system components. The benchmarks measure:

• Task dispatcher throughput (tasks/second)
• RBAC authorization latency (milliseconds)
• SQLite CRUD operation performance (operations/second)

Running Benchmarks

pytest tests/performance/benchmarks.py -v --benchmark-enable --benchmark-json=benchmarks/results.json

Interpreting Results

Performance metrics are logged to metrics/api_performance.log with timestamps.

Key Metrics

Component	Metric	Target	Unit
TaskDispatcher	Throughput	≥1000	tasks/sec
RBACEngine	Auth Latency	≤5	ms
SQLiteAdapter	INSERT	≤2	ms/op
SQLiteAdapter	SELECT	≤1	ms/op

Baseline Targets

These targets are based on system requirements:

1. Task Dispatcher

   • Must handle ≥1000 tasks/second under load
   • 95th percentile latency ≤10ms

2. RBAC Authorization

   • Average check time ≤5ms
   • 99th percentile ≤10ms

3. SQLite Operations

   • INSERT: ≤2ms average
   • SELECT: ≤1ms average for simple queries
   • Complex queries (joins): ≤10ms average

Performance Trends

Performance metrics are tracked over time in metrics/api_performance.log. Use this command to analyze trends:

grep "TaskDispatcher throughput" metrics/api_performance.log

Troubleshooting

If benchmarks fail to meet targets:

1. Check system resource usage during tests
2. Review recent code changes affecting components
3. Compare with historical data in performance logs