Load Testing Guide

Load Testing Guide

Guide for conducting load testing of PATAS API.

Quick Start

Basic Load Test

python scripts/load_test.py --url http://localhost:8000 --endpoint /api/v1/health --rps 100 --duration 60

Analyze Endpoint Test

python scripts/load_test.py \
  --url http://localhost:8000 \
  --endpoint /api/v1/analyze \
  --rps 500 \
  --duration 120 \
  --output load_test_500rps.json

Parameters

• --url - API base URL (default: http://localhost:8000)
• --endpoint - API endpoint to test (default: /api/v1/health)
• --rps - Requests per second (default: 100)
• --duration - Test duration in seconds (default: 60)
• --output - Output JSON report file
• --api-key - Optional API key for authentication

Output

JSON report with:

• Total requests, success/error rates
• Latency metrics (P50, P95, P99, avg, min, max)
• Status code distribution
• Error messages (first 10)

Performance Targets

• P95 latency: ≤ 200ms (rules-only), ≤ 700ms (with LLM)
• Error rate: < 1%
• Stable throughput: Maintain target RPS without degradation

Interpreting Results

Good Performance:

• P95 latency within targets
• Error rate < 1%
• Consistent latency across test duration

Issues to Watch:

• P95 latency exceeding targets
• Error rate > 1%
• Increasing latency over time
• High P99 latency

Troubleshooting

High Latency: Check database pool, CPU/memory, slow queries, LLM API response times

High Error Rate: Check API logs, database connectivity, rate limiting, external API status

Connection Errors: Verify API server running, network connectivity, firewall rules